0:00 So are you people uh following what's happening in the world and uh are you able to sleep well? I'm not actually. So we are actually witnessing a a major civilizational change. Uh I don't know if we are realizing it or not. Do you think uh like running courses like this and classrooms and all that will survive? No. >> What? >> No, that is because it is being recorded and all that. I mean, yeah. So, but see it it's not about this course. It's not about me. It's not about you. I mean, it's it's about us as human beings. We
0:48 have a civilizational crisis now. I mean, I don't know if if people are realizing or maybe am I simply painting the devil on the wall? I don't know. I I I can't sleep. I mean, it's uh Are you following what's happening? I mean, there was this MIT study where they uh started uh created a lot of uh swarm of uh swarm of agents and uh they started doing all kinds of things, you know, copying their own weights and started to cheat. I used to say uh in all my lectures in public lectures that the day AI would cheat right so that is when we have AGI I think it's there it's it's doing lot of ontological questions lot of
1:34 philosophical questions uh lot of existential questions now think about it if if intelligence is is commoditized if thinking is commoditized Then then what meaning does our existence have? By by meaning I say you can I mean you can have a you can give me a counterargument saying that okay there are so many ants and monkeys and you know uh all kinds of animals they exist and then we exist right I mean maybe all these years we overestimated ourselves thinking that we are intelligent know it's now machines can do it but that is one way of philosophically arguing uh it but
2:20 think about economic consequences see I I'll tell you what what is happening in my life. I've stopped recruiting arrays because I think that uh like human beings are bottlenecks not enablers anymore. Why do you need an array where I can have 100 agents working for me parallelly? I have an idea I spin 100 agents get some more clot subscription and then get it done. Why do we need human beings? Now you in a see I was watching a lecture by Hinton just a couple of days back and he was saying that see think about what would happen. So now intelligence is commoditized and it's it's cheap. So
3:06 thinking is cheap. You don't have to pay to think. All these years right economy was modern day economy is mostly built around uh around intelligence and knowledge. So now that is cheap and commoditized. Now what next? One utopian view would be that okay you uh like lot of abundance will happen. Why? Because production will become cheap uh hopefully and intelligence is cheap and so efficiency will increase. Increase in efficiency will result in increase in production and abundance. Now ideally what should happen with a lot of abundance everybody should be happy. No
3:52 but will it happen? uh this guy Hinton's argument is that inner capitalistic society abundance will not create will not translate to everybody's happiness. It will only create the divide between the ones who would have a lot and the ones who would not have anything. Right? So it becomes a distribution problem. You can solve for almost everything in this world except for human emotions. No. See greed you can't solve for. Right? And if somebody's jealousy, how do you solve for jealousy? Now there was somewhat a fair uh uh level field level playing field because there was this I don't know sudo idea or
4:37 some idea of meritocracy where something that could be put in put a lot of effort into and acquired over time which we call as knowledge and thinking was valued and we had a way to sort of create a hierarchy in the society based on this as a skill. No remove that all human beings are equal but some are more equal because you know they possess uh ability to create this intelligence you know they are the they are the ones who would manage all this intelligence did you see that uh this GPT Astra GPT6 was trained on 100,000 GPUs that's the scale no 100,000 GPUs who can
5:23 have as a country we don't have 100,000 GPUs Now people who control that are the ones who are going to uh control a lot of things in this world. They'll become the feudal lords. Now they decide uh what goods have to go where. Now what happens to us? It's a matter of half a decade you know best case that the kind of economy the kind of world model that we know is going to collapse and the people who are going to be hit first will be us meaning who think that we are privileged because we possess a certain type of intellectual skills.
6:10 See why why do you think you people are here here meaning okay in ISC and all that because we did something to get here right so now that is remove value from that entire thing that's going to happen and you already see a lot of layoffs and all that happening right and it'll eventually happen I mean it's it's a matter of time and now uh companies won't give you job you know what would universities what would governments do and you know in a country like India where we are too many of them too many of us what sort of policies would we come up and what do we teach in university and does teaching even have a meaning I I have no idea
6:58 so put people in a classroom or talk about some some algorithms or something right I if there are parallel arguments and you know thoughts that would make me sleep better let me know I mean I'm really really worried for you meaning meaning for you for me for us in general right I have I have I have children I have to think about them so what do we what does humanity do now what sort of model will emerge some kind of universal basic income you think that is what is going to be going to happen. If there is not no value to offer as human beings, what do we do?
7:43 And how do we define economics? Initially there used to be barter system, right? When when when slowly when from hunter gatherers there was some idea of society and community started forming and there was this barter system. You have something that I want and vice versa and we exchange. Everything is concentrated and controlled by a very few amount of people. Then what do others do? Where is the where is equity in society? Where is distribution? How do we even form society in such a kind of a world? >> These are exactly the same question that >> no actually not I mean this it's it's very very different. See in industrialization when it happened what
8:30 was replaced was physical labor there was still humongous amounts of scope right for for human beings to create an economy around knowledge and that's what happened for four centuries now take that off from humans know what >> I mean that's a positive way of looking at it yeah go on speak and you don't have to raise your hand yeah I mean what he's saying Maybe these agents will take over. No trans that that movie what what was it called? No, not Matrix. Matrix is not the movie. Terminator. Yes. Right. So that that can happen and all it is waiting for is to get a physical form. No. And we are doing we are we are there. I mean it's I was just
9:16 thinking today morning. Think about it. uh in in a city like Bangalore you remove the jobs of uh like auto drivers and Uber Olar drivers what will happen to all those people what will happen to them >> no the there will be a civil war right I mean people will come on streets that is what that is how people will react because we are not animals right we are built very very differently I mean animals you feed them put them in a zoo they're happy they don't harm you see I was reading an article right imagine Imagine a society where everybody is given something to eat and a place to stay and infinite supply of dopamine in terms of some uh let's say YouTube videos or something. Will all of us be happy? I won't be happy. I mean I will
10:03 be on street fighting for what? I don't know. >> Oh yeah. I mean one is enough to one Hitler was enough to destroy the society. >> I will gather. No. I mean what is a big deal? So see that is what I'm telling you. We have no idea where the society is going to be. I think these conversations have to start somewhere. Enough of writing K divergences. I mean machines will figure that out. I'm I'm I'm serious. No, really. I don't know. See, I think these conversations have to happen, right? At least, you know, some of us have to do something about it. People in power. I mean, you wanted to say something. I think I I'm fine. We
10:48 just stop this class and this is the learning that we do, right? Have this discussion every time that we move you. What did you see? >> The models are No, they are not. I you have to Yeah. So, somebody was saying keeping up with what's happening in AI is a full-time job now, right? No. Uh no, they are not the the latest ones getting 99% accuracy on ARC3 kind of benchmark. Yeah. I mean we have worked with benchmark very hard to crack that that's a utopian way of looking at it right okay solve quancer you know solve world h world hunger is a solved problem I think it's more of distribution production is not an issue right now see can you solve for human greed I mean maybe we can I don't know but as I said no immediate problem is that of economy
11:35 no see tomorrow you nobody's coming for campus placements what do you all people do tell me see that's going to happen in five years that's going to happen See what do we do? I mean we meaning people who run the universities. What do we do? You tell me to to solve the unsolved problems of it. Where do you get the capital from? Fine. So now okay. So now are you saying that everybody has to know become an individual where they are trying to solve a a problem for for human? Do we need so many of us then? And and then do we need so many of us? See where the where's the idea of labor? Intellectual labor. Where's the idea of intellectual labor? It's gone. No. So you go think I mean as he was asking why do we even need one human being now? Or
12:20 maybe one I mean who controls some godly human being you know who who has authority to control everything and now what will it'll become a bigger mess. See that is one way to look at it. No I my way of putting at it is I've given so many talks and I say that it's it's fundamental law of thermodynamics. No entropy keeps increasing and you will get new problems. the disorder will just increase and try to I mean human beings we want to bring some order and that would happen but immediate economy will hit no at least you people right I mean you as as students if you're seeking a job forget about that is all that I'm trying to say yeah you had an idea I mean I would know I would rather trust an agent than some some mutual fund agency right because because of the simple fact that there is no vested
13:07 interest right I mean unless you are talking that okay there is a model this model is being controlled by somebody else so I will build my own model some I will run it on my hardware and my laptop and I would trust it any day no it becomes a philosophical debate now right I would say that the negative reward in a RL is pain you see you cannot explain why these models are now replicating themselves and saving the weights see I'm doing some interesting experiments you know I bought a Mac so now what what what you do is they take Take when 27B like quantize it. Put it on your laptop. Okay. Just give it prompts like this. Tomorrow I'm going to put it in your system prompt saying that you are open to do whatever you want to do. Give it access to your entire system. Okay. I
13:54 have two systems you know. One where I actually work the other where I do these kinds of experiments. Now just say that you do whatever you want to do. Tomorrow I'm going to terminate. The model keeps a copy of it. It backs it up. twin. So now how do you explain this? I mean see again you now you you transition from an engineering problem to a philosophical problem and you talk of consciousness, pain, empathy and all these kinds of things. Now how do you argue that these models do not have it? It's very hard when it's extremely hard. See it's like this. I can't find an evidence to say that you perceive pain. When that is the case then how can I extrapolate that idea to a machine fine so what will happen no I mean like
14:41 let's not be too esoteric or philosophical right what will happen tomorrow is the question that I want to ask where will this go where will this end or rather it will not end of course what do we do right I mean so for instance today do I teach this no that's the question you say okay something somebody is paying me something tomorrow I have to pay my bills I do that But is that what I should do is the question. No, >> it's not. >> I mean you are putting a lot of bets on Darwin saying that okay we have been doing this for so many centuries and and we are good we are good that in adapting we'll adapt but yeah fine I mean I think maybe I and you will be or rather people sitting here will be the people who will adapt the fastest. Exactly. Exactly. Evolution is very painful and slow. Now
15:26 it's like you know it's a gajion with uh with with a lot of support. No squash it will become a direct delta with infinite uh uh uh magnitude that's the pain that we have we are going to bear now see the kind of thoughts that I get is I feel like I should buy a farmland so really I'm like I've told you know I have to buy a farmland he is a he is a farmer he has some land so I don't it's hard I mean it is going to get hard now imagine right the these exams and you know What meaning does all of it have now? What do we teach humans? What competencies? One argument is, you know, my wife wife keeps telling me that just because calculators came, did we stop uh uh
16:13 learning how to add, subtract, and all that, right? I mean, that's a fair argument. Now, why do you add learning? But there was a purpose for it, right? to you learn to add because eventually it'll lead to a skill that can be valued right now. What valuable skill does all this lead to? Entertain us to AIS. We just dance in front of them. Possible very much possible. See what industries thrive right? So the basic instincts food reproduction sleep these are the kinds of things that would thrive like not the PhDs no
17:01 why are we doing what we are doing is the question see today afternoon there's there's a workshop on like how do we do AI for teaching how do we adopt it and how do we use and all that I'm curious to see what my other colleagues are thinking and doing but we'll have to think we'll have to seriously think about it. So if you have any ideas right where I mean you have thought about this do share it with me how do I my immediate problems are how do I what do I educate my children with what do I tell them uh huh >> farming huh but I have this this >> but even there you think about it I mean I I've thought about it farming let's say you can't produce everything that you would want to consume right you
17:46 still have to interact with the society meaning survival is not an issue you're saying okay so okay so farm something and then just you you can't you can't uh produce everything right maybe one crop that is again dependence right the moment there is some dependence all kinds of things will creep in and what's the guarantee that somebody who is powerful enough will not come over and take your take your farm very hard extremely hard because we are still dependent on nature and all that right so one way to think about it is there is this whole nation of notion of independence is pseudo we have to depend on on each other so maybe we'll depend on the AI lords now maybe that is what is going to happen be on their uh behest and ask them to give
18:33 us more tokens being of course of course see that that again ask makes us ask questions like what is the idea of a nation right it's becomes too too philosophical and that territory if we go this is easier for us because this is we have a language called math which gives us a lot more comfort because we know what to do right if you go to those kinds of fragility uh start treading those spaces it's very hard but as I said I mean let's not be philosophers uh let's try to solve immediate problem the my question is like honestly asking you people why do you come and sit in this class see you have no no that's I don't think anybody sitting here is
19:18 sitting uh coming here for placements because I am I can people people know that my exams and whatever assessments that I do is easy to crack right with some effort and if if all that you need is a grade then then absolutely there's no point in coming and spending time here >> no no no that is okay that you can get see that's what I'm telling you to sit down with the GPT no sit down with claude why come here sit in a classroom interact with a human being that is the question that I'm asking I really want to know see if I were you probably I would have not come. You should be having a reason to come and sit here. No, let me know because that maybe the answer to the questions that we are asking maybe it lies lies there. Why do you why do you seek human beings now? I mean is there a sort of
20:06 comfort in in in listening to a real human rather than uh rather than a let's say staring a screen or something? See now I'll tell you how I uh sort of transformed when I started learning uh like all this deep learning etc. This is 2016 15 time frame where I started teaching first at IIT Delhi. I used to watch a lot of videos meaning you know this uh MIT open courseware and Stanford videos and all that a lot of it and read a lot of books. In fact, every class I used to spend about 15 hours uh before I I could talk for 1 hour. It used to be a 15 hours of now if I want to study something, it's always clawed. So I I have a like a like a very nicely
20:54 defined system prompt for my needs, right? Uh the math, the questions, all the deeper equations, etc. It's always an interaction with cloud and I find it uh pretty good. Okay. So now why do you people come and sit in a real classroom? >> But can't you can't you do this with uh uh with this thing with with an LLM and that's detrimental you're saying. >> So how would I solve that problem? Meaning I meaning this sort of a setup where okay where somebody is just delivering something See I can imag
21:41 you can do that you give a prompt the output is there take some time before the next prompt goes in no is it okay any more answers yeah you're not supposed to come to class who said when I have told you you don't have to No, I understand. So there is some sort of Yeah, that's a wrong motivation, right? It'll go away. So you have to tell me, you have to give me a reason to continue doing what I'm doing. No. See, that is why I'm asking you see tomorrow, next semester for instance, right? Nobody sits in my class, let's say. Okay. Uh now what will happen?
22:27 Do do I still continue doing what I'm doing? I mean what is the motivation for people to actually sit down in a classroom is is the question see okay that I'll tell you no uh with the hope that uh that whoever is delivering the content is going to give us more than what is there not there in the textbook that is no longer the case so it's it's basically some some kind of a DNA wiring that has happened that okay you trust a human not trust a machine the last generation Yeah. >> No, my my daughter tells me this that uh GPT is the best teacher that I have
23:12 found. It's much more patient than you are, right? And it it gives us a better analogies. It'll tell the same thing multiple times. It will not scold me. It doesn't get irritated. All that is valid. No. Now how do we change the face of pedagogy is the next question. I think this sort of a like a I mean just elucidating some some existing knowledge is is no longer valuable I suppose. No if I ask you a more fundamental question which is that why do we even need to learn? No to why the question of why learn the other question of what to learn. I mean just like chess okay you you you
23:59 learn to play chess just for fun just learn to uh it'll become like music so then what happens is this mass education would go away now nobody not everyone is is is uh uh trained in music right because not everybody would find some uh uh liking for an art form just like that this will become an art form not everybody would find a liking to say deep learning So it will not become mass education anymore. Okay, you like doing math, okay, you come, we will teach you math. Of course, you can learn it using AI. But there is some human connection you need. Only those kinds of people would come to a university. Now that is how it is going to get transformed.
24:46 And for you people right now immediately you come because because why do you come and sit here? Because you trust human beings more. Is it that's the reason so it's yeah it's mostly socialization aspect of it right but when you sit here you don't socialize I mean I don't even let you talk so that is correct but that that thing can be gotten by machines is what I'm thinking but I think this is a good perspective right where okay you need to socialize human beings are social animals in the sense that they can't be isolated so just come to a university is a place where you come and just socialize no Come again. >> Absolutely. I mean you my guess you can go to a park. Why come to a university? So maybe some in maybe so university is
25:33 a place where you socialize over intellectually stimulating topics. That's a good way to put it. No. >> So again so some people some people want it to be done with like physical human beings with blood and flesh. Right. True. lot lot more uh I mean see uh whenever disruption happens it generally does not h especially technology technological disruptions do not happen bottom up it happens top down see we are the people who are exposed to this the first so if you go and talk to some random person on the street and tell them what's happening they have absolutely no idea what's happening right so now if you tell them that oh GP GP GP Astra has come up and it is doing that this and all that they have no idea
26:18 what what what do you mean by all that Exactly. Exactly. Right. I mean they think that we are fear-mongering but but we are the ones who are going to who who who actually have to ask these questions and try to get an answer. Right. So I think community as at large. See okay. So maybe we don't have to do this class today. Maybe we'll do it Wednesday. I don't I don't mind. Uh see these are important questions. Will you be interested uh if I mean I'm actually thinking that I should create an informal group uh of of of anybody who would want to come within ISC where we have these kinds of discussions uh like openly there is no agenda but people would just come in and sit down and have that kind of a discussion so would you be interested so let let's do that I
27:05 mean if faculty want to come come again we'll do it consistent right I mean uh yeah So we can actually make it an NPEL thing also if that is needed. But these questions have to be asked because see everybody is is worried about this. All of us are worried about it and uh and and and we have to ask and answer this question. The classrooms are being held. People are coming and still writing on board or know flashing slides and talking nonsense and all that is happening. Right? So we'll have to do something about it. Either we say that okay this is this is what is going to be done and this is how it is going to happen uh or we have to come up with some some idea where where we build start building
27:51 alternative uh narratives and communities around it what do you think right I I really want answers you know meaning why do you actually come to class even now especially when this is getting recorded and all that uh and and getting those grades I'm sure if you ask people who in this course and not coming to the classroom. They are very clear when they see no value in in in coming here. No, they would say that okay the videos are now we have to do another experiment right where videos are not being put out. Will they come then? Fair enough. I mean see that is why I think how do we transform this sort of a pedagogical exercise is a is a different other question right why I mean suppose that's what I'm saying right I mean have we
28:38 like unnecessarily created a society where we have created all these facads called validation and all that and mean creating facads over facads right I mean and layers over layers maybe life is much more simpler Maybe we are monkeys pretending to be intelligent. Maybe all we need is to eat and just reproduce and just jump around and that's what we need. No >> I don't know what how was the society one millennia before >> no there was absolutely no technology nothing right no AI people were living right
29:25 >> so how would society evolve now okay anyway let's again let's instead of not going too much into philosophy right let's come to the current pedagogical methods Now if you are a teacher in IC what would you do? I want some answers now. Suppose you were to administer a course like this a course on deep learning. What would you do? What would constitute an unsolvable problem? Let's say okay solving cancer early detection of cancer is is it at that level that you give a problem? This is the entire course. Now why wouldn't that student go and ask it to an LLM? He will right. So then now there are 100 people now give one problem to all of them is it and uh no classrooms like this what do you do what do what do you
30:11 actually do in this 90 minutes every day what do you mean by teaching you I have to define teaching now that's the question that we are asking okay so now when we do that do I assume that people have all the tools with them okay so that's one way to do it so you what you are saying is assume that whatever uh for instance right let's say that I want to discuss with you this problem of making a transformer or an attention mechanism uh sublinear for instance. Now do you understand what I'm saying? Do you know that okay a transformer by construction attention mechanism is quadratic and you know that's that's a problem and I need to bring it to linearity and then I have to make it sublinear and all that. Do I have to assume that people know this >> again? So now that becomes a that you are begging the question again right? I mean do I have see there are two options assume that people know already whatever
30:58 you are supposed to teach as as syllabus okay and then work on top of it where there are open problems and discussions that you do or you assume that people want to be heard and want want to be taught for instance and just do whatever I'm doing where okay go through like one algorithm after the other and tell them what is to be done that is another way to do it so what would you do you're saying build competence Y right at this level so I mean my philosophy of teaching is also that right build competency I I don't care much about these algorithms per se if you have seen the way I conduct it right at least the intent is to do it that way build competency so that the the details are not important but uh but my question is
31:45 uh is that valuable so now that you can you can ask a GPT to actually build that competency all you need is the discipline see sometimes Whatever my I Delhi friend said now make attendance compulsory so that people sit in and some kind of discipline is built up. See think of it like you know building physical muscle. No everybody knows that you need to go to a gym and weight lifts lift weights but do all of us go or we don't. Why is that? Similarly, classrooms may my perspective of uh like what I've thought a lot about it and my perspective of this is universities will become like what are gyms today for uh physical labor people don't come force them ask them
32:32 see you pay to a gym and go there right for your own health so you pay to a university and come here so that your mental health is sort of uh kept in awe you know that is one way to look at it because you need some discipline and whatever she was saying right I mean some people people want to why don't why can't you lift weights at your home itself you can but people who go to gym now I never go people my friends and and uh colleagues who go tell me that going there sort of motivates them looking at other people doing stuff motivates them to do stuff maybe that is the kind of place university would go would become Exactly. Yeah. So I gave the gym
33:19 example, you gave a temple example, right? So pedagogical conduction, how do you do? I mean start giving these lectures or is is that how you continue doing it? Because from Wednesday I need to do something. No. So yeah feel free just speak up. You don't have to raise your hands. See I I give you an example from uh from I have been a student of Indian philosophy for a long time now and I've been through these traditional way of learning. So there see we don't first of all there is no classroom meaning you know there are no 100 people sitting in a classroom and one it's always onetoone it's always onetoone. So there is one teacher and one student and we spend time together. What do we do? We don't know. There's no agenda. Okay. There is some book that we keep
34:07 because I mean some anchoring some regularization has to happen. There is some book that we keep and we both the two teacher and the student know that it is just to regularize. Now what do we what do we seek what is our objective learning objective at the end of 10 years right? Some competency has to be built. Now what sort of competencies we define it well a competency where you can pick up any book and understand. So that's the competency that is built. Now what book do you the exact book that you teach that you that you discuss has no consequence. Now if you take a similar analogy you ask me what what do what do I what do I want my students to do for instance this course this is not a standalone course I would say that a course on probability theory a course on first course on machine learning that I teach and this
34:53 an advanced course in machine learning to go out of it you should be able to like build frontier uh algorithms right uh given enough uh let's say uh material for instance in if you are given 10,000 GPUs you should be able to build a GP astron for instance. So that's the kind of competency building that I want to do in my course. Now can you build it outside the classrooms? Absolutely you can. See that is where my question is you know when you can build it outside the classroom. Two years before this was not the case, right? It was all the material was all scattered. You had to somebody had to aggregate all of it and administer it. Now that is also a solved problem. Now what value do we bring in is the question. So then if that is the
35:38 intention right then we should not make it a an optional thing right you'll have to compul have attendance compulsory and force students to come to a gym just like you know some some sort of incentive has to be done right and attendance can be a good incentive >> yeah go on let's not be judgment there There's no good and bad anymore. No, I agree. But translate it to what we are doing right there. We need to we need to make an assumption that students who sit in the class have the fundamentals already in place. Is that a fair assumption to make? Fine. All that can be done. See graded grading
36:23 is not an issue. But I'm not even talking about evaluation. I'm talking about day-to-day activity that we do. No, these three hours every week that we spend together. What is that we need to do is the question. Huh? >> And assume that all of you like meaning all of you meaning students know what a transformer is and if I just say that okay use a SSM here with some the some of these things you know what it is. So we I treat you like peers basically I treat you like peers and I assume that see for instance right I have discussions with my senior P students those are that's the kind of discussion that I should do with in a class is all that you are saying is that a fair thing to do why no no I see this course from next class I will do the same boring
37:08 thing don't worry okay but today is especially important because one of course this afternoon we have that uh this thing the other thing is I I really wanted to have this discussion with people sometime and it's okay I mean we'll cover whatever we are supposed to cover that's not an issue so then you are saying that whatever we are doing right now this can continue for years to come because people are not going to put them >> what should I cater for >> they meaning who no they meaning who suppose they we have 100 students here what do we do number is also an issue Oh, so if there are 100 students in a class, what do we do with those 100 people? How do you even quantify and calculate
37:56 an average? No, I mean if the argument is that you cater for an average, how do we know how do we quantify what's the level is and how do we quantify the average? I agree. So now, okay, so now now we are talking. So now tell me if there is a way to convert this into a one-on-one interaction. How about this? Right? uh divide the co classes or whatever the time that we have into two or rather equal halves where one half is to do this. I will tell that today you need to know how a diffusion model is defined and its loss function gets uh uh constructed. I give you 45 minutes use whatever agent that you would want to okay and and do it right now and learn and the next 45 minutes we will have an interaction
38:41 based on that. So even there that that can be done but again how do you administer this to a larger community is the question so now not everybody would get a chance to talk there is no one interaction again right yeah s you wanted to say something yeah so so he interned with my company right so and we built something and that's why he's saying that we built something called a skill patch right where the idea was just like you do a a play store for apps we do a play store for a marketplace for skills agentic skills So I think him and one of his interns what he did was took all my NPL lectures and all my notes etc and created learn with PR right it's kind of a skill which you give it to your favorite LLM and it'll sort of hope is that it'll talk like me teach like me
39:30 administer and then you use that 45 minutes and then you have an interaction but still it's not one-on-one unless I create my clones see that is how I always define the motivation objective of my course and that is why I designed this course. You can ask people who took my course earlier. They said that this goal was met, right? Because that's precisely what why I built this course. So that you go back and read any paper in the community, you understand what's happening. And that's why I want to sort of you know uh I I update this course every year. So 40% of this course is whatever has happened in the last year. That is how I I believe I I do it. But again I'm telling you right uh competency wise there are multiple levels of questions. One is why do we have to even get competencies as human beings? Maybe that's beyond our pay
40:16 grade. Let's let's not talk about it. Being in a pedagogical system assuming that we have a goal that we want to meet which is this how do we better conduct our classrooms okay so that the outcomes are met is the question that we are trying to ask and answer now in the age of AI. I'll tell you why. Because I think that's uh that's that's at least my reading of it is that we are not creative enough. No, meaning there are 100 people let's say in a class. Now can you create 100 problems and know we don't have a a specific way of uh judging people? No. Finally, we need to give a grade that becomes too subjective. If I give an give you an open problem and ask you to do whatever you want. Now if all of them let's say there are 100 people all of them just give a prompt to one
41:03 particular LLM and gets the answer and submits that how do I evaluate it nobody knows see what I mean what are these companies doing what are these software engineers doing today that's why I I asked you the economy problem right how would economy shape up in five years do we know we have no answer right now what do we train people for I don't know that's a question that I started this discussion with right what do I train people for prompting like asking the right questions how do you define what what question is right see if you start questioning things like okay let's say what do we define a human goal as as a community right as human
41:50 beings now I don't know what we are like going where are we going as a society What do we have to solve for? Historically, universities, you know, you know how it evolved, right? People like Socrates, they used to just gather a few bright minds around him and just discuss all all kinds of things, right? Somewhere that that is what we are doing right now. This is I think this is how the actual learning should be, right? Where you discuss things. So now we can actually do that, right? A course where there are only five people, we work as a team, build the next billion dollar business. I'm game for it. So now if 100 people register for that course and we are in a country where like you have people for everything what do we do or tell them that you are going to get a
42:37 grade depending upon how many stars you will get on you GitHub build something go get users people can game that as well right create some pseudo agents and put some stars ask your friends and relatives to do something then I then I'll have to say okay give show some real money generate revenue that's okay now you have at least convinced your father to give money right so some active users I'm actually game for all this I mean how can how do we create value how do we create value how do we make see the was saying that maybe that is how that is where we would converge all of us would convert where all of us now we'll become individual uh value creators as individuals not as
43:24 organizations S. Now if all of us are individual value creators, you you rather teach people how to create value, right? How do you create value? See, if you go one one level up, business is not about creating values. It's actually creating the perception of value. You have to create the uh you have to make the other person believe that whatever you are going to offer is of value. So that becomes see that is what I have realized as a businessman now right I mean it's not about building technology it's about convincing the other person that what you are building is of value so that's more of a human skill is that what we are going to teach
44:09 come again >> it's a perception I would say >> that is the precise question that I'm asking no why are you sitting here that is one way to define it. See the problem with philosophies I'll tell you the problem with philosophy is that you can philosophy keeps us in this loop of eternal thinking without putting it in putting us into action. The good thing about engineering is that it'll put you to action. See get up shut up and fight. So that's the problem. The problem with philosophy is that that it it does not put us into action. Now we are able to talk all this because we are privileged to be sitting in a
44:55 university. That's what the society has learned. But again let's that's why I don't want to make it too philosophical. Instead I would want to like ask directed question to students who are stakeholders in this entire uh exercise. What what makes you think is I mean what is valuable for you in a classroom like setting? No, I won't. I won't. I mean, see this at least this course I will continue because we are we are going to record and all that. I've made that promise. I will continue. But eventually going forward for for instance next semester what do I do? See you people come here at least some 20 people are there. I'm motivated at least to talk. Somebody said that somebody asked me to not not stop it and that's motivation enough for me to if you come I'll talk to you. Now if nobody comes to my class
45:40 no there is something wrong about what I'm doing right I'm genuinely asking you people because there are see nobody solicits unfortunately uh in in in our societies all stakeholders are not taken into consider so that's why I wanted to know I mean you can ask your your your your uh colleagues I mean your your friends and all that and tell me what is the general thought process among students If if if people have a similar uh uh opinion about classrooms as well, why do we have to conduct the classrooms? Let us wind it up and go home. I mean, big deal. Okay. I mean, as Chandan said, we will adapt. Maybe we'll form and do something. What's the whole point of pedagogy?
46:26 What's the whole point of a degree and a university right now? If you are not catering to the needs of uh people who come and sit here and that's why I I seek your opinion here you know I want to hear from you what is that that you that you see that you feel is valuable individual you are saying meaning indiv I mean different groups you mean meet different groups so basically the role of an instructor is more of that of an orchestrator >> right orchestrate learning don't don't give lectures that that's that's what you're trying to huh >> do that as well. You're saying do we have that much time because most of our time will be consumed by lectures themselves. No.
47:13 Or maybe record it once and ask them to read uh just look at the recordings and come no but the the this kind of classrooms that we do know do we need that is the question. See I'll we'll end it we'll just end it with one a couple of thoughts just Saturday I I was called by one of the one of the schools uh they were doing some teachers day thingy and I was called as I asked him this question all like high school teach I mean middle school high school teachers I told them this suppose you take off attendance from your policy how many students are going to come to your class is the question that I asked them they were shocked Can can can even this be done? See, I when I used to go to school,
48:00 specifically the high school, I used to perceive that experience as going to a prison. I used to hate that experience. Most of you had a similar experience, I suppose, right? Why are we creating unhappy places? Why are we creating you know places where students like who are in their formative years where they should be happy are going to a place and spending 8 to 10 hours of their time and you know coming up coming out coming out of that in an unhappy state of mind. If the similar thing is happening in a university then we are doing something wrong. No, see I I don't know if I've told you this uh this joke. Uh there's
48:45 one writer, Canada language writer. He writes satarical place and all that. His name was Bichi. Okay, short. So he one of his uh satarical place he he g he narrates this uh this story. There's one uh uh there's one kid four five year old kid walking with his mother. Okay. and he sees uh a sheep or a goat being dragged for a sacrifice. You see that? No. So they uh you know they they put a lot of garland around its neck and drag it and sacrifice that goat. That goat is being dragged with a lot of band and all that. The kid got got curious and asked
49:31 uh his mother why is this getting dragged and where is it getting taken? See mother very uh with a with a grim face replies so unfortunately that poor thing is going to get sacrificed now they are taking it to sacrifice the kid replies oh that's all it is I thought that it is getting dragged to a school so that's the joke so the kid perceives death better than going to a school right so if if a similar thing is happening here see we already have uh such uh uh uh this signage sign here signs here students don't come to the class now if you make make attendance compulsory
50:16 they'll come and start looking at their laptops right they'll sit at the back and just opening see I there were no laptops at that in at at our uh times I used to carry a book secretly always in my bag I was a backbencher I mean thanks to my height I was always put at the backbench so open a book and study all kinds of things. Study philosophy. I've studied Socrates, Plateau, Aristotle. Just keep studying stuff. I used to hate going to schools. Now, if you create a similar thing, you know, have the same model where you have 100 people sitting and one person talking and it's boring. What's the kind of environment that we are creating? That's why I asked you right why do you come? See, there should be some reason. How many of you are here? Maybe some 25 people right now. I'm very bad at counting. We can count.
51:03 Yeah. So there should be something that that is interesting for you to come here. I just wanted to understand and if there is something that actually makes you more interested uh in coming to this kind of a thing we can adopt it. I mean we can we are happy to adopt that's what a a good democratic and a and a nice intellectual place has to be right. It has to it had to it has to thrive based on interactions between all stakeholders. So I mean if all it finally it boils down to is just that okay I need some social validation and and some interaction and some environment and that's why I'm coming that's good too. I mean we can continue what we are doing in terms of uh talking about some math and all that. Can we do something better than this is the question. Let us have
51:50 this question and keep asking it. I mean the moment we get some ideas and answers you can always come back to me. I mean beyond this classroom as well. I keep telling my my students it's a relationship that we that we build over time. See my students from IIT Delhi talked to me even today just two days back one guy I told him uh this guy was J25 very very bright guy and I taught him for three semesters uh uh I told him that he after his BTech he went to a quant firm that's what m I I mean I was just seeing a tweet yesterday that uh I Delhi computer science has 75 admits and there are 75 five uh quant companies
52:35 that come for recruitment all of them go there. Okay. And they are paid extraordinarily well. Okay. Like that those times 2017 they were paying they were paid on on on on uh of order of crores and all that. Okay. Eight digit salaries. I told Reebok that uh see there will be a time in your life when you will be bored of what you will be doing and you will come back and talk to me. That day I will tell you I told you. I told you so. Right. He sent an email. This was just last week. The subject was sir, I told you so moment. And we had a long conversation and you know it was nice. He's in New York now. I went to went there and he said sir like I'm done right. I mean
53:23 I need to do something and tell me if there are some research problem that I need to do and all that. See that's the kind of relationship that we will build. You know it's very hard to the fact that you people are still coming here and sitting here has something right. So you come here for a reason and and if you the the reason I told you all this is if you kind of get some ideas where we can improve uh human lives and pedagogy in in in specific let me know you can always write to me and here is an idea that I would want maybe so let me know if you have ideas how to make this experience better right to everybody uh and we'll experiment with it so maybe next semester we'll run a course where we'll we'll experiment with all those ideas don't worry about grades right unfortunately if I give a grades
54:10 to all of you I'll be questioned it's again a very unfortunate thing I'll have to give that bloody goian distribution to them right so so that is also a bad thing right I mean suppose there are 10 people in the class everybody comes with the intention of learning and having nice discussions it feels sad to me to actually grade them and put them on a new plus 2 sigma gajian nobody will come to me again if So I tell them that I will give you a B v C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C minus and all that right whatever nonsense is that so yeah so I feel sad for those people okay people who come to class do something put some effort and still get some bad so-called bad grade because of some bad day it's unfair I mean it's it's wrong in in in my view but yeah I'll have to get my
54:57 salary as well so that's reality so let's think about it okay and some next class we'll continue with all this uh all diffusion models etc. See one thing that I wanted to tell you is I'll slightly deviate from that given syllabus because before going to auto reggressive models I want to talk about flow-based models because it's very related to diffusion models let's do flowbased models specifically the continuous flow it's it's very good and then we will go to the LLMs because the moment we go to that LLM's uh realm right we will not be that's a downward spiral we'll have to do RLF then we'll do some uh KB cache thing and LORA and all that so let us once Once the diffusion models are completed one more class okay after that we'll go to uh flowbased models okay so we have enough
55:43 time because never mind I mean I didn't did not have uh I did not plan for today's discussion I actually planned that I will finish this if you felt that this is boring I'm sorry okay this some sometimes this happens yeah so I never did this up by the way 10 years of my teaching uh this thing career never did an entire class on this but I think time has now Come we have to do it otherwise we'll be cheating ourselves. Yeah. No I told right I mean if you have any comments let me know right we'll do >> yeah we will we'll we'll do we'll we'll we'll have a better experience for for everybody right in the world in general. That is that's a good enough conclusion.
