#74 – Michael I. Jordan: Machine Learning, Recommender Systems, and the Future of AI

[0] The following is a conversation with Michael I. Jordan, a professor of Berkeley and one of the most influential people in the history of machine learning, statistics, and artificial intelligence.

[1] He has been cited over 170 ,000 times, and he has mentored many of the world -class researchers defining the field of AI today, including Andrew Eng, Zubing Garamani, Bantascar, and Yoshio.

[2] NGO.

[3] All this, to me, is as impressive as the over 32 ,000 points in the six NBA championships of the Michael J. Jordan of basketball fame.

[4] There's a non -zero probability that I talk to the other Michael Jordan given my connection to and love with the Chicago Bulls of the 90s, but if I had to pick one, I'm going with the Michael Jordan of statistics and computer science.

[5] Or as Yon Lacoon calls him the Miles Davis of machine learning.

[6] In his blog post titled Artificial Intelligence, The Revolution Hasn't Happened Yet, Michael argues for broadening the scope or the artificial intelligence field.

[7] In many ways, the underlying spirit of this podcast is the same, to see artificial intelligence as a deeply human endeavor, to not only engineer algorithms and robots, but to understand and empower human beings at all levels of abstractions, from the individual to our civilization as a whole.

[8] This is the Artificial Intelligence Podcast.

[9] If you enjoy it, subscribe and YouTube, give it five stars at Apple Podcasts, support it on Patreon, or simply connect with me on Twitter, at Lex Friedman, spelled F -R -I -D -M -A -N.

[10] As usual, I'll do one or two minutes of ads now and never any ads in the middle that can break the flow of the conversation.

[11] I hope that works for you and doesn't hurt the listening experience.

[12] This show is presented by Cash App, the number one finance app in the app store.

[13] When you get it, use code Lex Podcast.

[14] Cash app lets you send money to friends buy Bitcoin and invest in the stock market with as little as $1.

[15] Since Cash App does fractional share trading, let me mention that the order execution algorithm that works behind the scenes to create the abstraction of the fractional orders is to me an algorithmic marvel.

[16] So big props for the cash app engineers for solving a hard problem that in the end provides an easy interface that takes a step up to the next layer of abstraction over the stock market, making trading more accessible for new investors and diversification much easier.

[17] So once again, if you get Cash App from the App Store or Google Play and use the code Lex podcast, you'll get $10 and Cash App will also donate $10.

[18] is the first, one of my favorite organizations that is helping to advance robotics and STEM education for young people around the world.

[19] And now, here's my conversation with Michael I. Jordan.

[20] Given that you're one of the greats in the field of AI, machine learning, computer science, and so on, you're trivially called a Michael Jordan of machine learning.

[21] Although, as you know, you were born first.

[22] So technically, MJ is the Michael I. Jordan of basketball.

[23] But anyway, my favorite is Jan Lacoon calling you the Miles Davis of machine learning, because as he says, you reinvent yourself periodically and sometimes leave fans scratching their heads after you change direction.

[24] So can you put at first your historian hat on and give a history of computer science and AI as you saw it, as you experienced it, including the four generations of AAS successes that I've seen you talk about.

[25] Sure.

[26] Yeah, first of all, I much prefer Yon's metaphor.

[27] Miles Davis was a real explorer in jazz, and he had a coherent story.

[28] So I think I have one, but it's not just the one you lived.

[29] It's the one you think about later.

[30] What a good historian does is they look back and they revisit.

[31] I think what happening right now is not AI.

[32] That was an intellectual aspiration that's still alive today as an aspiration.

[33] But I think this is akin to the development of chemical engineering from chemistry or electrical engineering from electromagnetism.

[34] So if you go back to the 30s or 40s, there wasn't yet chemical engineering.

[35] There was chemistry.

[36] There was fluid flow.

[37] There was mechanics and so on.

[38] But people pretty clearly viewed interesting goals to try to build factories that make chemicals, products, and do it viably, safely, make good ones, do it at scale.

[39] So people started to try to do that, of course, and some factories worked, some didn't, some were not viable, some exploded.

[40] But in parallel, developed a whole field called chemical engineering.

[41] And chemical engineering is a field.

[42] It's no bones about it.

[43] It has theoretical aspects to it.

[44] It has practical aspects.

[45] It's not just engineering, quote, unquote.

[46] It's the real thing, real concepts were needed.

[47] Same thing with electrical engineering.

[48] There was Maxwell's equations, which in some sense were everything you know about electrical magnetism, but you needed to figure out how to build circuits, how to build modules, how to put them together, how to bring electricity from one point to another, safely, and so on and so forth.

[49] So a whole field is developed called electrical engineering.

[50] I think that's what's happening right now, is that we have a proto field, which is statistics, computer more of the theoretical side of the algorithmic side of computer science that was enough to start to build things but what things systems that bring value to human beings and use human data and mix in human decisions the engineering side of that is all ad hoc that's what's emerging in fact if you want to call machine learning a field i think that's what it is that's a proto form of engineering based on statistical and computational ideas in previous generations but do you think there's something deeper about AI in his dreams and aspirations as compared to chemical engineering engineering and electrical engineering.

[51] Well, the dreams and aspirations may be, but those are from, those are 500 years from now.

[52] I think that that's like the Greek sitting there and saying, it would be neat to get to the moon someday.

[53] Right.

[54] I think we have no clue how the brain does computation.

[55] We're just a clueless.

[56] We're like, we're even worse than the Greeks on most anything interesting, scientifically of our era.

[57] Can you linger on that just for a moment because you stand not completely unique, but a little bit unique in the clarity of that.

[58] Can you elaborate your intuition of why we, like where we stand in our understanding of the human brain?

[59] And a lot of people say, neuroscientists say we're not very far in understanding human brain.

[60] But you're like, you're saying we're in the dark here.

[61] Well, I know I'm not unique.

[62] I don't even think in the clarity.

[63] But if you talk to real neuroscientists that really study real synapses or real neurons, they agree.

[64] They agree.

[65] It's a hundred years of year task and they're building it up slowly, surely.

[66] What the signal is there is not clear.

[67] We think we have all of our metaphors.

[68] We think it's electrical.

[69] Maybe it's chemical.

[70] It's a whole soup.

[71] It's ions and proteins and it's a cell.

[72] And that's even around like a single synapse.

[73] If you look at a electromagnograph of a single synapse, it's a city of its own.

[74] And that's one little thing on a dendritic tree, which is extremely complicated, you know, electrochemical thing.

[75] And it's doing these spikes and voltages are being flying around.

[76] And then proteins are taking that and taking it down into the DNA and who knows what.

[77] So it is the problem.

[78] of the next few centuries.

[79] It is fantastic.

[80] But we have our metaphors about it.

[81] Is it an economic device?

[82] Is it like the immune system?

[83] Or is it like a layered, you know, set of, you know, arithmetic computations?

[84] We have all these metaphors and they're fun.

[85] But that's not real science per se.

[86] There is neuroscience.

[87] That's not neuroscience.

[88] All right.

[89] That's like the Greek speculating about how to get to the moon.

[90] Fun, right?

[91] And I think that I like to say this fairly strongly because I think a lot of young people think we're on the verge.

[92] Because a lot of people who don't talk about it clearly, let it be understood that, yes, we kind of, this is brain inspired, we're kind of close, you know, breakthroughs are on the horizon.

[93] And unscrupulous people sometimes who need money for their labs, as I'm saying unscrupulous, but people will oversell.

[94] I need money for my lab.

[95] I'm going to, I'm studying, you know, computational neuroscience.

[96] I'm going to oversell it.

[97] And so there's been too much of that.

[98] So I'll step into the gray area between metaphor and engineering with, I'm not sure if you're familiar with brain computer interfaces.

[99] So a company like Elon Musk has NeuroLink that's working on putting electrodes into the brain and trying to be able to read both read and send electrical signals.

[100] Just as you said, even the basic mechanism of communication in the brain is not something we understand.

[101] But do you hope without understanding the fundamental principles of how the brain works will be able to do something interesting at that gray area of metaphor?

[102] It's not my area.

[103] So I hope in this sense like anybody else hopes for some interesting things to happen from research.

[104] I would expect more something like Alzheimer's will get figured out from modern neuroscience.

[105] There's a lot of human suffering based on brain disease.

[106] And we throw things like lithium at the brain.

[107] It kind of works.

[108] No one has a clue why.

[109] That's not quite true, but, you know, mostly we don't know.

[110] And that's even just about the biochemistry of the brain and how it leads to mood swings and so on.

[111] How thought emerges from that, we just, we were really, really, completely dim.

[112] So that you might want to hook up electrodes and try to do some signal processing on that and try to find patterns, fine, you know, by all means go for it.

[113] It's just not scientific at this point.

[114] It's just, it's like kind of sitting in a satellite and watching the emissions from a city and trying to infer things about the microeconomy, even though you don't have microeconomic concepts.

[115] I mean, it's really that kind of thing.

[116] And so, yes, can you find some signals that do something interesting or useful?

[117] Can you control a cursor or mouse with your brain?

[118] Yeah, absolutely.

[119] And I can imagine business models based on that and even medical applications of that.

[120] But from there to understanding algorithms that allow us to really tie in deeply from the brain to the computer.

[121] No, I don't agree with Elon Musk.

[122] think that's even, that's not for our generation, it's not even for the century.

[123] So just in the hopes of getting you to dream, you've mentioned Kolmogrov and touring might pop up.

[124] Do you think that there might be breakthroughs that will get you to sit back in five, ten years and say, wow?

[125] Oh, I'm sure there will be, but I don't think that there'll be demos that impress me. I don't think that having a computer call a restaurant and tend to be a human, is a breakthrough.

[126] And people, you know, some people present it as such.

[127] It's imitating human intelligence.

[128] It's even putting coughs in the thing to make a bit of a PR stunt.

[129] And so fine, the world runs on those things too.

[130] And I don't want to diminish all the hard work and engineering that goes behind, things like that, and the ultimate value to the human race.

[131] But that's not scientific understanding.

[132] And I know the people that work on these things, they are after scientific.

[133] understanding.

[134] You know, in the meantime, they've got to kind of, you know, the trains got to run, and they got mouths to feed, and they got things to do.

[135] And there's nothing wrong with all that.

[136] I would call that, though, just engineering.

[137] And I want to distinguish that between an engineering field like electrical engineering chemistry that originally emerged, that had real principles, and you really know what you're doing, and you had a little scientific understanding, maybe not even complete.

[138] So it became more predictable, and it was really gave value to human life because it was understood.

[139] And so we have to, we don't want to muddle too much these waters of what we're able to do versus what we really can do in a way that's going to impress the next.

[140] So I don't need to be wowed, but I think that someone comes along in 20 years, a younger person who's absorbed all the technology, and for them to be wowed, I think they have to be more deeply impressed.

[141] A young Kolmogorov would not be wowed by some of the stunts that you see right now coming from the big companies.

[142] The demos, but do you think the breakthroughs from Kualmogarov would be, and give this question a chance, Do you think they'll be in the scientific fundamental principles arena, or do you think it's possible to have fundamental breakthroughs in engineering?

[143] Meaning, you know, I would say some of the things that Elon Musk is working with SpaceX and then others, sort of trying to revolutionize the fundamentals of engineering, of manufacturing, of saying, here's a problem, we know how to do a demo of, and actually taking it the scale.

[144] Yeah, so there's going to be all kinds of breakthroughs.

[145] I just don't like that terminology.

[146] I'm a scientist, and I work on things day in and day out and things move along and eventually say, wow, something happened, but I don't like that language very much.

[147] Also, I don't like to prize theoretical breakthroughs over practical ones.

[148] I tend to be more of a theoretician, and I think there's lots to do in that arena right now.

[149] And so I wouldn't point to the Colmogoras.

[150] I might point to the Edisons of the era, and maybe Musk is a bit more like that.

[151] But, you know, Musk, God bless him, also will say things about AI that he knows very little about.

[152] And he leads people astray when he talks about things he doesn't know anything about.

[153] Trying to program a computer to understand natural language, to be involved in a dialogue we're having right now.

[154] It can happen in our lifetime.

[155] You could fake it.

[156] You can mimic, sort of take old sentences that humans use and retread them.

[157] With the deep understanding of language, no, it's not going to happen.

[158] And so from that, you know, I hope you can perceive that the deeper yet deeper kind of aspects and intelligence are not going to happen.

[159] Now, well, there'll be breakthroughs.

[160] I think that Google was a breakthrough.

[161] I think Amazon is a breakthrough.

[162] I think Uber is a breakthrough, that bring value to human beings at scale in brand new, brand new ways based on data flows and so on.

[163] A lot of these things are slightly broken because there's not a kind of an engineering field that takes economic value in context of data and at planetary scale and worries about all the externalities, the privacy.

[164] We don't have that field, so we don't think these things through very well.

[165] But I see that is emerging in, and that will be, you know, looking back from 100 years, that will be a constitute a breakthrough in this era, just like electrical engineering was a breakthrough in the early part of the last century, and chemical engineering was a breakthrough.

[166] So the scale, the markets that you talk about, and we'll get to, will be seen as sort of breakthrough, and we're in the very early days of really doing interesting stuff there.

[167] And we'll get to that, but it's just taking a quick step back.

[168] Can you give, we kind of threw off the historian hat?

[169] I mean, you briefly said that the history of AI kind of mimics the history of chemical engineering, but...

[170] I keep saying machine learning, you keep on to say AI, just to let you know I don't, you know, I'd resist that.

[171] I don't think this is about AI really was John McCarthy as almost a philosopher saying, wouldn't it be cool if we could put thought in a computer, if we could mimic the human capability to think or put intelligence in in some sense into a computer?

[172] that's an interesting philosophical question and he wanted to make it more than philosophy he wanted to actually write down logical formula and algorithms that would do that and that is a perfectly valid reasonable thing to do that's not what's happening in this era so so the reason I keep saying AI actually and I'd love to hear what you think about it machine learning has a very particular set of methods and tools maybe your version of it is mine doesn't no it doesn't very very open and it does optimization, it does sampling, it does...

[173] So systems that learn is what machine learning is?

[174] Systems that learn and make decisions.

[175] And make decisions.

[176] So it's not just pattern recognition and, you know, finding patterns.

[177] It's all about making decisions in real worlds and having close feedback loops.

[178] So something like symbolic AI, expert systems, reasoning systems, knowledge -based representation, all of those kinds of things, search.

[179] Does that neighbor fit into what you think of as machine learning?

[180] So I don't even like the word machine learning.

[181] I think that with the field you're talking about, about is all about making large collections of decisions under uncertainty by large collections of entities.

[182] Yes.

[183] Right?

[184] And there are principles for that at that scale.

[185] You don't have to say the principles are for a single entity that's making decisions, single agent or a single human.

[186] It really immediately goes to the network of decisions.

[187] Is it a good word for that or no?

[188] No, there's no good words for any of this.

[189] That's kind of part of the problem.

[190] So we can continue the conversation to use AI for all that.

[191] I just want to kind of raise the flag here that this is not about we don't know what intelligence is and real intelligence we don't know much about abstraction and reasoning at the level of humans we don't have a clue we're not trying to build that because we don't have a clue eventually it may emerge they'll make i don't know if there'll be breakthroughs but eventually we'll start to the glimmers of that it's not what's happening right now okay we're taking data we're trying to make good decisions based on that we're trying to at scale we're trying to economically viably we're trying to build markets we're trying to keep value at that scale And aspects of this will look intelligent.

[192] They will look, computers were so dumb before, they will seem more intelligent.

[193] We will use that buzzword of intelligence, so we can use it in that sense.

[194] But, you know, so machine learning, you can scope it narrowly as just learning from data and pattern recognition.

[195] But whatever, when I talk about these topics, maybe data science is another word you could throw in the mix.

[196] It really is important that the decisions are as part of it.

[197] It's consequential decisions in the real world.

[198] or am I going to have a medical operation?

[199] Am I going to drive down this street?

[200] Things that were their scarcity, things that impact other human beings or other, you know, the environments and so on.

[201] How do I do that based on data?

[202] How do I do that?

[203] How do I use computers to help those kind of things go forward?

[204] Whatever you want to call that.

[205] So let's call it AI.

[206] Let's agree to call it AI.

[207] But it's not say that what the goal of that is is intelligence.

[208] The goal of that is really good working systems at planetary scale that we've never seen before.

[209] So reclaim the word AI from the Dartmouth.

[210] conference from many decades ago of the dream of human I don't want to reclaim it I want a new word I think it was a bad choice I mean I you know if you read one of my little things um the history was basically that McCarthy needed a new name because cybernetics already existed and he didn't like you know no one really liked norbert veneer it was kind of an island to himself and he felt that he had encompassed all this and in some sense he did you look at the language of cybernetics it was everything we're talking about it was control theory and signal processing and some notions of intelligence and closed feedback loops and data.

[211] It was all there.

[212] It's just not a word that lived on partly because of maybe the personalities.

[213] But McCarthy needed a new word to say, I'm different from you.

[214] I'm not part of your show.

[215] I got my own.

[216] Invented this word.

[217] And again, as a kind of thinking forward about the movies that would be made about it, it was a great choice.

[218] But thinking forward about creating a sober academic and world world discipline, it was a terrible choice because it led to promises that are not true, that we understand.

[219] We understand our, perhaps, but we don't understand intelligence.

[220] It's a small tangent because you're one of the great personalities of machine learning, whatever the heck you call the field.

[221] Do you think science progresses by personalities or by the fundamental principles and theories and research that's outside of personalities?

[222] Yeah, both.

[223] And I wouldn't say there should be one kind of personality.

[224] I have mine and I have my preferences and I have a kind of network around me that feeds me and some of them agree with me and some of disagree, but all kinds of personalities.

[225] needed.

[226] Right now, I think the personality that it's a little too exuber and a little bit too ready to promise the moon is because a little bit too much in ascendance.

[227] And I do think that there's some good to that.

[228] It certainly attracts lots of young people to our field.

[229] But a lot of those people come in with strong misconceptions and they have to then unlearn those and then find something to do.

[230] And so I think there's just got to be some multiple voices and I didn't, I wasn't hearing enough of the more sober voice.

[231] So as a continuation, of a fun tangent and speaking of vibrant personalities, what would you say is the most interesting disagreement you have with Jan Lacoon?

[232] So, Jan's an old friend, and I just say that I don't think we disagree about very much, really.

[233] He and I both kind of have a, let's build that kind of mentality and does it work, a kind of mentality, and kind of concrete.

[234] We both speak French, and we French more together and we have we have a lot of a lot in common and so you know if one wanted to highlight a disagreement it's not really a fundamental one I think it's just kind of where we're emphasizing Jan has emphasized pattern recognition and has emphasized prediction all right so you know and it's interesting to try to take that as far as you can if you could do perfect prediction what would that give you kind of as a thought experiment and I think that's way too limited.

[235] We cannot do perfect prediction.

[236] We will never have the data sets allow me to figure out what you're about ready to do.

[237] What question you're going to ask next?

[238] I have no clue.

[239] I will never know such things.

[240] Moreover, most of us find ourselves during the day in all kinds of situations we had no anticipation of that are kind of various, various, that are novel in various ways.

[241] And in that moment, we want to think through what we want.

[242] And also there's going to be market forces acting on us.

[243] I'd like to go down that street, but now it's full because there's a crane in the street.

[244] I got to think about that.

[245] I got to think about what I might really want here.

[246] And I got to sort of think about how much it costs me to do this action versus this action.

[247] I got to think about the risks involved.

[248] A lot of our current pattern recognition and prediction systems don't do any risk evaluations.

[249] They have no error bars, right?

[250] I got to think about other people's decisions around me. I've got to think about a collection of my decisions.

[251] Even just thinking about like a medical treatment, you know, I'm not going to take the prediction of a neural net about my health, about something consequential.

[252] I'm about ready to have a heart attack because.

[253] some number is over 0 .7.

[254] Even if you had all the data in the world, they've ever been collected about heart attacks, better than any doctor ever had, I'm not going to trust the output of that neural net to predict my heart attack.

[255] I'm going to want to ask what if questions around that.

[256] I'm going to want to look at some us or other possible data I didn't have causal things.

[257] I'm going to want to have a dialogue with a doctor about things we didn't think about when you gathered the data.

[258] You know, I could go on and on.

[259] I hope you can see.

[260] And I think that if you say predictions, everything, that you're missing all of this stuff.

[261] and so prediction plus decision making is everything but both of them are equally important and so the field has emphasized prediction yon rightly so has seen how powerful that is but at the cost of people not being aware the decision making is where the rubber really hits the road where human lives are at stake where risks are being taken where you got to gather more data you got to think about the air bars you got to think about the consequences of your decisions on others you got about the economy around your decisions blah blah blah blah blah i'm not the only one working on those but we're a smaller tribe, and right now we're not the one that people talk about the most.

[262] But, you know, if you go out of the real world and industry, you know, at Amazon, I'd say half the people there are working on decision -making, and the other half are doing, you know, the pattern recognition.

[263] It's important.

[264] And the words of pattern recognition and prediction, I think the distinction there, not to linger on words, but the distinction there is more a constraint sort of in the lab data set versus decision -making is talking about consequential decisions in the real world, of the messiness and the uncertainty of the real world and just the whole of it, the whole mess of it that actually touches human beings and scale, like you said, market forces.

[265] That's the distinction.

[266] It helps add that perspective, that broader perspective.

[267] You're right.

[268] I totally agree.

[269] On the other hand, if you're a real prediction person, of course, you want it to be in the real world.

[270] You want to predict real world events.

[271] I'm just saying that's not possible with just data sets, that it has to be in the context of, you know, strategic things that someone's doing, data they might gather, things they could have gathered, the reasoning process around data.

[272] It's not just taking data and making predictions based on the data.

[273] So one of the things that you're working on, I'm sure there's others working on it, but I don't hear often it talked about, especially in the clarity that you talk about it.

[274] And I think it's both the most exciting and the most concerning area of AI in terms of decision -making.

[275] So you've talked about AI systems that help make decisions that scale in a distributed way, millions, billions decisions, is sort of markets of decisions.

[276] Can you, as a starting point, sort of give an example of a system that you think about when you're thinking about these kinds of systems?

[277] Yeah, so first of all, you're absolutely getting into some territory, which I will be beyond my expertise, and there are lots of things that are going to be very not obvious to think about.

[278] Just like, again, I like to think about history a little bit, but think about, put yourself back in the 60s, there was kind of a banking system that wasn't computerized, really.

[279] There was database theory emerging, and database people had, to think about how do I actually, not just move data around, but actual money and have it be, you know, valid and have transactions at ATMs happen that are actually, you know, all valid and so and so forth.

[280] So that's the kind of issues you get into when you start to get serious about sorts of things like this.

[281] I like to think about as kind of almost a thought experiment to help me think something simpler, which is the music market.

[282] And because there is, to first store, there is no music market in the world right now, in our country, for sure.

[283] There are things called record companies, and they make money, and they prop up a few really good musicians and make them superstars, and they all make huge amounts of money.

[284] But there's a long tale of huge numbers of people that make lots and lots of really good music that is actually listened to by more people than the famous people.

[285] They are not in a market.

[286] They cannot have a career.

[287] They do not make money.

[288] The creators, the creators, the so -called influencers or whatever, that diminishes who they are, right?

[289] So there are people who make extremely good music, especially in the hip -hop or Latin world these days.

[290] They do it on their laptop.

[291] That's what they do on the weekend.

[292] And they have another job during the week.

[293] They put it up on SoundCloud or other sites.

[294] Eventually, it gets streamed.

[295] It down gets turned into bits.

[296] It's not economically valuable.

[297] The information is lost.

[298] It gets put up there.

[299] People stream it.

[300] You walk around, in a big city, you see people with headphones all, you know, especially young kids, listening to music all the time.

[301] If you look at the data, very little of the music they're listening to is the famous people's music, and none of it's old music.

[302] It's all the latest stuff.

[303] But the people who made that latest stuff are like some 16 -year -old somewhere who will never make a career out of this, who will never make money.

[304] Of course, there will be a few counter examples.

[305] The record companies incentivize to pick out a few and highlight them.

[306] Long story short, there's a missing market there.

[307] There is not a consumer -producer relationship at the level of the actual creative acts.

[308] the pipelines and Spotify's of the world that take this stuff and stream it along, they make money off of subscriptions or advertising and those things.

[309] They're making the money, all right?

[310] And then they will offer bits and pieces of it to a few people, again, to highlight that, you know, they're the simulated market.

[311] Anyway, a real market would be, if you're a creator of music, that you actually are somebody who's good enough that people want to listen to you, you should have the data available to you.

[312] There should be a dashboard showing a map of the United States.

[313] So in last week, here's all the places your songs were listened to.

[314] It should be transparent, vettable, so that if someone in down in Providence sees that you're being listened to 10 ,000 times in Providence, that they know that's real data, you know, it's real data.

[315] They will have you come give a show down there.

[316] They will broadcast to the people who have been listening to you that you're coming.

[317] If you do this right, you could go down there and make $20 ,000.

[318] You do that three times during or you start to have a career.

[319] So in this sense, AI creates jobs.

[320] It's not about taking away human jobs.

[321] it's creating new jobs because it creates a new market.

[322] Once you've created a market, you've now connected up producers and consumers.

[323] You know, the person who's making the music can say to someone who comes to their shows a lot, hey, I'll play your daughter's wedding for $10 ,000.

[324] You'll say $8 ,000.

[325] They'll say $9 ,000.

[326] Then you, again, you can now get an income up to $100 ,000.

[327] You're not going to be a millionaire.

[328] All right.

[329] And now even think about really the value of music is in these personal connections, even so much so that a young kid wants to, wear a t -shirt with their favorite musician's signature on it, right?

[330] So if they listen to the music on the internet, the internet should be able to provide them with a button that they push and the merchandise arrives the next day.

[331] We can do that, right?

[332] And now, why should we do that?

[333] Well, because the kid who bought the shirt will be happy, but more the person who made the music will get the money.

[334] There's no advertising needed, right?

[335] So you could create markets between producers of consumers, take 5 % cut, your company will be perfectly sound.

[336] It'll go forward into the future.

[337] And it will create new markets.

[338] And that raises human happiness.

[339] Now, this seems like it was easy.

[340] Just create this dashboard, kind of create some connections and all that.

[341] But if you think about Uber or whatever, you think about the challenges in the real world of doing things like this.

[342] And there are actually new principles going to be needed.

[343] You're trying to create a new kind of two -way market at a different scale that's ever been done before.

[344] there's going to be, you know, unwanted aspects of the market.

[345] There'll be bad people.

[346] There'll be, you know, the data will get used in the wrong ways.

[347] You know, it'll fail in some ways.

[348] It won't deliver value.

[349] You have to think that through.

[350] Just like anyone who, like, ran a big auction or, you know, ran a big matching service in economics will think these things through.

[351] And so that maybe doesn't get at all the huge issues that can arise when you start to create markets, but it starts, at least for me, solidify my thoughts and allow me to move forward in my own thinking.

[352] Yeah, so I talked to how to research at Spotify, actually.

[353] I think their long -term goal, they've said, is to have at least one million creators make a comfortable living putting on Spotify.

[354] So I think you articulate a really nice vision of the world and the digital and the cyber space of markets.

[355] What do you think companies like Spotify or YouTube or?

[356] Netflix can do to create such markets.

[357] Is it an AI problem?

[358] Is it an interface problem?

[359] So an interface design?

[360] Is it some other kind of, is an economics problem?

[361] Who should they hire to solve these problems?

[362] Well, part of it's not just top down.

[363] So the Silicon Valley has this attitude that they know how to do it.

[364] They will create the system, just like Google did with the search box, that will be so good that they'll just everyone will adopt that.

[365] right um it's not it's it's it's everything you said but really i think missing the kind of culture all right so it's literally that 16 year old who's who's able to create the songs you don't create that as a silicon valley entity you don't hire them per se right you have to create an ecosystem in which they are wanted and that they belong right and so you have to have some cultural credibility to do things like this you know Netflix to their credit wanted some of that sort of credibility they created shows you know content they call it content it's such a terrible word but it's cold it's cultural Yeah.

[366] Right.

[367] And so with movies, you can kind of go give a large sum of money to somebody graduate from the USC film school.

[368] It's a whole thing of its own, but it's kind of like rich white people's thing to do.

[369] You know, and, you know, American culture has not been so much about rich white people.

[370] It's been about all the immigrants, all the Africans who came and brought that culture and those those rhythms and that to this world and created this whole new thing, you know, American culture.

[371] And so companies can't.

[372] artificially create that they can't just say hey we're here we're going to buy it up you got a partner right and um so but anyway you know not to denigrate these companies are all trying and they should and they they they they are i'm sure they're asking these questions and some of them are even making an effort but it is it is partly a respect the culture as you are as a technology person you got to blend your technology with cultural uh with cultural uh you know meaning how much of a role do you think the algorithm so machine learning has in connecting the consumer to the the creator, sort of the recommender system aspect of this.

[373] Yeah, it's a great question.

[374] I think pretty high.

[375] You know, there's no magic in the algorithms, but a good recommender system is way better than a bad recommender system.

[376] And recommender systems was a billion dollar industry back even, you know, 10, 20 years ago.

[377] And it continues to be extremely important going forward.

[378] What's your favorite recommender system just so we can put something?

[379] Well, just historically, I was one of the, you know, when I first went to Amazon, I first didn't like Amazon because they put the book people out of business or the library that, you know, the local booksellers went out of business.

[380] I've come to accept that there, you know, there probably are more books being sold now and poor people reading them than ever before.

[381] And then local books, stores are coming back.

[382] So, you know, that's how economics sometimes work.

[383] You go up and you go down.

[384] But anyway, when I finally started going there and I bought a few books, I was really pleased to see another few books being recommended to me that I never would have thought of.

[385] and I bought a bunch of them, so they obviously had a good business model, but I learned things, and I still, to this day, kind of browse using that service.

[386] And I think lots of people get a lot, you know, that is a good aspect of a recommendation system.

[387] I'm learning from my peers in an indirect way.

[388] And their algorithms are not meant to have them impose what we learn.

[389] It really is trying to find out what's in the data.

[390] It doesn't work so well for other kind of entities, but that's just the complexity of human life, like shirts, you know, I'm not going to get recommendations on shirts, but that's, that's, that's interesting.

[391] If you try to recommend, um, uh, restaurants, it's, it's, it's, it's, it's hard.

[392] It's hard to do it at scale.

[393] And, and, um, but, uh, a blend of recommendation systems with other, um, economic ideas, uh, matchings and so on is really, really still very open research wise and there's new companies that could emerge that do that well.

[394] what do you think is going to the messy difficult land of say politics and things like that that youtube and twitter have to deal with in terms of recommendation systems being able to suggest i think facebook just launched facebook news so they're having uh recommend the kind of news that are most likely for you to be interesting you think this is a i solvable again whatever term we want to use do you think it's a solvable problem for machines or is it a deeply human problem that's unsolvable.

[395] So I don't even think about at that level.

[396] I think that what's broken with some of these companies, it's all monetization by advertising.

[397] They're not, at least Facebook, let's, I want to critique them.

[398] They didn't really try to connect a producer and a consumer in an economic way, right?

[399] No one wants to pay for anything.

[400] And so they all, you know, starting with Google and Facebook, they went back to the playbook of, you know, the television companies back in the day.

[401] No one wanted to pay for this signal.

[402] They will pay for the TV box, but not for the signal, at least back in the day.

[403] And so advertising kind of filled that gap.

[404] And advertising was new and interesting, and it somehow didn't take over our lives quite, right?

[405] Fast forward, Google provides a service that people don't want to pay for.

[406] And so somewhat surprisingly in the 90s, they ended up making huge amounts so they cornered the advertising market.

[407] It didn't seem like that was going to happen, at least to me. These little things on the right hand side of the screen just did not seem all that economically interesting, but companies had maybe no other choice.

[408] The TV market was going to, way and billboards and so on.

[409] So they've, they got it.

[410] And I think that sadly that Google just was doing so well with that and making such money, they didn't think much more about how, wait a minute, is there a producer -consumer relationship to be set up here, not just between us and the advertisers market to be created?

[411] Is there an actual market between the producer -consumer?

[412] There, the producers, the person who created that video clip, the person that made that website, the person who could make more such things, the person who could adjust it as a function of demand.

[413] The person on the other side who's asking for different kinds of things, you know.

[414] So you see glimmers of that now.

[415] There's influencers and there's kind of a little glimmering of a market.

[416] But it should have been done 20 years ago.

[417] It should have been thought about.

[418] It should have been created in parallel with the advertising ecosystem.

[419] And then Facebook inherited that.

[420] And I think they also didn't think very much about that.

[421] So fast forward.

[422] And now they are making huge amounts of money off of advertising.

[423] And the news thing and all these clicks is just is feeding the advertising.

[424] It's all connected up to the advertising.

[425] So you want more people to click on certain things because that money flows to you, Facebook.

[426] You're very much incentivized to do that.

[427] And when you start to find it's breaking, people are telling you, well, we're getting into some troubles.

[428] You try to adjust it with your smart AI algorithms, right, and figure out what are bad clicks.

[429] Oh, maybe shouldn't be clicked through rate.

[430] It should be something.

[431] I find that pretty much hopeless.

[432] It does get into all the complies of human life.

[433] And you can try to fix it.

[434] You should.

[435] But you could also fix the whole business model.

[436] And the business model is that really, are there some human producers and consumers out there?

[437] Is there some economic value to be liberated by connecting them directly?

[438] Is it such that it's so valuable that people will be going to pay for it?

[439] All right?

[440] And micro payments, like micro, but even after you be micro.

[441] So I like the example.

[442] Suppose I'm going, next week I'm going to India.

[443] Never been to India before, right?

[444] I have a couple of days in Mumbai.

[445] I have no idea what to do there, right?

[446] And I could go on the web right now and search.

[447] It's going to be.

[448] kind of hopeless.

[449] I'm not going to find.

[450] I'll have lots of advertisers in my face.

[451] What I really want to do is broadcast to the world that I am going to Mumbai and have someone on the other side of a market look at me and there's a recommendation system there.

[452] So they're not looking at all possible people coming to Mumbai.

[453] They're looking at the people who are relevant to them.

[454] So someone, my age group, someone who kind of knows me in some level, I give up a little privacy by that, but I'm happy because what I'm going to get back is this person can make a little video for me. or they're going to write a little two -page paper on here's the cool things that you want to do and move by this week especially, right?

[455] I'm going to look at that.

[456] I'm not going to pay a micropayment.

[457] I'm going to pay, you know, $100 or whatever for that.

[458] It's real value.

[459] It's like journalism.

[460] As an odd subscription, it's that I'm going to pay that person in that moment.

[461] Company's going to take 5 % of that.

[462] And that person has now got a, it's a gig economy, if you will.

[463] But, you know, done for thinking about a little bit behind YouTube, there was actually people who could make more of those things.

[464] If they were connected into a market, they would more of those things independently.

[465] You don't have to tell them what to do.

[466] You don't have to incentivize them in any other way.

[467] And so, yeah, these companies, I don't think of thought long and heard about that.

[468] So I do distinguish on, you know, Facebook on the one side who's just not thought about these things at all.

[469] I think, thinking that AI will fix everything.

[470] And Amazon thinks about them all the time because they were already out in the real world.

[471] They were delivering packages, people's doors.

[472] They were worried about a market.

[473] They were worrying about sellers and, you know, they worry and some things they do are great.

[474] Some things maybe not so great, but, you know, they're in that business model.

[475] And then, I'd say Google sort of hover somewhere in between.

[476] I don't, I don't think for a long, long time they got it.

[477] I think they probably see that YouTube is more pregnant with possibility than they might have thought and that they're probably heading that direction.

[478] But Silicon Valley's been dominated by the Google Facebook kind of mentality and the subscription and advertising.

[479] And that is, that's the core problem, right?

[480] The fake news actually rides on top of that because it means that you're monetizing with clip through rate.

[481] And that is the core problem.

[482] You got to remove that.

[483] So advertisement, if we're going to linger on that, I mean, that's an interesting thesis.

[484] I don't know if everyone really deeply thinks about that.

[485] So you're right.

[486] The thought is the advertising model is the only thing we have, the only thing we'll ever have, so we have to fix, we have to build algorithms that despite that business model, you know, find the better angels of our nature and do good by society and by the individual.

[487] But you think we can slowly you think first of all there's a difference between you should and could so you're saying we should slowly move away from the advertising model and have a direct connection between the consumer and the creator the the question i also have is can we because the advertising model is so successful now in terms of just making a huge amount of money and therefore being able to build a big company that provides has really smart people working that create a good service Do you think it's possible?

[488] And just to clarify, you think we should move away.

[489] Well, I think we should, yeah.

[490] But we is, you know, not me. Society.

[491] Yeah, well, the companies.

[492] I mean, so first of all, full disclosure, I'm doing a day a week at Amazon because I kind of want to learn more about how they do things.

[493] So, you know, I'm not speaking for Amazon in any way.

[494] But, you know, I did go there because I actually believe they get a little bit of this or trying to create these markets.

[495] And they don't really use, advertisement is not a crucial part of Amazon.

[496] That's a good question.

[497] So it has become not crucial, but it's become more and more present.

[498] you go to Amazon website, and without revealing too many deep secrets about Amazon, I can tell you that a lot of people in the company question this, and there's a huge questioning going on.

[499] You do not want a world where there's zero advertising.

[500] That actually is a bad world, okay?

[501] So here's a way to think about it.

[502] You're a company that, like Amazon, is trying to bring products to customers, right?

[503] And the customer in any given moment you want to buy a vacuum cleaner, say, you want to know what's available for me. And, you know, it's not going to be that obvious.

[504] You have to do a little bit of work at it.

[505] The recommendation system will sort of help.

[506] Right.

[507] But now suppose this other person over here has just made the world, you know, they spent a huge amount of energy.

[508] They had a great idea.

[509] They made a great vacuum cleaner.

[510] They know.

[511] They really did it.

[512] They nailed it.

[513] It's an MIT, you know, Wiz kid that made a great new vacuum cleaner.

[514] It's not going to be in the recommendation system.

[515] No one will know about it.

[516] The algorithms will not find it.

[517] Okay.

[518] At all.

[519] Right.

[520] How do you allow that vacuum cleaner to start to get in front of people be sold?

[521] Well, advertising.

[522] And here what advertising is, it's a signal that you're, you you believe in your product enough that you're willing to pay some real money for it.

[523] And to me as a consumer, I look at that signal.

[524] I say, well, first of all, I know these are not just cheap little ads because we have now right now.

[525] I know that, you know, these are super cheap, you know, pennies.

[526] If I see an ad where it's actually, I know the company is only doing a few of these and they're making, you know, real money is kind of flowing.

[527] And I see an ad, I may pay more attention to it.

[528] And I actually might want that because I see, hey, that guy spent money on his vacuum cleaner.

[529] Oh, maybe there's something good there.

[530] So I will look at it.

[531] And so that's part of the overall information flow in a good market.

[532] So advertising has a role.

[533] But the problem is, of course, that signal is now completely gone because it just, you know, dominole by these tiny little things that add up to big money for the company.

[534] You know, so I think it will change because societies just don't, you know, stick with things that annoy a lot of people.

[535] And advertising currently annoys people more than it provides information.

[536] And I think that at Google probably is smart enough to figure out that this is a dead, this is a bad model.

[537] even though it's a hard huge amount of money and they'll have to figure out how to pull it away from it slowly and I'm sure the CEO there will figure it out but they need to do it and they need to so if you reduce advertising not to zero but you reduce it at the same time you bring up producer consumer actual real value being delivered so real money is being paid and they take a 5 % cut that 5 % could start to get big enough to cancel out the lost revenue from the kind of the poor kind of advertising and I think that a good company will do that will realize that and they're a company you know facebook you know again god bless them they they bring you know grandmothers uh you know they bring children's pictures into grandmother's lives it's fantastic um but they need to think of a new business model and and they that's that's the core problem there um until they start to connect producer consumer i think they will just just continue to make money and then buy the next social network company and then buy the next one and the innovation level will not be high and the health issues will not go away.

[538] So I apologize that we've kind of returned to words.

[539] I don't think the exact terms matter, but in sort of defense of advertisement, don't you think the kind of direct connection between consumer and creator, producer, is the best, like the, is what advertisement strives to do, right?

[540] so that it's best advertisement is literally now Facebook is listening to our conversation and heard that you're going to India and will be able to actually start automatically for you making these connections and start giving this offer.

[541] So like, I apologize if it's just a matter of terms, but just to draw a distinction.

[542] Is it possible to make advertisements just better and better and better algorithmically to where it actually becomes a connection, almost a direct connection?

[543] That's a good question.

[544] So let's put on that, push on it.

[545] First of all, what we just talked about, defending advertising.

[546] Okay.

[547] So I was defending it as a way to get signals into a market that don't come any other way, especially algorithmically.

[548] It's a sign that someone spent money on it.

[549] It's a sign they think it's valuable.

[550] And if I think that if other things, someone else thinks it's valuable, and if I trust other people, I might be willing to listen.

[551] I don't trust that Facebook, though, is who's an intermediary between this, I don't think they care about me. Okay?

[552] I don't think they do.

[553] And I find it creepy that they know I'm going to India next week because of our conversation.

[554] Why do you think that?

[555] Can we, so what, can you just put your PR hat on?

[556] Why do you think you find Facebook, creepy and not trust them, as do majority of the population?

[557] So they're out of the Silicon Valley companies, I saw like not approval rate, but there's, there's ranking of how much people trust companies and Facebook is in the gutter.

[558] In the gutter, including people inside of Facebook.

[559] So what, what do you attribute that to?

[560] Because when I - Come on, you don't find it creepy.

[561] that right now we're talking that I might walk out on the street right now that some unknown person who I don't know kind of comes up to me and says I hear you going to India I mean that's not even Facebook that's just a if I want transparency in human society I want to have if you know something about me there's actually some reason you know something about me that's something that if I look at it later and audit it kind of I approve you know something about me because you care in some way there's a caring relationship even or an economic one or something not just that you're someone who could exploit it in ways I don't know about or care about or I'm troubled by or whatever and we're in a world right now where that happened way too much and that Facebook knows things about a lot of people and could exploit it and does exploit it at times I think most people do find that creepy it's not for them it's not it's not that Facebook does not do it because they care about them right in any real sense and they shouldn't they should not be a big brother caring about us that is not the role of a company like that.

[562] Why not?

[563] Wait, not the Big Brother part, but the caring, the trust thing.

[564] I mean, don't those companies, just to linger under it, because a lot of companies have a lot of information about us.

[565] I would argue that there's companies like Microsoft that has more information about us than Facebook does, and yet we trust Microsoft more.

[566] Well, Microsoft is pivoting.

[567] Microsoft, you know, under Sotia Nadell, has decided this is really important.

[568] We don't want to do creepy things.

[569] Really want people to trust us to actually only use information in ways that they really would approve of.

[570] that we don't decide, right?

[571] And I'm just kind of adding that the health of a market is that when I connect to someone who produces a consumer, it's not just a random producer of consumer, it's people who see each other, they don't like each other, but they sense that if they transact, some happiness will go up on both sides.

[572] If a company helps me to do that in moments that I choose of my choosing, then fine.

[573] And also think about the difference between, you know, browsing versus buying.

[574] right.

[575] There are moments in my life.

[576] I just want to buy, you know, a gadget or something.

[577] I need something for that moment.

[578] I need some ammonia for my house or something because I got a problem and a spill.

[579] I want to just go in.

[580] I don't want to be advertised at that moment.

[581] I don't want to be led down very straight.

[582] You know, that's annoying.

[583] I want to just go and have it to be extremely easy to do what I want.

[584] Other moments I might say, no, it's like today I'm going to the shopping mall.

[585] I want to walk around and see things and see people and be exposed to stuff.

[586] So I want control over that, though.

[587] I don't want the competent.

[588] these algorithms to decide for me, right?

[589] And I think that's the thing.

[590] It's a total loss of control if Facebook thinks they should take the control from us of deciding when we want to have certain kinds of information when we don't, what information that is, how much it relates to what they know about us that we didn't really want them to know about us.

[591] They're not, I don't want them to be helping me in that way.

[592] I don't want them to be helping them by they decide what they have control over what I want and when.

[593] I totally agree.

[594] So Facebook, by the way, I have this optimistic thing where I think Facebook has the kind of personal information about us that could create a beautiful thing.

[595] So I'm really optimistic of what Facebook could do.

[596] It's not what is doing, but what it could do.

[597] I don't see that.

[598] I think that optimism is misplaced because there's not a you have to have a business model behind these things.

[599] Create a beautiful thing is really, let's be clear.

[600] It's about something that people would value.

[601] And I don't think they have that business model.

[602] And I don't think they will suddenly discover it.

[603] by what you know a long hot shower I disagree I disagree in terms of you can discover a lot of amazing things in a shower so I didn't say that I said they won't cover they won't do it in the shower I think a lot of other people will discover it I think that this guy so I should also full disclosure there's a company called United Masters which I'm on their board and they've created this music market and they have 100 ,000 artists now signed on and they've done things like gone to the NBA and the NBA the music you find behind NBA eclipse right now is their music.

[604] That's a company that had the right business model in mind from the get -go, right, executed on that.

[605] And from day one, there was value brought to, so here you have a kid who made some songs who suddenly their songs were on the NBA website, right?

[606] That's real economic value to people.

[607] And so, you know.

[608] So you and I differ on the optimism of being able to sort of uh change the direction of the titanic right so i yeah i'm older than you so i've seen titanic's crash got it but uh so and just to elaborate because i totally agree with you and i just want to know how difficult you think this problem is of so for example i um i want to read some news and i would there's a lot of times in the day where something makes me either smile or think in a way where I, like, consciously think this really gave me value.

[609] Like, I sometimes listen to the daily podcast in the New York Times, way, way, way better than the New York Times themselves, by the way, for people listening.

[610] That's like real journalism is happening for some reason in the podcast space.

[611] It doesn't make sense to me. But often I listen to it, 20 minutes, and I would be willing to pay for that, like, $5, $10, for that experience.

[612] And how difficult, that's kind of what you're getting at, is that little transaction, How difficult is it to create a frictional system like Uber has, for example, for other things?

[613] What's your intuition there?

[614] So, first of all, I pay a little bits of money to, you know, to send.

[615] There's something called quartz that does financial things.

[616] I like Medium as a site.

[617] I don't pay there, but I would.

[618] You had a great post on Medium.

[619] I would have loved to pay you a dollar and not others.

[620] I wouldn't have wanted it.

[621] I wouldn't have wanted it per se because there should be also sites where that's not actually the goal.

[622] The goal is to actually have a broadcast channel that I monetize in some other way if I chose to.

[623] I mean, I could now.

[624] People know about it.

[625] I could.

[626] I'm not doing it, but that's fine with me. Also, the musicians who are making all this music, I don't think the right model is that you pay a little subscription fee to them, because people can copy the bits too easily, and it's just not that somewhere the value is.

[627] The value is that a connection was made between real human beings, then you can follow up on that, right, and create yet more value.

[628] So, no, I think...

[629] There's a lot of open questions.

[630] Yeah, hot open questions, but also, yeah, I do want good recommendation systems that recommend cool stuff to me, but it's pretty hard, right?

[631] I don't like them to recommend stuff just based on my browsing history.

[632] I don't like that based on stuff they know about me, quote, unquote.

[633] What's unknown about me is the most interesting.

[634] So this is the really interesting question.

[635] We may disagree, maybe not.

[636] I think that I love recommender systems, and I want to give them everything about me in a way that I trust.

[637] Yeah, but you don't.

[638] Because, so for example, this morning, I clicked on, you know, I was pretty sleepy this morning, I clicked on a story about the Queen of England, right?

[639] I do not give a damn about the Queen of England.

[640] I really do not.

[641] But it was clickbait.

[642] It kind of looked funny and I had to say, what the heck are they talking about them?

[643] I don't want to have my life, you know, heading that direction.

[644] Now that's in my browsing history.

[645] The system, in any reasonable system will think that I care about the queen of order.

[646] Right, but you're saying all the trace, all the digital exhaust or whatever, that's been kind of the models.

[647] If you collect to all this stuff, you're going to figure all of us out.

[648] Well, if you're trying to figure out, like, kind of one person, like Trump or something, maybe you could figure him out.

[649] But if you're trying to figure out, you know, 500 million people, you know, no way, no way.

[650] Do you think so?

[651] No, I think so.

[652] I think we are, humans are just amazingly rich and complicated.

[653] Every one of us has our little quirks.

[654] Everyone else has our little things that could intrigue us that we don't even know and will intrigue us.

[655] And there's no sign of it in our past.

[656] But by God, there it comes.

[657] And, you know, you fall in love with it.

[658] And I don't want a company trying to figure that out for me and anticipate that.

[659] I want them to provide a forum, a market, a place that I kind of go.

[660] And by hook or by crook, this happens.

[661] I'm walking down the street and I hear some Chilean music being played, and I never knew I like Chile music.

[662] Wow.

[663] So there is that side.

[664] And I want them to provide a limited, but interesting place to go, right?

[665] And so don't try to use your AI to kind of, you know, figure me out and then put me in a world where you figured me out.

[666] You know, no, create spaces for human beings where our creativity.

[667] and our style will be enriched and come forward.

[668] And it'll be a lot of more transparency.

[669] I won't have people randomly, anonymously putting comments up and especially based on stuff they know about me, facts that you know, we are so broken right now, especially if you're celebrity, but it's about anybody that anonymous people are hurting lots and lots of people right now.

[670] And that's part of this thing that Silicon Valley is thinking that, you know, just collect all this information and use it in a great way.

[671] So, no, I'm not a pessimism.

[672] I'm very much an optimist.

[673] but I think that's just been the wrong path for the whole technology to take.

[674] Be more limited, create, let humans rise up.

[675] Don't try to replace them.

[676] That's the AI mantra.

[677] Don't try to anticipate them.

[678] Don't try to predict them because you're not going to be the those things.

[679] You're going to make things worse.

[680] Okay.

[681] So right now, just give this a chance.

[682] Right now the recommender systems are the creepy people in the shadow watching your every move.

[683] so they're looking at traces of you they're not directly interacting with you sort of your close friends and family the way they know you is by having conversation by actually having interactions back and forth do you think there's a place for recommender systems sort of to step because you just emphasize the value of human to human connection but let's give a chance AI human connection is there a role for an AI system to have conversations with you in terms of to try to figure out what kind of music you like Not by just watching what you listen in it, but actually having a conversation, natural language or otherwise.

[684] Yeah, no, I'm, so I'm not against it.

[685] I just wanted to push back against the maybe you're saying, you have autism for Facebook.

[686] So there I think it's misplaced.

[687] But I think that distributing - I'm the one guy pending Facebook.

[688] Yeah, no, so good for you.

[689] Go for it.

[690] That's a hard spot to be.

[691] Yeah, no, good.

[692] Human interaction, like on our daily, the context around me in my own home is something that I don't want some big company to know about it all, but I would be more than happy to have technology help me with it.

[693] which kind of technology well you know just Alexa Amazon well a good Alexa's done right I think Alexa's a research platform right now more than anything else but Alexa done right you know could do things like I I leave the water running in my garden and I say hey Alexa the water's running my garden and even have Alexa figure out that that means when my wife comes home that she should be told about that that's a little bit of reasoning I would call that AI and by any kind of stretch it's a little bit of reasoning and it actually kind of would make my life a little easier and better and you know I don't I wouldn't call this a wow moment, but I kind of think that overall rises human happiness up to have that kind of thing.

[694] But not when you're lonely.

[695] Alexa knowing loneliness.

[696] No, no. I don't want Alexa to feel intrusive and I don't want just the designer of the system to kind of work all this out.

[697] I really want to have a lot of control and I want transparency and control.

[698] And if a company can stand up and give me that in the context of new technology, I think they're going to first of all be way more successful than our current generation.

[699] And like I said, I was mentioning Microsoft.

[700] I really think they're pivoting to kind of be the trusted old uncle, but, you know, I think that they get that this is a way to go, that if you let people find technology empowers them to have more control and have control not just over privacy, but over this rich set of interactions, that that people are going to like that a lot more.

[701] And that's the right business model going forward.

[702] What does control over privacy look like?

[703] Do you think you should be able to just view all the data that?

[704] No, it's much more than that.

[705] I mean, first of all, it should be an individual decision.

[706] Some people don't want privacy.

[707] They want their whole life out there.

[708] Other people's want it.

[709] Privacy is not a zero one.

[710] It's not a legal thing.

[711] It's not just about which date is available, which is not.

[712] I like to recall to people that, you know, a couple hundred years ago, everyone, there was not really big cities.

[713] Everyone lived in the countryside and villages.

[714] And in villages, everybody knew everything about you.

[715] You didn't have any privacy.

[716] Is that bad?

[717] Are we better off now?

[718] Well, you know, arguably no, because what did you get for that loss of at least certain kinds of privacy?

[719] well people help each other because they know everything about you they know something bad's happening they will help you with that right and now you live in a big city no one knows the amount of you get no help so it kind of depends the answer I want certain people who I trust and there should be relationships I should kind of manage all those but who knows what about me I should have some agency there it shouldn't I shouldn't be a drift in a sea of technology where I have no agency I don't want to go reading things and checking boxes so I don't know how to do I'm not a privacy researcher per se.

[720] I recognize the vast complexity of this.

[721] It's not just technology.

[722] It's not just legal scholars meeting technologists.

[723] There's got to be kind of a whole layers around it.

[724] And so when I allude to this emerging engineering field, this is a big part of it.

[725] When electrical engineering came, I wasn't around in the time, but you just didn't plug electricity into walls and all kind of worked.

[726] You don't have like underwriters' laboratory that reassured you that that plug's not going to burn up your house.

[727] and that that machine will do this and that and everything.

[728] There'll be whole people who can install things.

[729] There'll be people who can watch the installers.

[730] There'll be a whole layer, an onion of these kind of things.

[731] And for things as deep andly interesting as privacy, which is at least as interesting as electricity, that's going to take decades to kind of work out, but it's going to require a lot of new structures that we don't have right now.

[732] So it's kind of hard to talk about it.

[733] And you're saying there's a lot of money to be made if you get it right.

[734] Absolutely.

[735] A lot of money to be made.

[736] And all these things that provide human services and people recognize them as useful parts of their lives.

[737] So, yeah.

[738] So, yeah, the dialogue sometimes goes from the exuberant technologists to the no technology is good, kind of.

[739] And that's, you know, in our public discourse, you know, in news stories, you see too much of this kind of thing.

[740] And the sober discussions in the middle, which are the challenging ones to have, are where we need to be having our conversations.

[741] And, you know, actually, there's not many forum fora for those.

[742] You know, that's kind of what I would look for maybe I could go and I could read a comment section of something and it would actually be this kind of dialogue going back and forth you don't see much of this right which is why actually there's a resurgence of podcasts out of all because good people are really hungry for conversation yeah there's technology is not helping much so comment sections of anything including youtube yeah is not hurting hurting and not helping and you think technically speaking is possible to help I don't know the answers, but it's less anonymity, a little more locality, you know, worlds that you kind of enter in and you trust the people there in those worlds so that when you start having a discussion, you know, not only is that people not going to hurt you, but it's not going to be a total waste of your time because there's a lot of wasting of time that, you know, a lot of us, I pulled out of Facebook early on because it was clearly going to waste a lot of my time, even though there was some value.

[743] And so, yeah, worlds that are somehow you enter in and you know what you're getting and it's kind of appeals to you.

[744] You might, new things might happen.

[745] but you kind of have some trust in that world and there's some deep interesting complex psychological aspects around anonymity how that changes human behavior that's quite dark and quite dark yeah i think a lot of us are especially those of us who really loved the advent of technology i love social networks when they came out i was just i didn't see any negatives there at all but then i started seeing comment sections i think it was maybe you know with the cnn or something and i started to go wow this this darkness i just did not know about and our technology is now amplifying it so sorry for the big philosophical question but on that topic do you think human beings because you've also out of all things had a foot in psychology too the do you think human beings are fundamentally good like all of us have good intent that could be mined or is it depending on context and environment everybody could be evil So my answer is fundamentally good, but fundamentally limited.

[746] All of us have very, you know, blinkers on.

[747] We don't see the other person's pain that easily.

[748] We don't see the other person's point of view that easily.

[749] We're very much in our own head, in our own world.

[750] And on my good days, I think the technology could open us up to, you know, more perspectives and more less blinkered and more understanding.

[751] You know, a lot of wars in human history happened because of just ignorance.

[752] They didn't, they thought the other person was doing this.

[753] Well, that person wasn't doing this.

[754] And we have a huge amounts of that.

[755] But in my lifetime, I've not seen technology really help in that way yet.

[756] And I do believe in that.

[757] But, you know, no, I think fundamentally humans are good.

[758] People suffer.

[759] People have grievances.

[760] You have grudges and those things cause them to do things they probably wouldn't want.

[761] They regret it often.

[762] So, no, I think it's a, you know, part of the progress of technology is to indeed allow it to be a little easier to be the real good person you actually are.

[763] Well, but do you think.

[764] Do you think individual human life or society can be modeled as an optimization problem?

[765] Not the way I think typically.

[766] I mean, that's your time about one of the most complex phenomena in the whole, you know, in all the universe.

[767] Which individual human life or society as a whole?

[768] Both.

[769] Both.

[770] I mean, individual human life is amazingly complex.

[771] And so, you know, optimization is kind of just one branch of mathematics that talks about certain kind of things.

[772] And it just feels way too limited for the complexity of.

[773] such things.

[774] What properties of optimization problems?

[775] Do you think most interesting problems that could be solved through optimization?

[776] What kind of properties does that surface have?

[777] Non -convexity, convexity, linearity, all those kinds of things, saddle points.

[778] Well, so optimization is just one piece of mathematics.

[779] You know, there's like, even in our era, we're aware that it's a sampling.

[780] It's coming up examples of something.

[781] Coming up with a distribution.

[782] What's optimization?

[783] What's sampling?

[784] Well, you can, if you're a certain kind of mathematics, you can try to blend them and make them seem to be sort of the same thing.

[785] But optimization, roughly speaking, trying to find a point that, a single point, that is the optimum of a criterion function of some kind.

[786] And sampling is trying to, from that same surface, treat that as a distribution or density and find points that have high density.

[787] So I want the entire distribution in a sampling paradigm, and I want the single point, that's the best point in the optimization paradigm.

[788] Now, if you were optimizing in the space of probability measures, the output of that could be a whole probability distribution.

[789] So you can start to make these things the same.

[790] But in mathematics, if you go too high up that kind of abstraction hierarchy, you start to lose the ability to do the interesting theorems, so you kind of don't try to overly over -abstract.

[791] So as a small tangent, what kind of world do you find more appealing?

[792] one that is deterministic or stochastic?

[793] Well, that's easy.

[794] I mean, I'm a statistician.

[795] You know, the world is highly stochastic.

[796] I don't know what's going to happen in the next five minutes, right?

[797] What you're going to ask, what we're going to do?

[798] Due to the uncertainty.

[799] Do you do the...

[800] Massive uncertainty.

[801] You know, massive uncertainty.

[802] And so the best I can do is have come rough sense or probability distribution on things and somehow use that in my reasoning about what to do now.

[803] So how does the distribution, it at scale when you have multi -agent systems look like so optimization can optimize sort of it makes a lot more sense sort of at least from a robotics perspective for a single robot for a single agent trying to optimize some objective function when you start to enter the real world this game theoretic concept starts popping up and that how do you see optimization in this because you've talked about markets and a scale, what does that look like?

[804] Do you see it as optimization?

[805] Do you see it as a sampling?

[806] Do you see, like, how should you monitor?

[807] Yeah, these all blend together and a system designer thinking about how to build an incentivized system will have a blend of all these things.

[808] So, you know, a particle in a potential well is optimizing a functional called a Lagrangian, right?

[809] The particle doesn't know that.

[810] There's no algorithm running that does that.

[811] It just happens.

[812] So it's a description mathematically of something that helps us understand as analysts what's happening.

[813] And so the same will happen when we talk about, you know, mixtures of humans and computers and markets and so and so forth.

[814] There'll be certain principles that allow us to understand what's happening and whether or not the actual algorithms are being used by any sense is not clear.

[815] Now, at some point I may have set up a multi -agent or market kind of system, and I'm now thinking about an individual agent in that system, and they're asked to do some task and they're incentivizing some way.

[816] They get certain signals and they have some utility.

[817] What they will do at that point is they just won't know.

[818] the answer, they may have to optimize to find an answer.

[819] So an optimist could be embedded inside of an overall market.

[820] You know, and game theory is very, very broad.

[821] It is often studied very narrowly for certain kinds of problems.

[822] But it's roughly speaking, there's just the, I don't know what you're going to do, so I kind of anticipate that a little bit, and you anticipate what I'm anticipating, and we kind of go back and forth in our own minds.

[823] We run kind of thought experiments.

[824] You've talked about this interesting point in terms of game theory.

[825] You know, most optimization problems really hate saddle points.

[826] Maybe you can describe what saddle points are.

[827] But I've heard you kind of mentioned that there's a branch of optimization that you could try to explicitly look for saddle points as a good thing.

[828] Oh, not optimization.

[829] That's just game theory.

[830] There's all kinds of different equilibrium in game theory.

[831] And some of them are highly explanatory behavior.

[832] They're not attempting to be algorithmic.

[833] They're just trying to say, if you happen to be at this equilibrium, you would see certain kind of behavior.

[834] and we see that in real life.

[835] That's what an economist wants to do, especially a behavioral economist.

[836] In continuous differential game theory, you're in continuous spaces.

[837] Some of the simplest equilibrium are saddle points.

[838] Nash equilibrium is a saddle point.

[839] It's a special kind of saddle point.

[840] So classically in game theory, you were trying to find Nash equilibrium.

[841] And an algorithm in a game theory, you're trying to find algorithms that would find them.

[842] And so you're trying to find saddle points.

[843] I mean, so that's literally what you're trying to do.

[844] But, you know, any economist knows that Nash equilibrium have their limitations.

[845] They are definitely not that explanatory in many situations.

[846] They're not what you really want.

[847] There's other kind of equilibria, and there's names associated with these because they came from history with certain people working on them, but there will be new ones emerging.

[848] So, you know, one example is a Stackelberg equilibrium.

[849] So, you know, Nash, you and I are both playing this game against each other or for each other, maybe it's cooperative, and we're both going to think it through, then we're going to decide, and we're going to do our thing simultaneously.

[850] You know, in a Stockleburg, no, I'm going to be the first mover.

[851] I'm going to make a move.

[852] You're going to look at my move, and then you're going to make yours.

[853] Now, since I know you're going to look at my move, I anticipate what you're going to do, and so I don't do something stupid.

[854] But then I know that you are also anticipating me, so we're kind of going back in some form of mind, but there is then a first mover thing.

[855] And so there's a different equilibrium, all right?

[856] And so just mathematically, yeah, these things have certain topologies, in certain shapes that are like satellite what's and then algorithmically or dynamically how do you move towards them how do you move away from things um you know so some of these questions have answers they've been studied others do not and especially if it becomes stochastic especially if there's large numbers of decentralized things there's just uh you know young people getting in this field who kind of think it's all done because we have you know tensor flow well no these are all open problems and they're really important and interesting and it's about strategic settings how do i I collect data.

[857] Suppose I don't know what you're going to do because I don't know you very well, right?

[858] Well, I got to collect data about you.

[859] So maybe I want to push you in a part of the space where I don't know much about you so I can get data.

[860] And then later I'll realize that you'll never go there because of the way the game is set up.

[861] But you know, that's part of the overall, you know, data analysis context is that.

[862] Even the game of poker is fascinating space.

[863] Yeah.

[864] Whenever there's any uncertainty, a lack of information, it's a super exciting space.

[865] Yeah.

[866] Just Lingard optimization for a second.

[867] So when we look at deep learning, it's essentially minimization of a complicated loss function.

[868] So is there something insightful or hopeful that you see in the kinds of function surface that loss functions, that deep learning in the real world is trying to optimize over?

[869] Is there something interesting?

[870] As is just the usual kind of problems of optimization?

[871] I think from an optimization point of view, that surface, it's pretty smooth.

[872] And secondly, if it's over -parameterized, there's kind of lots of paths down to reasonable optima.

[873] And so kind of the getting downhill to an optimum is viewed as not as hard as you might have expected in high dimensions.

[874] The fact that some optima tend to be really good ones and others not so good and you tend to, sometimes you find the good ones is sort of still needs explanation.

[875] Yes, that's a total mystery.

[876] But the particular surface is coming from the particular generation of neural nets.

[877] I kind of suspect those will, those will change.

[878] In 10 years, it will not be exactly those surfaces.

[879] There'll be some others that are, and optimization theory will help contribute to why other surfaces or why other algorithms.

[880] Layers of arithmetic operations with a little bit of non -linearity, that's not, that didn't come from neuroscience per se.

[881] I mean, maybe in the minds of some of the people working on it, they were thinking even about brains, but they were arithmetic circuits in all kinds of fields, you know, computer science control theory and so on.

[882] And that layers of these could transform things in certain ways, and that if it's smooth, maybe you could, you know, find parameter values, you know, is a big, is a sort of big discovery that it's working, it's able to work at this scale.

[883] But I don't think that, we're stuck with that, and we're certainly not stuck with that because we're understanding the brain.

[884] So in terms of, on the algorithm side, sort of gradient descent, do you think we're stuck with gradient descent, is variance of it?

[885] What variance do you find interesting?

[886] or do you think there'll be something else invented that is able to walk all over these optimization spaces in more interesting ways?

[887] So there's a co -design of the surface or the architecture and the algorithm.

[888] So if you just ask, if we stay with the kind of architectures we have now, not just neural nets, but, you know, phase retrieval architectures or matrix completion architecture and so on, you know, I think we've kind of come to a place where, yeah, a stochastic gradient algorithms are dominant And there are versions that are a little better than others.

[889] They have more guarantees.

[890] They're more robust and so on.

[891] And there's ongoing research to kind of figure out which is the best argument for which situation.

[892] But I think that that will start to co -evolve, that that'll put pressure on the actual architecture.

[893] And so we shouldn't do it in this particular way.

[894] We should do it in a different way because this other algorithm is now available if you do it in a different way.

[895] So that I can't really anticipate that co -evolution process.

[896] You know, gradients are amazing mathematical objects.

[897] They have a lot of people who start to study them more deeply mathematically are kind of shocked about what they are and what they can do.

[898] I mean, to think about this way, if suppose that I tell you, if you move along the x -axis, you get, you know, you go uphill in some objective by, you know, three units.

[899] Whereas if you move along the y -axis, you go uphill by seven units, right?

[900] Now I'm going to only allow you to move a certain, you know, unit distance.

[901] distance, right?

[902] What are you going to do?

[903] Well, the most people will say, I'm going to go along the y -axis.

[904] I'm getting the biggest bang for my buck, you know, and my buck is only one unit, so I'm going to put all of it in the y -axis, right?

[905] And why should I even take any of my strength, my step -size, and put any of it in the x -axis because I'm getting less bang for my buck?

[906] That seems like a completely, you know, clear argument, and it's wrong because the gradient direction is not to go along the y -axis.

[907] It's to take a little bit of the x -axis.

[908] It's to take a little bit of the x -axis.

[909] And to understand that, you have to know some math.

[910] And so even a trivial so -called operator -like rating is not trivial, and so, you know, exploiting its properties is still very, very important.

[911] Now we know that just creating descent has got all kinds of problems.

[912] It gets stuck in many ways and it had to have, you know, good dimension dependence and so on.

[913] So my own line of work recently has been about what kinds of stochasticity, how can we get dimension dependence, how can we do the theory of that?

[914] and we've come up pretty favorable results with certain kinds of stochasticity.

[915] We have sufficient conditions generally.

[916] We know if you do this, we will give you a good guarantee.

[917] We don't have necessary conditions that it must be done a certain way in general.

[918] So stochasticity, how much randomness to inject into the walking along the gradient?

[919] And what kind of randomness?

[920] Why is randomness good in this process?

[921] Why is stochasticity good?

[922] Yeah, so I can give you simple answers, but in some sense, again, it's kind of amazing.

[923] Stochasticity just, you know, particular features of a surface that could have hurt you if you were doing one thing deterministically won't hurt you because, you know, by chance, you know, there's very little chance that you would get hurt.

[924] And, you know, so here stochasticity, you know, is just kind of saves you from some of the particular features of surfaces that, you know, and in fact, if you think about, you know, surfaces that are discontinuous in a first derivative, like, you know, an absolute value function, you will go down and hit that point where there's non -differentiability, right?

[925] And if you're running a deterministic at that point, you can really do something bad, right?

[926] Where stochasticity just means it's pretty unlikely that's going to happen.

[927] You're going to get, you're going to hit that point.

[928] So, you know, it's, again, non -trivial analyzed, but especially in higher dimensions, also stochasticity, our intuition isn't very good about it, but it has properties that kind of are very appealing in high dimensions for kind of law of large number reasons.

[929] So it's all part of the mathematics.

[930] It's what's fun to work in the field is that you get to try to understand this mathematics.

[931] But long story short, you know, partly empirically it was discovered.

[932] Stochastic gradient is very effective.

[933] And theory kind of followed, I'd say, that.

[934] But I don't see that we're getting clearly out of that.

[935] What's the most beautiful, mysterious, a profound idea to you in optimization?

[936] I don't know the most, but let me just say that, you know, Nesterov's work on Nesterov acceleration to me is pretty surprising and pretty deep.

[937] Can you elaborate?

[938] Well, Nestrug acceleration is just that suppose we are going to use gradients to move around into space for the reasons I've alluded to.

[939] They're nice directions to move.

[940] And suppose that I tell you that you're only allowed to use gradients.

[941] You're not going to be allowed to use this local person.

[942] It can only sense kind of a change in the surface.

[943] But I'm going to give you kind of a computer that's able to store all your previous gradients.

[944] And so you start to learn something about the surface.

[945] And I'm going to restrict you to maybe move in the direction of like linear span of all the gradients.

[946] So you can't kind of just move in some arbitrary direction.

[947] Right.

[948] So now we have a well -defined method.

[949] mathematical complexity model.

[950] There's a certain classes of algorithms that can do that, and others that can't.

[951] And we can ask for certain kinds of surfaces, how fast can you get down to the optimum?

[952] So there's answers to these.

[953] So for a smooth convex function, there's an answer, which is one over the number of steps squared.

[954] You will be within a ball of that size after K steps.

[955] Grading descent in particular has a slower rate.

[956] It's one over K. okay um so you could ask is gradient is that actually even though we know it's a good algorithm is it the best algorithm in the sense of the answer is no well well not clear yet because what one over k score is a lower bound that's that's provably the best you can do what gradient is one or k but is there something better and so it was i think it's a surprise to most the nestrov discovered a new algorithm that has got two pieces to it it uses two gradients um and puts those together in a certain kind of obscure way, and the thing doesn't even move downhill all the time.

[957] It sometimes goes back uphill.

[958] And if you're a physicist, that kind of makes some sense.

[959] You're building up some momentum, and that is kind of the right intuition, but that intuition is not enough to understand kind of how to do it and why it works.

[960] But it does.

[961] It achieves one over a K -squared, and it has a mathematical structure, and it's still kind of, to this day, a lot of us are writing papers in trying to explore that and understand it.

[962] So there are lots of cool ideas in optimization, but just kind of using gradients, I think, is number one.

[963] That goes back, you know, 150 years.

[964] And then Nestrov, I think, has made a major contribution with this idea.

[965] So like you said, gradients themselves are, in some sense, mysterious.

[966] Yeah.

[967] They're not as trivial as mathematically speaking.

[968] The coordinate descent is more of a trivial when you just pick one of the coordinates and go down the one.

[969] That's how our human minds think.

[970] And gradients are not that easy for our human mind to grapple with.

[971] an absurd question but what is statistics so here it's a little bit it's somewhere between math and science and technology it's somewhere in that convex hole so it's a set of principles that allow you to make inferences that have got some reason to be believed and also principles allow you to make decisions where you can have some reason to believe you're not going to make errors so all that requires some assumptions about what do you mean by an error what do you mean by you know the probabilities And, but, you know, after you start making some of those assumptions, you're led to conclusions that, yes, I can guarantee that, you know, if you do this in this way, your probability making error will be small.

[972] Your probability of continuing to not make errors over time will be small.

[973] And probability you found something that's real will be small, will be high.

[974] So decision making is a big part.

[975] So making is a big part, yeah.

[976] So the original, so statistics, you know, short history was that, you know, You know, it sort of goes back as a formal discipline, you know, 250 years or so.

[977] It was called inverse probability, because around that era, probability was developed, sort of especially to explain gambling situations.

[978] Of course.

[979] Interesting.

[980] So you would say, well, given the state of nature is this, there's a certain roulette board that has a certain mechanism and it, what kind of outcomes do I expect to see?

[981] And especially if I do things long, long amounts of time, what outcomes will I see?

[982] And the physicists start to pay attention to this.

[983] And then people say, well, given, let's turn the problem around.

[984] What if I saw certain outcomes, could I infer what the underlying mechanism was?

[985] That's an inverse problem.

[986] And in fact, for quite a while, statistics was called inverse probability.

[987] That was the name of the field.

[988] And I believe that it was Laplace, who was working in Napoleon's government, who was trying, who needed to do a census of France, learn about the people there.

[989] So he went and gathered data, and he analyzed that data to determine policy.

[990] and said, well, let's call this field that does this kind of thing, statistics, because the word state is in there.

[991] In French, that's eta.

[992] But, you know, it's the study of data for the state.

[993] So anyway, that caught on, and it's been called statistics ever since.

[994] But by the time it got formalized, it was sort of in the 30s.

[995] And around that time, there was game theory and decision theory developed nearby.

[996] people in that era didn't think of themselves as either computer science or statistics or controlled or econ.

[997] They were all the above.

[998] And so, you know, von Neumann is developing game theory, but also thinking of that as decision theory.

[999] Wald is an econometrician developing decision theory and then, you know, turn that into statistics.

[1000] And so it's all about here's a, here's not just data and you analyze it.

[1001] Here's a loss function.

[1002] Here's what you care about.

[1003] Here's the question you're trying to ask.

[1004] Here is a probability model.

[1005] And here is the risk you will face if you make certain decisions and to this day in most advanced statistical curricula you teach decision theory is the starting point and it branches out into the two branches of basine and frequentist but that's it's all about decisions in statistics what is the most beautiful mysterious maybe surprising idea that you've come across uh yeah good question um i mean there's a bunch of surprising ones.

[1006] There's something that's way too technical for this thing, but something called James Stein estimation, which is kind of surprising and really takes time to wrap your head around.

[1007] Can you try to maybe...

[1008] I think I don't even want to try.

[1009] Let me just say a colleague at Stephen Stigler at University of Chicago wrote a really beautiful paper on James Stein Estimation, which helps to, its views a paradox, it kind of defeats the mind's attempts to understand it, but you can, and Steve has a nice perspective on that.

[1010] there so one of the troubles with statistics is that it's like in physics that are in quantum physics you have multiple interpretations there's a wave and particle duality in physics and you get used to that over time but it still kind of haunts you that you don't really you know quite understand the relationship the electrons a wave and electrons a particle well the same thing happens here there's Bayesian ways of thinking in frequentist and they are different they they sometimes become sort of the same in practice, but they are philosophically different.

[1011] And then in some practice, they are not the same at all.

[1012] They give you rather different answers.

[1013] And so it is very much like wave and particle duality.

[1014] And that is something you have to kind of get used to in the field.

[1015] Can you define Bayesian and a frequentist?

[1016] Yeah, in decision theory, you can make, I have a, like I have a video that people could see.

[1017] It's called, are you a Bayesian or a frequentist and kind of help try to make it really clear.

[1018] It comes from decision theory.

[1019] So, you know, decision theory, you're talking about loss functions, which are a function of data, X, and parameter theta.

[1020] They're a function of two arguments, okay?

[1021] Neither one of those arguments is known.

[1022] You don't know the data a priori, it's random, and the parameter's unknown, all right?

[1023] So you have this function of two things you don't know, and you're trying to say, I want that function to be small.

[1024] I want small loss, right?

[1025] Well, what are you going to do?

[1026] So you sort of say, well, I'm going to average over these quantities or maximize over them or something so that, you know, I turn that uncertainty into something certain.

[1027] So you could look at the first argument and average over it, or you could look at the second argument average over it.

[1028] That's Bayesian Frequentus.

[1029] So the Frequantus says, I'm going to look at the X, the data, and I'm going to take that as random, and I'm going to average over the distribution.

[1030] So I take the expectation of loss under X. Theta is held fixed, right?

[1031] That's called the risk.

[1032] And so it's looking at all the data sets you could get, right, and saying, how well, will a certain procedure do under all those data sets?

[1033] That's called a frequent as guarantee.

[1034] So I think of this is very appropriate when you're building a piece of software and you're shipping it out there and people are to use it on all kinds of data sets.

[1035] You want to have a stamp, a guarantee on it that has people run it on many, many datasets that you never even thought about that 95 % of the time it will do the right thing.

[1036] Perfectly reasonable.

[1037] The Bayesian perspective says, well, no, I'm going to look at the other argument of the loss function, the theta part.

[1038] Okay.

[1039] That's unknown and I'm uncertain about it.

[1040] So I could have my own personal probability for what it is.

[1041] You know, how many tall people are there out there?

[1042] I'm trying to infer the average height of the population.

[1043] Well, I have an idea roughly what the height is.

[1044] So I'm going to average over the theta.

[1045] So now that loss function has only now, again, one argument's gone.

[1046] Now it's a function of X. And that's what a Bayesian does is they say, well, let's just focus on the particular X we got, the data set we got.

[1047] We condition on that.

[1048] conditional on the X, I say something about my loss.

[1049] That's a Bayesian approach to things.

[1050] And the Bayesian will argue that it's not relevant to look at all the other data sets you could have gotten and average over them, the frequentest approach.

[1051] It's really only the data set you got, right?

[1052] And I do agree with that, especially in situations where you're working with a scientist, you can learn a lot about the domain, and you're really only focused on certain kinds of data, and you've gathered your data, and you make inferences.

[1053] I don't agree with it, though, in the sense that there are needs for frequent as guarantees.

[1054] You're writing software, people are using it out there, you want to say something.

[1055] So these two things have got to fight each other a little bit, but they have to blend.

[1056] So long story short, there's a set of ideas that are right in the middle.

[1057] They're called empirical bays.

[1058] And empirical bays sort of starts with the Bayesian framework.

[1059] It's kind of arguably philosophically more, you know, reasonable and kosher.

[1060] write down a bunch of the math that kind of flows from that and then realize there's a bunch of things you don't know because it's the real world and you don't know everything so you're uncertain about certain quantities.

[1061] At that point, ask, is there a reasonable way to plug in an estimate for those things?

[1062] And in some cases, there's quite a reasonable thing to do to plug in.

[1063] There's a natural thing you can observe in the world that you can plug in and then do a little bit more mathematics and assure yourself it's really good.

[1064] So based on math or based on human expertise, what are good...

[1065] They're both going in.

[1066] The Bayesian framework allows you to put a lot of human expertise in.

[1067] But the math kind of guides you along that path and then kind of reassures at the end you could put that stamp of approval.

[1068] Under certain assumptions, this thing will work.

[1069] So you asked the question, what's my favorite, you know, what's the most surprising nice idea?

[1070] So one that is more accessible is something called false discovery rate, which is, you know, you're making not just one hypothesis test or making one decision, you're making a whole bag of them.

[1071] And in that bag, decisions, you look at the ones where you made a discovery.

[1072] You announced that something interesting that happened.

[1073] That's going to be some subset of your big bag.

[1074] In the ones you made a discovery, which subset of those are bad?

[1075] There are false, false discoveries.

[1076] You like the fraction of your false discoveries among your discoveries to be small.

[1077] That's a different criterion than accuracy or precision or recall or sensitivity and specificity.

[1078] It's a different quantity.

[1079] Those latter ones that are almost all of them have more of a frequentest flavor.

[1080] They say, given the truth is that the null hypothesis is true, here's what accuracy I would get.

[1081] Or given that the alternative is true, here's what I would get.

[1082] So it's kind of going forward from the state of nature to the data.

[1083] The Bayesian goes the other direction from the data back to the state of nature.

[1084] And that's actually what false discovery rate is.

[1085] It says, given you made a discovery, okay, that's conditioned on your data.

[1086] What's the probability of the hypothesis?

[1087] It's going the other.

[1088] the direction.

[1089] And so the classical frequency is looking at that.

[1090] So I can't know that there's some priors needed in that.

[1091] And the empirical Bayesian goes ahead and plows forward and starts writing down these formulas and realizes at some point some of those things can actually be estimated in a reasonable way.

[1092] And so it's kind of it's a beautiful set of ideas.

[1093] So this kind of line of argument has come out.

[1094] It's not certainly mine, but it sort of came out from Robbins around 1960.

[1095] Brad Efron has written beautifully about this in various papers and books.

[1096] And the FDR is, you know, Benyamini in Israel, John Story did this Bayesian interpretation and so on.

[1097] So I've just absorbed these things over the years and find it a very healthy way to think about statistics.

[1098] Let me ask you about intelligence to jump slightly back out into philosophy, perhaps.

[1099] you said that maybe you can elaborate but you said that defining just even the question of what is intelligence is a very difficult question is that a useful question do you think will one day understand the fundamentals of human intelligence and what it means you know have good benchmarks for general intelligence that we put before our machines so I don't work on these topics so much you're really asking a question know, for a psychologist, really, and I studied some, but I don't consider myself, at least an expert at this point.

[1100] You know, a psychologist aims to understand human intelligence, right?

[1101] And I think maybe the psychologist I know are fairly humble about this.

[1102] They might try to understand how a baby understands, you know, whether something's a solid or liquid, or whether something's hidden or not, and maybe how, you know, child starts to learn the meaning of certain words, what's a verb, what's a noun, and also, you know, slowly, but, you know, slowly, but surely trying to figure out things.

[1103] But humans' ability to take a really complicated environment, reason about it, abstract about it, find the right abstractions, communicate about it, interact, and so on, is just, you know, really staggeringly rich and complicated.

[1104] And so, you know, I think in all humidally, we don't think we're kind of aiming for that in the near future.

[1105] Certainly a psychologist doing experiments with babies in the lab or with people talking has a much more limited aspiration.

[1106] And, you know, Kahnem and Tversky would look at our reasoning and they're not deeply understanding all the how we do our reasoning, but they're sort of saying, here's some, here's some oddities about the reasoning and some things you need to think about it.

[1107] But also, as I emphasize in some things I've been writing about, you know, AI, the revolution hasn't happened yet.

[1108] Yeah.

[1109] Great blockposts.

[1110] I've been emphasizing that, you know, if you step back and look at intelligent systems of any kind and whatever you mean by intelligence, it's not just the humans or the animals or, you know, the plants or whatever, you know, so a market that brings goods into a city, you know, food to restaurants or something every day, is a system.

[1111] It's a decentralized set of decisions.

[1112] Looking at it from far enough away, it's just like a collection of neurons.

[1113] Every one, every neuron is making its own little decisions, presumably in some way.

[1114] And if you step back enough, every little part of an economic system is making it solid of his decisions.

[1115] And just like with a brain, who knows what, any individual neuron does, and know what the overall goal is, right?

[1116] But something happens at some aggregate level.

[1117] Same thing with the economy.

[1118] People eat in a city.

[1119] And it's robust.

[1120] It works at all scales, small villages to big cities.

[1121] It's been working for thousands of years.

[1122] It works rain or shine.

[1123] So it's adaptive.

[1124] So all the kind of, you know, those are adjectives.

[1125] One tends to apply to intelligent systems.

[1126] Robust, adaptive, you know, you don't need to keep adjusting it.

[1127] It's self -healing, whatever.

[1128] Plus, not perfect.

[1129] You know, intelligences are never perfect.

[1130] And markets are not perfect.

[1131] But I do not believe in this area that you can not, that you, can say, well, our computers are, our humans are smart, but, you know, no markets are not, more markets are.

[1132] So they are intelligent.

[1133] Now, we humans didn't evolve to be markets.

[1134] We've been participating in them, right?

[1135] But we are not ourselves a market per se.

[1136] The neurons could be viewed as a market.

[1137] You can.

[1138] There's economic, you know, neuroscience kind of perspective.

[1139] That's interesting to pursue all that.

[1140] The point, though, is that if you were to study humans and really be the world's best psychologist studied for thousands of years and come up with the theory of human intelligence, you might have never discovered principles of markets, you know, supply demand curves and, you know, matching and auctions and all that.

[1141] Those are real principles and they lead to a form of intelligence that's not maybe human intelligence.

[1142] It's arguably another kind of intelligence.

[1143] There probably are third kinds of intelligence or fourth that none of us are really thinking too much about right now.

[1144] So if you really, and all those are relevant to computer systems in the future.

[1145] Certainly the market one is relevant right now, whereas understanding of human intelligence is not so clear that it's relevant right now.

[1146] now, probably not.

[1147] So if you want general intelligence, whatever one means by that or, you know, understanding intelligence in a deep sense and all that, it is definitely has to be not just human intelligence.

[1148] It's got to be this broader thing.

[1149] And that's not a mystery.

[1150] Markets are intelligence.

[1151] So, you know, it's definitely not just a philosophical stance to say, we've got to move beyond human intelligence.

[1152] That sounds ridiculous.

[1153] Yeah.

[1154] But it's not.

[1155] And in that block boils you define different kinds of like intelligent infrastructure, I, I, which I really like.

[1156] because some of the concept you've just been describing, do you see ourselves, if we see Earth, human civilization is a single organism, do you think the intelligence of that organism when you think from a perspective of markets and intelligence infrastructure is increasing?

[1157] Is it increasing linearly?

[1158] Is it increasing exponentially?

[1159] What do you think the future of that intelligence?

[1160] I don't know.

[1161] I don't tend to answer questions like that because, you know, that's science.

[1162] I was hoping to catch you off guard.

[1163] well again because you said it's so far in the future it's fun to ask and you'll probably you know like you said predicting the future is really nearly impossible but say as an axiom one day we create a human level of superhuman level intelligent not the scale of markets but the scale of an individual what do you think is what do you think it would take to do that or maybe to ask another question is how would that system be different than the biological human beings that we see around us today?

[1164] Is it possible to say anything interesting to that question or is it just a stupid question?

[1165] It's not stupid question, but it's science fiction.

[1166] Science fiction.

[1167] And so I'm totally happy to read science fiction and think about it from my own life.

[1168] I love the, there was this like brain and a vat kind of, you know, a little thing that people were talking about when I was a student.

[1169] I remember, you know, imagine that, you know, between your brain and your body, there's a, you know, there's a bunch of wires, right?

[1170] And suppose that every one of them was replaced with a literal wire.

[1171] And then suppose that wire was turned in actually a little wireless, you know, there's a receiver and sender.

[1172] So the brain has got all the senders and receiver, you know, on all of its exiting, you know, axons and all the dendrites down in the body are replaced with senders and receivers.

[1173] Now you could move the body off somewhere and put the brain in a vat.

[1174] Right.

[1175] And then you could do things like start killing off those centers or receivers one by one.

[1176] And after you've killed off all of them, where is that person?

[1177] You know, they thought they were out in the body walking around the world and they moved on.

[1178] So those are science fiction things.

[1179] Those are fun to think about.

[1180] It's just intriguing about what is thought, where is it and all that.

[1181] And I think every 18 -year -old, it's to take philosophy classes and think about these things.

[1182] And I think that everyone should think about what could happen in society that's kind of bad and all that.

[1183] But I really don't think that's the right thing for most of us that are my age group to be doing and thinking about.

[1184] But I really think that we have so many more present, you know, for challenges and dangers and real things to build and all that, such that, you know, spending too much time on science fiction, at least in public fora like this, I think, is not what we should be doing.

[1185] Maybe over beers in private.

[1186] That's right.

[1187] I'm welcome.

[1188] I'm not going to broadcast where I have beers because this is going to go on Facebook, and I don't want a lot of people showing up there.

[1189] But, yeah.

[1190] I love Facebook, Twitter, Amazon, YouTube.

[1191] I'm optimistic and hopeful, but maybe I don't have grounds for such optimism and hope.

[1192] Let me ask, you've mentored some of the brightest, sort of some of the seminal figures in the field.

[1193] Can you give advice to people who undergraduates today?

[1194] What does it take to take, you know, advice on their job?

[1195] journey.

[1196] If they're interested in machine learning, in AI, in the ideas of markets from economics and psychology and all the kinds of things that you're exploring, what steps should they take on that journey?

[1197] Well, yeah, first of all, the doors open, and second, it's a journey.

[1198] I like your language there.

[1199] It is not that you're so brilliant and you have great, brilliant ideas, and therefore that's just, you know, that's how you have success or that's how you enter into the field.

[1200] it's that you apprentice yourself you you spend a lot of time you work on hard things you try and pull back and you be as broad as you can you talk to lots people and it's like entering any kind of a creative community there's years that are needed and human connections are critical to it so you know I think about you know being a musician or being an artist or something you don't just you know immediately from day one you know you're a genius and therefore you do it.

[1201] No, you, you know, practice really, really hard on basics, and you be humble about where you are and then you realize you'll never be an expert on everything.

[1202] So you kind of pick, and there's a lot of randomness and a lot of kind of luck, but luck just kind of picks out which branch of the tree you go down, but you'll go down some branch.

[1203] So, yeah, it's a community.

[1204] So graduate school is, I still think, is one of the wonderful phenomena that we have in our world.

[1205] It's very much about apprenticeship with an advisor.

[1206] It's very much about a group of people you belong to.

[1207] It's a four or five year process, so it's plenty of time to start from kind of nothing to come up to something, you know, more expertise and then start to have your own creativity start to flower, even surprising to your own self.

[1208] And it's a very cooperative endeavor.

[1209] It's, I think a lot of people think of science is highly competitive, and I think in some other fields it might be more so.

[1210] And here it's way more cooperative than you might imagine.

[1211] And people are always teaching each other with something.

[1212] And people are always more than happy to be clear that.

[1213] So I feel I'm an expert on certain kind of things, but I'm very much not expert on lots of other things.

[1214] And a lot of them are relevant.

[1215] And a lot of them are, I should know, but it should in some time, you know, you don't.

[1216] So I'm always willing to reveal my ignorance to people around me so they can teach me things.

[1217] And I think a lot of us feel that way about our field.

[1218] So it's very cooperative.

[1219] I might add it's also very international.

[1220] because it's so cooperative.

[1221] We see no barriers.

[1222] And so that the nationalism that you see, especially in the current era and everything, is just at odds with the way that most of us think about what we're doing here, where this is a human endeavor and we cooperate and are very much trying to do it together for the benefit of everybody.

[1223] So last question, where and how and why did you learn French and which language is more beautiful, English or French?

[1224] Great question.

[1225] So, first of all, I think Italian is actually more beautiful than French and English, and I also speak that.

[1226] So I'm married to an Italian, and I have kids, and we speak Italian.

[1227] Anyway, all kidding aside, every language allows you to express things a bit differently.

[1228] And it is one of the great fun things to do in life is to explore those things.

[1229] So, in fact, when I kids or teens or college students ask me what they study, I say, well, do what your heart, where your heart is, certainly do a a lot of math.

[1230] Math is good for everybody, but do some poetry and do some history and do some language too.

[1231] Throughout your life, you don't want to be a thinking person.

[1232] You'll want to have done that.

[1233] For me, yeah, French I learned when I was, I'd say, a late teen.

[1234] I was living in the middle of the country in Kansas, and not much was going on in Kansas, with all due respect to Kansas.

[1235] And so my parents happened to have some French books on the shelf and just in my boredom, I pulled them down and I found this is fun, and I kind of learned the language by reading, and when I first heard it spoken, I had no idea what was being spoken, but I realized I had somehow knew it from some previous life, and so I made the connection.

[1236] But then, you know, I traveled, and just I love to go beyond my own barriers and my own comfort or whatever, and I found myself on trains in France next to, say, older people who had, you know, lived the whole life of their own, and the ability to communicate with them was special and ability to also see myself in other people's shoes and have empathy and kind of work on that.

[1237] Language is part of that.

[1238] So after that kind of experience and also embedding myself in French culture, which is quite amazing.

[1239] Languages are rich, not just because there's something inherently beautiful about it, but it's all the creativity that went into it.

[1240] So I learned a lot of songs, read poems, read books.

[1241] And then I was here actually at MIT, where we're doing the podcast today, and a young professor, you know, not yet married and, you know, not having a lot of friends in the area.

[1242] So I just didn't have, I was going kind of a board person.

[1243] I said, I heard a lot of Italians around there.

[1244] It happened to be a lot of Italians at MIT, an Italian professor for some reason.

[1245] And so I was kind of vaguely understanding what they were talking about.

[1246] I said, well, I should learn this language too.

[1247] So I did.

[1248] And then later met my spouse and, you know, Italian became a more important part of my life.

[1249] But I go to China a lot these days.

[1250] I go to Asia.

[1251] I go to Europe.

[1252] And every time I go, I'm amazed by the richness of human experience.

[1253] And the people don't have any idea if you haven't traveled, kind of how amazingly rich.

[1254] And I love the diversity.

[1255] It's not just a buzzword to me. It really means something.

[1256] I love the, you know, embed myself with other people's experiences.

[1257] And so, yeah, learning language is a big part of that.

[1258] I think I've said in some interview at some point that if I had millions of dollars and infinite time, whatever, what would you really work on if you really wanted to do AI?

[1259] And for me, that is natural language and really done right.

[1260] You know, deep understanding of language.

[1261] That's to me an amazingly interesting scientific challenge.

[1262] One were very far away.

[1263] One were very far away.

[1264] But good natural language, people are kind of really invested in that.

[1265] I think a lot of them see, that's where the core of AI is.

[1266] If you understand that, you really help human communication.

[1267] You understand something about the human mind, the semantics that come out of the human mind.

[1268] And I agree.

[1269] I think that will be such a long time.

[1270] So I didn't do that in my career just because I kind of, I was behind in the early days.

[1271] I didn't kind of know enough of that stuff.

[1272] I was at MIT.

[1273] I didn't learn much language.

[1274] And it was too late at some point to kind of spend a whole career doing that.

[1275] But I admire that field.

[1276] And so in my little way, by learning language, you know, kind of that part of my brain has been trained up.

[1277] Yeah, and was right.

[1278] You truly are the Miles Davis and machine learning.

[1279] I don't think there's a better place than it.

[1280] Mike, it was a huge honor talking to you today.

[1281] Merci Bucco.

[1282] All right.

[1283] It's been my pleasure.

[1284] Thank you.

[1285] Thanks for listening to this conversation with Michael I. Jordan.

[1286] And thank you to our presenting sponsor, Cash App.

[1287] Download it, use code Lex Podcast.

[1288] You'll get $10 and $10 will go to first, an organization that inspires and educates young minds to become science and technology innovators of tomorrow.

[1289] If you enjoy this podcast, subscribe on YouTube, give it five stars on Apple Podcasts, support it on Patreon, or simply connect with me on Twitter at Lex Friedman.

[1290] And now, let me leave you with some words of wisdom from Michael I. Jordan from his blog post titled Artificial Intelligence, The Revolution Hasn't Happened Yet, calling for broadening the scope of the AI field.

[1291] We should embrace the fact that what we are witnessing is the creation of a new branch of engineering.

[1292] The term of engineering is often invoked in a narrow sense, in academia and beyond, with overtones of cold, affectless machinery, and negative connotations of loss of control by humans.

[1293] But an engineering discipline can be what we wanted to be.

[1294] In the current era, we have a real opportunity to conceive of something historically new, a human -centric engineering discipline.

[1295] I will resist giving this emerging discipline a nature.

[1296] I will resist giving this emerging discipline a name, but if the acronym AI continues to be used, let's be aware of the very real limitations of this placeholder.

[1297] Let's broaden our scope, tone down the hype, and recognize the serious challenges ahead.

[1298] Thank you for listening and hope to see you next time.