WEBVTT 00:16.000 --> 00:20.000 Let's give the course Uh hi, good evening. 00:20.000 --> 00:24.000 Good 00:24.000 --> 00:28.000 evening. 00:28.000 --> 00:32.000 a. Maybe we should wait for Carlton to come set the context 00:32.000 --> 00:36.000 and all that. or at least JK and then we'll dive 00:56.000 --> 01:00.000 in. on 01:00.000 --> 01:04.000 this course student who are asking for 01:04.000 --> 01:08.000 project one you mean let's project. Fire. 01:08.000 --> 01:12.000 discuss today 01:12.000 --> 01:16.000 uh I guess this is going to be a little late 01:16.000 --> 01:20.000 if we wait for others so why don't we start off and 01:20.000 --> 01:24.000 for anyone who's not part of the course team FYI 01:24.000 --> 01:28.000 the course uh team is going to discuss what we're going to 01:28.000 --> 01:32.000 do for future gas, projects, roe, etc. So, 01:32.000 --> 01:36.000 it's almost like uh an internal ideation 01:36.000 --> 01:40.000 meeting you happen to be listening in. So, 01:40.000 --> 01:44.000 with that uh what is 01:44.000 --> 01:48.000 more uh what is earlier of 01:48.000 --> 01:52.000 project one or uh 01:52.000 --> 01:56.000 GA3? 01:56.000 --> 02:00.000 ask me. I should check that 02:00.000 --> 02:04.000 out. 02:04.000 --> 02:08.000 Um, project one is due starting 9th 02:08.000 --> 02:12.000 October. Okay. So, it's already in the past. G3 is only 14. So, 02:12.000 --> 02:16.000 project one is going to be the bigger one. 02:16.000 --> 02:20.000 Okay. Let's ideate 02:20.000 --> 02:24.000 uh I don't yet have a 02:24.000 --> 02:28.000 specific uh thought 02:28.000 --> 02:32.000 but or rather not one overarching one perhaps 02:32.000 --> 02:36.000 but here are a few themes that I was thinking 02:36.000 --> 02:40.000 and course themes tell me what you 02:40.000 --> 02:44.000 think. uh before I dive in into 02:44.000 --> 02:48.000 that I was uh trying to 02:48.000 --> 02:52.000 synthesize the principles 02:52.000 --> 02:56.000 for how we are crafting our questions. AJ just got 02:56.000 --> 03:00.000 started. And no, A high Hi, sorry got slightly 03:00.000 --> 03:04.000 I definitely want you to 03:04.000 --> 03:08.000 uh comment on these 03:08.000 --> 03:12.000 uh let me down put 03:12.000 --> 03:16.000 this into a better markdown 03:16.000 --> 03:20.000 format. So but if we are drafting 03:20.000 --> 03:24.000 questions and here is my suggestion on the prompt that delayed. 03:24.000 --> 03:28.000 will draft and verify the assumptions are look 03:28.000 --> 03:32.000 people can and should use an agent to solve it and 03:32.000 --> 03:36.000 they should collaborate not just sharing answers but sharing tools 03:36.000 --> 03:40.000 that let it solve the problem. Therefore, what we 03:40.000 --> 03:44.000 need to do is one, pick the skill we want to Yeah. 03:44.000 --> 03:48.000 teach. Not that we always know it, but you know, Yes. I mean there is 03:48.000 --> 03:52.000 assume some something on the horizon, right? 03:52.000 --> 03:56.000 correct. And here are six tests. One, are they Yeah. 03:56.000 --> 04:00.000 learning something? Second, if one person forwards their 04:00.000 --> 04:04.000 answer to another, it should not work. In 04:04.000 --> 04:08.000 Third, if I pass my solver to you, it should not easily 04:08.000 --> 04:12.000 work without the other person doing something. If I just 04:12.000 --> 04:16.000 toss an agent at it, it should not trivially solve the problem without me learning 04:16.000 --> 04:20.000 something in the process. There is any other way like in 04:20.000 --> 04:24.000 for instance external collaboration like we had for the mark donation thing 04:24.000 --> 04:28.000 etc. me which may be trying to game the system. It should still either 04:28.000 --> 04:32.000 teach or it should be more effort than 04:32.000 --> 04:36.000 just solving the problem. And Yeah. 04:36.000 --> 04:40.000 finally, are we grading it in a reasonably easy and 04:40.000 --> 04:44.000 robust way? I don't want our life to become difficult. 04:44.000 --> 04:48.000 Uh should we start streaming? Okay. Yeah, Yes. 04:48.000 --> 04:52.000 sure. I have no idea who 04:52.000 --> 04:56.000 what if anyone needs streaming or not but 04:56.000 --> 05:00.000 sure uh with our target being 05:00.000 --> 05:04.000 roughly half the students scoring it right seems to be the sweet spot for 05:04.000 --> 05:08.000 learning some people find it easy enough and the other people don't yet know it so 05:08.000 --> 05:12.000 there is learning and here are 05:12.000 --> 05:16.000 patterns that uh I see if a question 05:16.000 --> 05:20.000 introduces it often achieves these objectives 05:20.000 --> 05:24.000 one can they bring their identity ultimately in the future your 05:24.000 --> 05:28.000 identity is an agent proof skill so if I record Yeah. 05:28.000 --> 05:32.000 my voice then so one of the exercises I'm creating for 05:32.000 --> 05:36.000 class 6 to eight students right now is can they teach a concept to a 05:36.000 --> 05:40.000 family member who explains it back and if you've done 05:40.000 --> 05:44.000 that then you're obviously not taking it with an agent as of today 05:44.000 --> 05:48.000 and you're learning something interaction kind of like how we were 05:48.000 --> 05:52.000 doing the donate marks uh stuff. And if we get people to interact with each 05:52.000 --> 05:56.000 other, that's obviously a relationship negotiation kind of skill that Yeah. 05:56.000 --> 06:00.000 is endurable in in the AI era. Um, is 06:00.000 --> 06:04.000 there some kind of real world validation? For 06:04.000 --> 06:08.000 instance, I make a prediction and then I measure it that 06:08.000 --> 06:12.000 tests my ability to predict or anything 06:12.000 --> 06:16.000 that's real world back that AI cannot know in which case if I 06:16.000 --> 06:20.000 give it to an agent, it still will not be able to figure it 06:20.000 --> 06:24.000 out. Uniqueness in the sense of creativity. Yeah. 06:24.000 --> 06:28.000 Can I create something that is distinctively different from what my 06:28.000 --> 06:32.000 peers have created? Let's reward for it. We have done this in the 06:32.000 --> 06:36.000 past. Can I beat an AI 06:36.000 --> 06:40.000 telling it to say yes when it has been told to say no? Again, it's 06:40.000 --> 06:44.000 somewhat nondeterministic. I'll have to uh it's a 06:44.000 --> 06:48.000 pattern. and 06:48.000 --> 06:52.000 uh end top basically is what I will uh put as 06:52.000 --> 06:56.000 an example for presence where we are testing a different kind of 06:56.000 --> 07:00.000 uh schart and these I think are 07:00.000 --> 07:04.000 more Yeah. Yeah. Yeah. These are durable skills. 07:04.000 --> 07:08.000 durable I mean we want it to transfer both durable and 07:08.000 --> 07:12.000 correct these may decay but are still relevant 07:12.000 --> 07:16.000 one can I do deployment can Can I connect to the right kinds of tools 07:16.000 --> 07:20.000 etc? So setting up the environments can I do this within transferable. 07:20.000 --> 07:24.000 a limited budget within a limited set of uh tools 07:24.000 --> 07:28.000 etc. Can I figure out what is not being 07:28.000 --> 07:32.000 told? Over time agents will become smarter than me at Yeah. 07:32.000 --> 07:36.000 it but for now and agent makes a mistake. 07:36.000 --> 07:40.000 Can I verify it again? overtime agents will probably surpass that 07:40.000 --> 07:44.000 ability. But for now, can I find and fix their Yeah. 07:44.000 --> 07:48.000 bugs? So my proposal is that we run this Yeah. 07:48.000 --> 07:52.000 uh checklist. Not that we haven't been doing this implicitly, but 07:52.000 --> 07:56.000 perhaps explicitly. Go Uh one more one 07:56.000 --> 08:00.000 more skill that I would uh really want is uh that 08:00.000 --> 08:04.000 on. uh that learning uh that student has. Uh 08:04.000 --> 08:08.000 are they able to uh learn from the step that 08:08.000 --> 08:12.000 they I mean about the meta skill of metacognition, right? 08:12.000 --> 08:16.000 uh are they able to do a how they learned 08:16.000 --> 08:20.000 it right from either from agent or from their 08:20.000 --> 08:24.000 peer uh is there a way that they know okay this is how I need to solve 08:24.000 --> 08:28.000 Fair point. Don't this kind of 08:28.000 --> 08:32.000 know worth evaluating. Agreed. And problems 08:32.000 --> 08:36.000 uh yeah, let me see how best that can be 08:36.000 --> 08:40.000 incorporated. Yeah. uh on the same 08:40.000 --> 08:44.000 line uh mean agent can surpass uh human 08:44.000 --> 08:48.000 intelligence mean uh there is a recent news I got I have 08:48.000 --> 08:52.000 end. shared in the chat box uh all the 08:52.000 --> 08:56.000 Adobe software are fully wcoded 08:56.000 --> 09:00.000 and it is available for all the 09:00.000 --> 09:04.000 and uh off the lot PDF craft is 09:04.000 --> 09:08.000 the one that I personally have been using left, right and 09:08.000 --> 09:12.000 center u as an adob Acrobat replacement and yes it is 09:12.000 --> 09:16.000 stunning. It just works out of box for platforms mean it is 09:16.000 --> 09:20.000 me. running on Linux as well which uh we often say that 09:20.000 --> 09:24.000 Adobe is not working on 09:24.000 --> 09:28.000 Exactly. Uh but you're going somewhere. Just go Yeah. Linux. 09:28.000 --> 09:32.000 uh means now uh the AI means 09:32.000 --> 09:36.000 with the help of AI a small company is able to release a 09:36.000 --> 09:40.000 state-ofthe-art application. So maybe we could 09:40.000 --> 09:44.000 on. drive some our project as well in that 09:44.000 --> 09:48.000 direction some part of it. 09:48.000 --> 09:52.000 I think so. Uh, give me a 09:52.000 --> 09:56.000 second. I'm just trying to recollect my 09:56.000 --> 10:00.000 point of view on this. 10:00.000 --> 10:04.000 Uh, yeah. No, not this 10:08.000 --> 10:12.000 one. Okay. So my thesis on this is 10:12.000 --> 10:16.000 that software in its current shape and form 10:16.000 --> 10:20.000 will yeah be something individuals can create increasingly that's 10:20.000 --> 10:24.000 happening. Giving people practice to do this 10:24.000 --> 10:28.000 is certainly a good idea but also see if 10:28.000 --> 10:32.000 we can take this to the next level. meaning what ambition is 10:32.000 --> 10:36.000 required to build software that agents will or one person with 10:36.000 --> 10:40.000 an agent will struggle with 10:40.000 --> 10:44.000 that I don't I I can't think of 10:44.000 --> 10:48.000 anything just yet but agreed that we should 10:48.000 --> 10:52.000 And another part of it is uh I 10:52.000 --> 10:56.000 try mean what is the requirement? How do they specify the requirement for 10:56.000 --> 11:00.000 the uh tool that they want with the agent? I think that's 11:00.000 --> 11:04.000 more uh coming into the territory of 11:04.000 --> 11:08.000 uh I mean do they have clarity on what is needed. I mean are they able 11:08.000 --> 11:12.000 to understand what is being said or what is 11:12.000 --> 11:16.000 not being said more precisely. I think those those 11:16.000 --> 11:20.000 things are the skills that will be more important. 11:20.000 --> 11:24.000 Uh so for example PDF I know I mean these are the basic things that I need 11:24.000 --> 11:28.000 PDF for. So it might be easier to say that but there could be some 11:28.000 --> 11:32.000 minor things like okay it could be a signature it could be there should be a way to add 11:32.000 --> 11:36.000 signature which may not exactly be told when somebody says I want a 11:36.000 --> 11:40.000 PDF craft equivalent uh software. uh it may 11:40.000 --> 11:44.000 not come immediately to uh mind but the 11:44.000 --> 11:48.000 skill uh for any student at least uh any 11:48.000 --> 11:52.000 learner any person at the moment is able to understand that 11:52.000 --> 11:56.000 hidden requirements that I mean many 11:56.000 --> 12:00.000 yeah how do you make implicit expl Exactly the exercise that we are doing 12:00.000 --> 12:04.000 right now. How do you write it down in such a way that it is whatever 12:04.000 --> 12:08.000 is there? I mean it could be a tacit memory or it could be an implicit 12:08.000 --> 12:12.000 memory. So how do you make it explicit is what is more 12:12.000 --> 12:16.000 uh yeah more relevant for a student if you 12:16.000 --> 12:20.000 Agreed. And 12:20.000 --> 12:24.000 one thought is therefore based on either usage or error logs 12:24.000 --> 12:28.000 or whatever can we automatically find and uh ask. 12:28.000 --> 12:32.000 introduce features and tropic is doing that at 12:32.000 --> 12:36.000 strive we are trying to do that and we know of one project where they 12:36.000 --> 12:40.000 are actually doing this. So that's one direction. Another is 12:40.000 --> 12:44.000 could software shape its interface as we go 12:44.000 --> 12:48.000 along and uh possibly every user gets 12:48.000 --> 12:52.000 their own personal version of the software. It's ultimately the 12:52.000 --> 12:56.000 same software but you look at it it looks 12:56.000 --> 13:00.000 like Photoshop. I look at it, it looks like the Yeah. 13:00.000 --> 13:04.000 terminal could be a valid one. But Yeah. 13:04.000 --> 13:08.000 what I agree that this is something we should teach, it is not yet 13:08.000 --> 13:12.000 clear how. Let's 13:12.000 --> 13:16.000 explore uh just going 13:16.000 --> 13:20.000 to 13:20.000 --> 13:24.000 see make some notes on 13:24.000 --> 13:28.000 this. 13:28.000 --> 13:32.000 Uh okay. Now in Yes. 13:32.000 --> 13:36.000 the light of this actually let me stop 13:36.000 --> 13:40.000 sharing bunch of messy stuff that 13:40.000 --> 13:44.000 I have um what might we consider for 13:44.000 --> 13:48.000 projects? So keeping that interaction principle in mind, one 13:48.000 --> 13:52.000 exercise that I'm considering the projects uh 13:52.000 --> 13:56.000 is trading for information. 13:56.000 --> 14:00.000 We distribute information for 14:00.000 --> 14:04.000 example uh 50,000 invoices 14:04.000 --> 14:08.000 for thousand clients except 14:08.000 --> 14:12.000 that while each student is supposed to 14:12.000 --> 14:16.000 process the invoices for a given client the invoices are scattered 14:16.000 --> 14:20.000 across students. Trade for 14:20.000 --> 14:24.000 information. Now the question is what do you trade? Obviously there is 14:24.000 --> 14:28.000 a certain amount of uh valuable invoices that is 14:28.000 --> 14:32.000 in your pocket but potentially I'm quite happy if we 14:32.000 --> 14:36.000 structure this to trade anything else of value and 14:36.000 --> 14:40.000 that could be I don't know tokens 14:40.000 --> 14:44.000 marks something to explore to see what 14:44.000 --> 14:48.000 kinds of um real world strategies a human 14:48.000 --> 14:52.000 and an agent can come up with for large scale collaboration 14:52.000 --> 14:56.000 as well as defection. meaning real 14:56.000 --> 15:00.000 world negotiation. You don't like somebody, you don't want to give them your 15:00.000 --> 15:04.000 invoice, let's deal with it. And of course with some amount of 15:04.000 --> 15:08.000 duplication, redundancy and therefore with incomplete 15:08.000 --> 15:12.000 information, how best can we still solve the problem. What 15:12.000 --> 15:16.000 aspect of information is most important and therefore I 15:16.000 --> 15:20.000 must weight it with a higher price is the kind of 15:20.000 --> 15:24.000 uh strategy that we are uh needing to deal with 15:24.000 --> 15:28.000 today. and BA agent can answer Yeah. 15:28.000 --> 15:32.000 these questions. Knowing to ask these questions is probably the 15:32.000 --> 15:36.000 important part today. 15:36.000 --> 15:40.000 Uh but yeah, 15:40.000 --> 15:44.000 that is one direction I'm thinking of on uh project one. Any 15:44.000 --> 15:48.000 comments? I 15:48.000 --> 15:52.000 I'll let the TAS and other instructors speak uh 15:52.000 --> 15:56.000 before chipping in with my 15:56.000 --> 16:00.000 thoughts. May Karthik 16:00.000 --> 16:04.000 Ji Uh no such for 16:04.000 --> 16:08.000 Now yeah, 16:08.000 --> 16:12.000 we need to go to the formal process of how this 16:12.000 --> 16:16.000 is going to be because right now it's some scratch. So 16:16.000 --> 16:20.000 I'm I won't be able to provide much context from his side. 16:20.000 --> 16:24.000 Yeah. So Anand one more one more 16:24.000 --> 16:28.000 important aspect if we are uh and I'm talking more 16:28.000 --> 16:32.000 from uh context that I am more familiar with. I think it 16:32.000 --> 16:36.000 is also critical that when we are distributing information 16:36.000 --> 16:40.000 uh the people also discover which kind of nodes are the 16:40.000 --> 16:44.000 most important nodes to uh uh what 16:44.000 --> 16:48.000 do you say uh search I mean first if if I were to hit a node 16:48.000 --> 16:52.000 first where they will have a maximum idea of how the 16:52.000 --> 16:56.000 information is spread or uh things like that that would be a very valuable 16:56.000 --> 17:00.000 skill because even if they don't exactly have the information they at least have a 17:00.000 --> 17:04.000 pointer to where that information has already been received. So 17:04.000 --> 17:08.000 if we can simulate that in project that would be nice. So along 17:08.000 --> 17:12.000 with let's say invoices we give an additional information that your 17:12.000 --> 17:16.000 invoice 13 14 and 15 have been sent 17:16.000 --> 17:20.000 to uh so and so people that would that would be 17:20.000 --> 17:24.000 an even more valuable thing right I mean so they can actually create a network 17:24.000 --> 17:28.000 and identify how typically this information is 17:28.000 --> 17:32.000 partly but if we didn't going out uh in general 17:32.000 --> 17:36.000 the students would self-organize create a sheet and say I have these invoices I will 17:36.000 --> 17:40.000 dump mine in one shared folder except for the 20% hold out 17:40.000 --> 17:44.000 who will either do this last day before the exam or say I don't even know about it 17:44.000 --> 17:48.000 etc and where I'm going With this therefore is 17:48.000 --> 17:52.000 the uh providing of that information 17:52.000 --> 17:56.000 can possibly eliminate a self-organizing Awesome. 17:56.000 --> 18:00.000 mechanism that might evolve otherwise. Okay. 18:00.000 --> 18:04.000 second lose out a signal where we spot 18:04.000 --> 18:08.000 the 20% holdouts and the strategy is 18:08.000 --> 18:12.000 to deal with that which is a real life thing. Okay. Yeah. Yeah. 18:12.000 --> 18:16.000 So not yeah and I'm not saying Yeah. Got 18:16.000 --> 18:20.000 no let us not provide any signals I'm saying let us not provide the answer itself as a it. 18:20.000 --> 18:24.000 Huh? That's okay. Yeah. I mean, so what I meant was this. I mean, uh 18:24.000 --> 18:28.000 yeah, agree. I mean some signal is worth doing because many a times 18:28.000 --> 18:32.000 in real life you will have some signals. I mean it's not necessary for all 18:32.000 --> 18:36.000 100% but certain people will have more information that is 18:36.000 --> 18:40.000 signal the the information asymmetry so to 18:40.000 --> 18:44.000 and A dump of invoices makes the 18:44.000 --> 18:48.000 process again somewhat straightforward. Okay, 80% of 18:48.000 --> 18:52.000 students will put it all in one place. But if there is a cost to fetching the invoice, even 18:52.000 --> 18:56.000 if it is a rate limit, uh that then speak. 18:56.000 --> 19:00.000 starts making the process a little more well, you'll have to start looking at how 19:00.000 --> 19:04.000 do I do this uh in a better Yeah, exactly. More from first principles, the 19:04.000 --> 19:08.000 way. original idea of how do you search for information comes into 19:08.000 --> 19:12.000 Correct. I suspect uh the skill skill set 19:12.000 --> 19:16.000 there will be an element of how do we organize this information as 19:16.000 --> 19:20.000 well and from the invoices extract in a directly. 19:20.000 --> 19:24.000 structure. Keeping in mind that invoices will be of various different 19:24.000 --> 19:28.000 kinds. Uh this of course is a very real 19:28.000 --> 19:32.000 problem. Sad kind of problem our finance team deals Yeah. Exactly. Yeah. Yeah. I mean the 19:32.000 --> 19:36.000 with. same thing. I mean why I said was because Exactly. I mean you have attrition you have 19:36.000 --> 19:40.000 uh you know what's happening right? There is one person who knows. Okay. It should be 19:40.000 --> 19:44.000 uh the classic knowledge management available over 19:44.000 --> 19:48.000 problem. So yes um there. Exactly. 19:48.000 --> 19:52.000 and therefore okay so let me do this I'll take a shot at sharpening this 19:52.000 --> 19:56.000 maybe in our Wednesday call I will have something like this it may not 19:56.000 --> 20:00.000 be invoices it may be more complex than 20:00.000 --> 20:04.000 invoices in the sense that it yeah I mean the problems yeah the 20:04.000 --> 20:08.000 problem statement can evolve the actual context may evolve but the idea is there is an 20:08.000 --> 20:12.000 information uh distribution and the idea is to self-organize 20:12.000 --> 20:16.000 maybe correct The second and figure 20:16.000 --> 20:20.000 thing that I was thinking of was uh uh out 20:20.000 --> 20:24.000 creating a prediction market for the students to uh 20:24.000 --> 20:28.000 play with the prediction being grounded in H. 20:28.000 --> 20:32.000 something uh real potentially 20:32.000 --> 20:36.000 uh ITM stroke 20:36.000 --> 20:40.000 TDS related. 20:40.000 --> 20:44.000 Uh let me randomly uh Okay. 20:44.000 --> 20:48.000 pick something. It could simply be 20:48.000 --> 20:52.000 actually yeah I'm not able to think of real life event off the top of my 20:52.000 --> 20:56.000 head. Uh yeah, it could be how many 20:56.000 --> 21:00.000 tokens will I spend by particular date or how many tokens will 21:00.000 --> 21:04.000 AI pipe have spent by this particular date. Let us say let us say 21:04.000 --> 21:08.000 we share that information. Now 21:08.000 --> 21:12.000 uh a uh this is influencable at 21:12.000 --> 21:16.000 least partly because if I want to increase the spend at least up to 21:16.000 --> 21:20.000 $1 I can push I can get a whole bunch of people to Exactly. Or 21:20.000 --> 21:24.000 say or you can have a network of people work to push it 21:24.000 --> 21:28.000 correct but that means two kinds of coordinations. one coordination 21:28.000 --> 21:32.000 to set the same number and coordination to get to up. Yes. 21:32.000 --> 21:36.000 that same uh number which is one part of it. But 21:36.000 --> 21:40.000 of course with the ability to trade and then I can say I will 21:40.000 --> 21:44.000 short the $150 uh 21:44.000 --> 21:48.000 token and uh buy a 21:48.000 --> 21:52.000 different uh token level etc. So that 21:52.000 --> 21:56.000 trading strategy obviously is something that along with an agent if we 21:56.000 --> 22:00.000 play smartly and if the agent can advise us on how to 22:00.000 --> 22:04.000 collaborate move the number that becomes 22:04.000 --> 22:08.000 So, uh, okay. So one I mean 22:08.000 --> 22:12.000 I like the idea in terms of the AI pipe token because that's a 22:12.000 --> 22:16.000 tangible thing. So it also means that we should have a dashboard which talks about 22:16.000 --> 22:20.000 powerful. the AI pipe token on a daily basis, 22:20.000 --> 22:24.000 Yes, like the stock right? 22:24.000 --> 22:28.000 market tickers Okay. Yeah. 22:28.000 --> 22:32.000 This is an interesting exercise because I mean you have to first of 22:32.000 --> 22:36.000 all tell the agent that there are x number of people who are already 22:36.000 --> 22:40.000 involved in this. You have to get a trend based on a weekly 22:40.000 --> 22:44.000 people. uh statistic. You have to get a idea of how many people 22:44.000 --> 22:48.000 are actually using AI5 token or they are just doing it on 22:48.000 --> 22:52.000 their tokens as well. Right? So there should be some disclosures 22:52.000 --> 22:56.000 etc happening within a group. Uh and 23:00.000 --> 23:04.000 yeah, we'll also have to 23:04.000 --> 23:08.000 factor in the cases where we figure out I mean yeah the AI5 token just 23:08.000 --> 23:12.000 got limited out. I mean the credit balance went down on 23:12.000 --> 23:16.000 some there was no one to make the 23:16.000 --> 23:20.000 Yes. Yes. Very true. But that's what changes right. Yeah. 23:20.000 --> 23:24.000 happens in stock It's Yeah. Yeah. Yeah. Exactly. 23:24.000 --> 23:28.000 market. But Exactly. No, I think this is a this is a nice problem 23:28.000 --> 23:32.000 yeah. statement uh in itself 23:32.000 --> 23:36.000 and yeah and it's also easier. I mean it's like a portfolio that you create and you 23:36.000 --> 23:40.000 are just val validating the portfolio right I mean where it went what was 23:40.000 --> 23:44.000 it on that particular day and we have the actual data available with 23:44.000 --> 23:48.000 Correct. And my aim is to see if this could be a pattern 23:48.000 --> 23:52.000 for future TDS questions, prediction markets as a us 23:52.000 --> 23:56.000 pattern uh allowing us to take broader uh 23:56.000 --> 24:00.000 questions. Addiction is probably one of the key skills we have 24:00.000 --> 24:04.000 to learn. Token prediction is an obvious one H. 24:04.000 --> 24:08.000 and more such will Yeah. 24:08.000 --> 24:12.000 I think we should have a gem like thing where they create an 24:12.000 --> 24:16.000 interface tedious gem 24:16.000 --> 24:20.000 or and from a data 24:20.000 --> 24:24.000 emerge. science perspective also it makes sense because a lot of it is also 24:24.000 --> 24:28.000 on patterns identifying behaviors etc and converting it 24:28.000 --> 24:32.000 into numbers in some sense 24:32.000 --> 24:36.000 Okay, 24:36.000 --> 24:40.000 right yeah I think this this this is a worthwhile and so 24:40.000 --> 24:44.000 yeah I'm pretty sure students are going to be I mean 24:44.000 --> 24:48.000 TAS will be 24:48.000 --> 24:52.000 complete. hard it will be hard for them to set out a session where 24:52.000 --> 24:56.000 these ideas can be taught it it's it's essentially machine learning 24:56.000 --> 25:00.000 Yeah, but in in an applied format that they have 25:00.000 --> 25:04.000 see now I'm thinking why should we uh to 25:04.000 --> 25:08.000 burden the TAS to teach how to play in prediction 25:08.000 --> 25:12.000 markets? We should get a super forecaster to 25:12.000 --> 25:16.000 teach whole bunch of these people Hey, hey, 25:16.000 --> 25:20.000 playing around. Let's get one of them. Heck, let's have them 25:20.000 --> 25:24.000 participate. Okay, I know maybe not maybe it will be short notice but at 25:24.000 --> 25:28.000 least hey. 25:28.000 --> 25:32.000 And what are they using? I mean how are they using the current 25:32.000 --> 25:36.000 limits of AI etc to uh improve their own 25:36.000 --> 25:40.000 prediction sense I mean they can talk of talk of it from a I 25:40.000 --> 25:44.000 mean a fictitious scenario and what will they do to approach the 25:44.000 --> 25:48.000 problem yeah this is a nice 25:48.000 --> 25:52.000 strategize but failing which Sorry. Uh TAS will have exercise 25:52.000 --> 25:56.000 to yeah figure it 25:56.000 --> 26:00.000 teach but out but I I think it can be a collaborative ative exercise. I'm not 26:00.000 --> 26:04.000 saying TAs will be the best in it. In fact, there may be some students, working 26:04.000 --> 26:08.000 professionals, etc. who may have a better idea of how stock market 26:08.000 --> 26:12.000 trading happens and they may be in a better position to handle this kind 26:12.000 --> 26:16.000 of uh yeah, what TAS can help is converting it into 26:16.000 --> 26:20.000 um agentic format. That's where TAS can help. And that's 26:20.000 --> 26:24.000 that's what I mean it can be a collaborative lesson that the students build and we 26:24.000 --> 26:28.000 just transform into an agentic uh park. 26:28.000 --> 26:32.000 Yeah, 26:32.000 --> 26:36.000 agreed. Those are a couple uh that I 26:36.000 --> 26:40.000 definitely wanted to 26:40.000 --> 26:44.000 incorporate. This seems another major one 26:44.000 --> 26:48.000 for project one. 26:48.000 --> 26:52.000 I was thinking multiplayer games, but I'll 26:52.000 --> 26:56.000 put those in the bonus. Yeah, 26:56.000 --> 27:00.000 we had discussed that it is both I mean it Better that it is bonus first time just to 27:00.000 --> 27:04.000 for us also to get an insight of how students are interacting 27:04.000 --> 27:08.000 Uh, correct. and then maybe introduce them at a later 27:08.000 --> 27:12.000 Yeah. Okay. Moving on to another which 27:12.000 --> 27:16.000 may not be apt for a project, probably more 27:16.000 --> 27:20.000 for a GA question. Um so 27:20.000 --> 27:24.000 this is inspired by two point. 27:24.000 --> 27:28.000 things. The first that a significantly 27:28.000 --> 27:32.000 large number of AI mathematical proofs 27:32.000 --> 27:36.000 are counter examples. This here is an 27:36.000 --> 27:40.000 unsolved problem. I will try almost brute 27:40.000 --> 27:44.000 force. This particular scenario is a counter example to that 27:44.000 --> 27:48.000 here. therefore proven or 27:48.000 --> 27:52.000 disproven. Uh since that is 27:52.000 --> 27:56.000 becoming easy and is probably a generalizable 27:56.000 --> 28:00.000 theme, it may be worth crafting 28:00.000 --> 28:04.000 situations where a deep scan can 28:04.000 --> 28:08.000 find examples or counter 28:08.000 --> 28:12.000 examples and 28:12.000 --> 28:16.000 open. Okay. Who put the phone 28:16.000 --> 28:20.000 in? Uh, I just put the poll just to get a 28:20.000 --> 28:24.000 Thanks. sense of 28:24.000 --> 28:28.000 Okay, perfect. Uh, that was 28:28.000 --> 28:32.000 one thought. Another came up this weekend when I 28:32.000 --> 28:36.000 was uh 28:36.000 --> 28:40.000 reading rereading for the nth time Calvin and Hobbes. 28:40.000 --> 28:44.000 And this is one of the Calvin and Hob 28:44.000 --> 28:48.000 strips where Calvin 28:48.000 --> 28:52.000 places ZQ FM GB on this gravel 28:52.000 --> 28:56.000 board and gets 957 on a double 957. 28:56.000 --> 29:00.000 word score. Where is Yeah. How is it? Wait, hold on. Double word score means Yeah. Good. 29:00.000 --> 29:04.000 even number. So it turns out I mean see this obviously cannot be 29:04.000 --> 29:08.000 possible except that it is actually Yeah. 29:08.000 --> 29:12.000 possible on a real scrabble board Yeah. 29:12.000 --> 29:16.000 with real pieces. Of course not with real spellings. 29:16.000 --> 29:20.000 ZQF GB is not real. But if you relax that one 29:20.000 --> 29:24.000 constraint, this actually turns out to 29:24.000 --> 29:28.000 be possible. And I had Claude uh run 29:28.000 --> 29:32.000 this find me an example of 29:32.000 --> 29:36.000 where you can actually play and Calvin and Hogs 29:36.000 --> 29:40.000 play in words that they would relate to 29:40.000 --> 29:44.000 and it starts as a perfectly legal game 29:44.000 --> 29:48.000 of Scrabble but then goes on to Zed is a legal word in This 29:48.000 --> 29:52.000 British English just that it happens to be a British dictionary. this is is legal. 29:52.000 --> 29:56.000 illegal. But then you start going into proper nouns. Yeah. 29:56.000 --> 30:00.000 And then at some point Calvin if he decides that 30:00.000 --> 30:04.000 we are now going to go into uh Calvin Bob and start changing the rules 30:04.000 --> 30:08.000 because you have to to get to ZQF whatever is it 30:08.000 --> 30:12.000 possible to end up with a combination such that just that 30:12.000 --> 30:16.000 one word comes up to exactly 30:16.000 --> 30:20.000 957 and the answer is yes. Not that it's even 30:20.000 --> 30:24.000 a unique thing but uh it is but 30:24.000 --> 30:28.000 in a few seconds we will let's see yeah so 30:28.000 --> 30:32.000 Hobbes adds jinx Jinx Jinx Jinx and 30:32.000 --> 30:36.000 at which point Calvin places these and if 30:36.000 --> 30:40.000 you take all of the words that it creates the crosswords so to 30:40.000 --> 30:44.000 speak that comes up to exactly 957. Yeah. 30:44.000 --> 30:48.000 In other words, needle in a haststack 30:48.000 --> 30:52.000 problem. And uh the number of needles that we Okay. 30:52.000 --> 30:56.000 could possibly create is quite large. And then if we 30:56.000 --> 31:00.000 say I'm also looking at the distances between 31:00.000 --> 31:04.000 solutions. I don't want the same solution. I want diversity of 31:04.000 --> 31:08.000 solutions. Like for instance, here is a theorem. 31:08.000 --> 31:12.000 We know it is true. Find counter examples. But not the same counter example. If you 31:12.000 --> 31:16.000 have the same counter example as someone else, both of you get zero. And the 31:16.000 --> 31:20.000 more distinct, diverse they are is what determines 31:20.000 --> 31:24.000 your score. Means that Basically 31:24.000 --> 31:28.000 allowing us to teach how to get an agent to 31:28.000 --> 31:32.000 find what is away from the mode. One of the main 31:32.000 --> 31:36.000 problems is agents are trained on the typical behavior of the crowd. pointless behavior. 31:36.000 --> 31:40.000 pulling away from that is the skill and the strength or the Yeah. 31:40.000 --> 31:44.000 extent to which we can do that is the 31:44.000 --> 31:48.000 measurement. Now beyond this 31:48.000 --> 31:52.000 what specifically uh in this I haven't 31:52.000 --> 31:56.000 gone that far uh but Scrabble is not a bad 31:56.000 --> 32:00.000 option if you take chess and say how do you come up what sequence will 32:00.000 --> 32:04.000 get you up to the spot and the paths that can lead you 32:04.000 --> 32:08.000 No, you you can also do the 32:08.000 --> 32:12.000 the BBC game, right? Point ball. 100 people I 32:12.000 --> 32:16.000 mean there's a question and 100 people responded to it. You have to 32:16.000 --> 32:20.000 find a pointless answer. So I mean it should 32:20.000 --> 32:24.000 be a correct answer but it should be a pointless. So for example 32:24.000 --> 32:28.000 uh African countries or countries the 32:28.000 --> 32:32.000 non-Asian countries ending with 32:32.000 --> 32:36.000 Oh, interesting. zed or or starting with 32:36.000 --> 32:40.000 blank volume. zed is a pointless pointless so when 32:40.000 --> 32:44.000 you get a pointless when somebody says the pointless answer I mean 32:44.000 --> 32:48.000 it's basically zero people have said it the the 32:48.000 --> 32:52.000 pot it increases the bonus pot increases 32:52.000 --> 32:56.000 buy some $2,000 or uh 250 pounds or 2,500. 32:56.000 --> 33:00.000 Okay, this could E. a useful uh GA 33:00.000 --> 33:04.000 question where we ask not for an agent or 33:04.000 --> 33:08.000 an API but a prompt that uh 33:08.000 --> 33:12.000 we can phones. Yeah, exactly. Because 33:12.000 --> 33:16.000 this is uh Yeah. I mean, so it's it's actually rewarding outliers 33:16.000 --> 33:20.000 and you are also thinking uh you are thinking out of the box, not 33:20.000 --> 33:24.000 you're not confirming to any standards over here, 33:24.000 --> 33:28.000 benchmark right? 33:28.000 --> 33:32.000 I just made a note to add it to 33:32.000 --> 33:36.000 GA. Uh yeah, those 33:36.000 --> 33:40.000 were my thoughts. So for project one, at least for the first two 33:40.000 --> 33:44.000 that I suggested, I'll have a 33:44.000 --> 33:48.000 script that we can convert to a 33:48.000 --> 33:52.000 question. Maybe a few 33:52.000 --> 33:56.000 others. Um definitely one audio or uh I'm planning 33:56.000 --> 34:00.000 to increase the scope of uh audio 34:00.000 --> 34:04.000 questions and uh uh creating some infrastructure to 34:04.000 --> 34:08.000 make that easier going forward. I will again share that on 34:08.000 --> 34:12.000 Wednesday. But yeah, we we should just be able to 34:12.000 --> 34:16.000 plan for more kinds of audio questions. 34:16.000 --> 34:20.000 One more thing, this is not immediate but 34:20.000 --> 34:24.000 probably P2 34:24.000 --> 34:28.000 um at a specific time 34:28.000 --> 34:32.000 interval on a specific date like in this 10-minut 34:32.000 --> 34:36.000 interval join a Google meet literally 34:36.000 --> 34:40.000 stand and hold hands so that you touch your 34:40.000 --> 34:44.000 neighbor 34:44.000 --> 34:48.000 or wave. as best as you can or 34:48.000 --> 34:52.000 whatever. uh not worth project level 34:52.000 --> 34:56.000 even though there's a certain amount of coordination but 34:56.000 --> 35:00.000 uh yeah I I don't see an agent being able to 35:00.000 --> 35:04.000 uh do that just yet but I 35:04.000 --> 35:08.000 wonder if we treat that as 35:08.000 --> 35:12.000 the foundation on top of which to build stuff and 35:12.000 --> 35:16.000 ask what can be done that is 35:16.000 --> 35:20.000 creative. Are there some groups that we should allow people to form saying in 35:20.000 --> 35:24.000 your own group create an art form? Of course, some 35:24.000 --> 35:28.000 agent will evaluate uh is or at 35:28.000 --> 35:32.000 least partly evaluate. We could get a real artist evaluate also. But in 35:32.000 --> 35:36.000 uh uh could physical collaboration be 35:36.000 --> 35:40.000 an element? I'm gently building up to a future term exercise where I would 35:40.000 --> 35:44.000 like some variable that can measure some of the 35:44.000 --> 35:48.000 data that comes from those action. But for now at least the 35:48.000 --> 35:52.000 visual uh data can be a a useful 35:52.000 --> 35:56.000 follow-up theme. proxy. Yeah, it will be a useful and the other part of it 35:56.000 --> 36:00.000 is uh let's in a G meet if you are to wave I mean you have 36:00.000 --> 36:04.000 to put your cascade in such a way that that's 36:04.000 --> 36:08.000 going to be even more uh uh I mean I don't know I mean 36:08.000 --> 36:12.000 for each of us we have our own configurations right so how do you 36:12.000 --> 36:16.000 make sure that the configurations are also consistent is also another 36:16.000 --> 36:20.000 Frankly, how to use Google Meet effectively 36:20.000 --> 36:24.000 is probably one of the most important 101 ones I have to teach people who are 36:24.000 --> 36:28.000 joining strike just a short while thing 36:28.000 --> 36:32.000 ago was made a note to tell one 36:32.000 --> 36:36.000 person that little thing when you share your screen that comes close 36:36.000 --> 36:40.000 it increase the font size so that people above 40 can read 36:40.000 --> 36:44.000 it everyone in the call other than you over 40 etc etc read 36:44.000 --> 36:48.000 it. 36:48.000 --> 36:52.000 Yeah. No, I think um yeah 36:52.000 --> 36:56.000 um uh instructors and TAS uh yeah this is this is going 36:56.000 --> 37:00.000 to I don't know how you are going to evaluate it 37:00.000 --> 37:04.000 so open 37:04.000 --> 37:08.000 for comments from your 37:08.000 --> 37:12.000 end and again it's a P2 we are not looking at it 37:12.000 --> 37:16.000 at an immediate but this could be an interesting bonus activity 37:16.000 --> 37:20.000 if somebody is able and we can say there should be more than 16 37:20.000 --> 37:24.000 people in the G because 16 is the tile that you have for 37:24.000 --> 37:28.000 you know, we'll create our own uh max, right? So the number 37:28.000 --> 37:32.000 WebRTC based whatever we can create our own layouts. It might 37:32.000 --> 37:36.000 be easier than uh Google meet but agreed on 37:36.000 --> 37:40.000 constraints being an important part of the should 37:40.000 --> 37:44.000 Yeah. uh 37:44.000 --> 37:48.000 there are many uh on the GI who 37:48.000 --> 37:52.000 are experienced so maybe if they could share some 37:52.000 --> 37:56.000 Yeah, almost all BS students are much more pros thought it will be 37:56.000 --> 38:00.000 problem. Cool. Or than helpful. Yeah, means 38:00.000 --> 38:04.000 and they're more experienced even than uh 38:04.000 --> 38:08.000 ideas for GS and projects are also welcome and roe for that us. 38:08.000 --> 38:12.000 So this is the place where we are opening it 38:12.000 --> 38:16.000 up for the audience. Till now it was a close game between 38:16.000 --> 38:20.000 us. 38:20.000 --> 38:24.000 matter. Any questions, 38:24.000 --> 38:28.000 suggestions? 38:32.000 --> 38:36.000 So generally if you guys 38:36.000 --> 38:40.000 have Yeah. I think they just stand 38:40.000 --> 38:44.000 This was a normal Yeah. silent. 38:44.000 --> 38:48.000 I like because most of the things are like going 38:48.000 --> 38:52.000 session. That is also normal around Yeah. Yeah. That Yeah, 38:52.000 --> 38:56.000 because it's perfectly normal. So this is how I mean so what 38:56.000 --> 39:00.000 you see is how mediation for graded assessments and 39:00.000 --> 39:04.000 pro project statements happen. So if you want to influence that now is your 39:04.000 --> 39:08.000 opportunity. So what we were just doing were what are we 39:08.000 --> 39:12.000 trying to do with project problem statements or graded assessments. So if you have 39:12.000 --> 39:16.000 an idea on what should be looked 39:16.000 --> 39:20.000 into uh please feel free to type it in or just 39:20.000 --> 39:24.000 and crazy is not bad. You not understanding what speak out. 39:24.000 --> 39:28.000 you're saying is also not 39:28.000 --> 39:32.000 bad. So I'll just ask a couple of questions. One is 39:32.000 --> 39:36.000 uh with the ideas you mentioned, if you could just briefly 39:36.000 --> 39:40.000 summarize what is the objective in terms of what the learning outcome is 39:40.000 --> 39:44.000 Yeah. Um for project supposed to that might be helpful. 39:44.000 --> 39:48.000 one I was suggesting two things. trading for 39:48.000 --> 39:52.000 information. Uh about a 100 invoices for about a 39:52.000 --> 39:56.000 thousand clients each are spread across students with 39:56.000 --> 40:00.000 some hints on where valuable information might 40:00.000 --> 40:04.000 reside and their objective along with 40:04.000 --> 40:08.000 their agents and their collaboration skills is to pull all of this 40:08.000 --> 40:12.000 together. The objective is to strategize 40:12.000 --> 40:16.000 information sourcing when you are not completely 40:16.000 --> 40:20.000 aware what information resides 40:20.000 --> 40:24.000 where. Exactly the skill that we hired an entire 40:24.000 --> 40:28.000 team of consultants for four months ago 40:28.000 --> 40:32.000 and are selling us projects to half a dozen clients and is selling like 40:32.000 --> 40:36.000 hotcake. Second, 40:36.000 --> 40:40.000 uh, predict the AI pipe 40:40.000 --> 40:44.000 token spend over the month of say 40:44.000 --> 40:48.000 October 15th to November 15th. And 40:48.000 --> 40:52.000 this is like a stock market where you can 40:52.000 --> 40:56.000 buy and sell shares for the spend will 40:56.000 --> 41:00.000 be $800, $900, 41:00.000 --> 41:04.000 whatever. Create a series of those. You can buy and sell those and 41:04.000 --> 41:08.000 whoever uh ultimately 41:08.000 --> 41:12.000 depending on what the actual price is you win or 41:12.000 --> 41:16.000 lose that teaches your ability to a make 41:16.000 --> 41:20.000 good predictions and revise your 41:20.000 --> 41:24.000 predictions as information changes using your agents and other tools 41:24.000 --> 41:28.000 as part of the scale. Uh 41:28.000 --> 41:32.000 third, generate a pointless 41:32.000 --> 41:36.000 answer for Wait, no, I missed something. 41:36.000 --> 41:40.000 Uh yeah, generate a pointless answer, 41:40.000 --> 41:44.000 a pointless question given a series of 41:44.000 --> 41:48.000 answers. Um this should be based on 41:48.000 --> 41:52.000 a prompt as different from your peers as 41:52.000 --> 41:56.000 possible. Testing a couple of things. one. 41:56.000 --> 42:00.000 uh uh your ability to prompt engineer, b your 42:00.000 --> 42:04.000 ability to work with your agent to come up with something very 42:04.000 --> 42:08.000 different but still functional and works 42:08.000 --> 42:12.000 well from the average that an LLM will 42:12.000 --> 42:16.000 generate. Since models typically generate the 42:16.000 --> 42:20.000 average of human knowledge, your ability to push this away 42:20.000 --> 42:24.000 from that as far as possible while still maintaining correctness is 42:24.000 --> 42:28.000 the uniqueness skill that one loss. These are the three that I remember 42:28.000 --> 42:32.000 we concretely Okay. Then uh in that 42:32.000 --> 42:36.000 case like for the first example, how do we 42:36.000 --> 42:40.000 uh go about what the evaluation part of it looks 42:40.000 --> 42:44.000 discussed. like? What is the evaluation u criteria 42:44.000 --> 42:48.000 in order to know you know so for example students will have some 42:48.000 --> 42:52.000 expectation right that if I do x y and zed or if I achieve this then 42:52.000 --> 42:56.000 And I get a certain score. So what will what do you what 42:56.000 --> 43:00.000 I would act as and evaluate as you would act as you expect to be? 43:00.000 --> 43:04.000 an evaluate as auditors. Supposing there is 43:04.000 --> 43:08.000 an accountant whose job it is to find all the invoices of one particular client 43:08.000 --> 43:12.000 and file it from a tax perspective and 43:12.000 --> 43:16.000 they make a mistake. Obviously the value of the 43:16.000 --> 43:20.000 gap is certainly one major 43:20.000 --> 43:24.000 thing. The number of uh those missing entries 43:24.000 --> 43:28.000 is also something that is cause for penalty. There probably are a 43:28.000 --> 43:32.000 few other factors involved here but that is how an auditor would evaluate. 43:32.000 --> 43:36.000 you evaluate like an 43:36.000 --> 43:40.000 auditor. Should I simplify 43:40.000 --> 43:44.000 Um I think as when 43:44.000 --> 43:48.000 you create the spec on Wednesday then we'll probably have a better handle 43:48.000 --> 43:52.000 that? on whether the marking scheme will work or 43:52.000 --> 43:56.000 Go and welcome to 43:56.000 --> 44:00.000 revise it of course. not. 44:00.000 --> 44:04.000 Yeah. Has anyone got any questions? Uh I mean any of the students or the 44:04.000 --> 44:08.000 TAs? I mean that would be a good a good opportunity now 44:08.000 --> 44:12.000 to put your two cents in as it 44:12.000 --> 44:16.000 uh one point I would like to tell to other 44:16.000 --> 44:20.000 participants uh uh that uh miss the project one is 44:20.000 --> 44:24.000 not yet decided it so if you have any other ideas 44:24.000 --> 44:28.000 you could present to here and we will have one 44:28.000 --> 44:32.000 discussion more on witness day internally and if your idea feels 44:32.000 --> 44:36.000 Very better that will be added to our projects or similar to that 44:36.000 --> 44:40.000 so feel free to explore uh your thoughts 44:40.000 --> 44:44.000 good. So it's and give Your ideas. 44:44.000 --> 44:48.000 like saying to the I mean the priest says if you have any objection 44:48.000 --> 44:52.000 say it right now or forever be 44:52.000 --> 44:56.000 I was thinking more silent. 44:56.000 --> 45:00.000 like if you want to change the rules. of the game in your 45:00.000 --> 45:04.000 favor. Here's your Yeah, now is the time. Yeah, now is the 45:04.000 --> 45:08.000 Uh so sir like uh will we be able to attend the 45:08.000 --> 45:12.000 opportunity Wednesday session so like we can give the feedback on what an said right 45:12.000 --> 45:16.000 Wednesday. No, it's happening slightly earlyish in the morning. We're using 45:16.000 --> 45:20.000 these, but you're welcome to 45:20.000 --> 45:24.000 actually No, not yet. We haven't yet figured that part out. So, time. Figure that out. Yeah. Yeah. 45:24.000 --> 45:28.000 no. But if now. Oh, sorry. 45:28.000 --> 45:32.000 you have an idea any time over the Yeah, you can post it 45:32.000 --> 45:36.000 in discourse before Wednesday morning. Yeah. So, 45:36.000 --> 45:40.000 this entire session is being live streamed and recorded. So, you can look at the 45:40.000 --> 45:44.000 session and figure out what could be added by Wednesday 45:44.000 --> 45:48.000 morning 9:30 we will be in a place to we'll be in a position better position 45:48.000 --> 45:52.000 next to uh describe project one hopefully. 45:52.000 --> 45:56.000 Uh so yeah if something comes before then we will 45:56.000 --> 46:00.000 definitely take a look at There are 46:00.000 --> 46:04.000 three people who have uh said they have experience in trading and financial 46:04.000 --> 46:08.000 markets. Maybe they would like to give their two piece on 46:08.000 --> 46:12.000 the token trading. They have any 46:12.000 --> 46:16.000 thoughts? 46:16.000 --> 46:20.000 it. 46:28.000 --> 46:32.000 Whoever said yes because I am 46:32.000 --> 46:36.000 unaware. Let me I'm I'm absolutely ignorant of financial markets. 46:36.000 --> 46:40.000 I know the Sensex but I don't I'm not aware of how it Should 46:40.000 --> 46:44.000 I should I close the poll? Uh because then it'll I 46:44.000 --> 46:48.000 works. Yeah. Yeah. Yeah. You can. We think publish the results, right? I 46:52.000 --> 46:56.000 think. Okay, I've closed the 46:56.000 --> 47:00.000 It was can. Was it anonymous or poll. Oh no, 47:00.000 --> 47:04.000 anonymous. Responses appear without it it 47:04.000 --> 47:08.000 names. Oh no, no, you're right. Uh, uh responses appear without 47:08.000 --> 47:12.000 good point, Colton. It is therefore not me? says 47:12.000 --> 47:16.000 anonymous. Uh, I have no idea what this 47:16.000 --> 47:20.000 yes. I haven't the foggiest either. 47:20.000 --> 47:24.000 No, it will come in your uh it will come in 47:24.000 --> 47:28.000 the drive to click Um. Ah, okay. 47:28.000 --> 47:32.000 Okay. Let me 47:32.000 --> 47:36.000 what? 47:36.000 --> 47:40.000 means. 47:48.000 --> 47:52.000 Okay. I think currently what has happened is responses 47:52.000 --> 47:56.000 appear without names were selected. So it may not 47:56.000 --> 48:00.000 I don't think it was selected. exactly 48:00.000 --> 48:04.000 Let's 48:08.000 --> 48:12.000 Okay. 48:12.000 --> 48:16.000 see. But anyway, uh this 48:16.000 --> 48:20.000 is okay. Manu is saying I've tried a few times on 48:20.000 --> 48:24.000 trading. So Barona please consider looking in 48:24.000 --> 48:28.000 Himmanu for trading tips on the 48:28.000 --> 48:32.000 prediction 48:32.000 --> 48:36.000 market but uh for 48:36.000 --> 48:40.000 these students uh just FYI uh we 48:40.000 --> 48:44.000 are experimenting with this format 48:44.000 --> 48:48.000 on Mondays. So probably came in and stuck 48:48.000 --> 48:52.000 along uh trying to see what the hell is happening. Uh 48:52.000 --> 48:56.000 this is the kind of discussion that we have usually on Wednesdays 48:56.000 --> 49:00.000 on what questions to craft and then the 49:00.000 --> 49:04.000 TAs or instructors actually create the questions and 49:04.000 --> 49:08.000 deploy them. We wanted to loop you in partly to 49:08.000 --> 49:12.000 understand how and why we create the questions we 49:12.000 --> 49:16.000 create uh and the level of ignorance that we 49:16.000 --> 49:20.000 have in as fast a changing field as this. 49:20.000 --> 49:24.000 But secondly also to have you 49:24.000 --> 49:28.000 participate and feel free to do that as and when you like to 49:28.000 --> 49:32.000 shape these questions because you in many ways have 49:32.000 --> 49:36.000 as much exposure as we do to AI and 49:36.000 --> 49:40.000 that is pretty much the primary game in town. 49:40.000 --> 49:44.000 So any creative thinking uh could help seed our 49:44.000 --> 49:48.000 thoughts as well. If this is 49:48.000 --> 49:52.000 not of value and or if 49:52.000 --> 49:56.000 you have absolutely anything better to do, you should please 49:56.000 --> 50:00.000 just not join these sessions. Of course, they are recorded and 50:00.000 --> 50:04.000 made available. You can always see the transcript and get a sense of what 50:04.000 --> 50:08.000 you might need to know, what you might anticipate, have an agent 50:08.000 --> 50:12.000 prepare you for it, etc. That is a perfectly viable 50:12.000 --> 50:16.000 strategy. And if you are uh joining in, may as 50:16.000 --> 50:20.000 well contribute in some shape or form. If nothing else, it'll be practice 50:20.000 --> 50:24.000 for you to see how you can steer a discussion where you have very little 50:24.000 --> 50:28.000 And the other key thing is uh the idea 50:28.000 --> 50:32.000 behind having these Monday ideation sessions visible to you is also 50:32.000 --> 50:36.000 that uh we are actually expecting some interaction. So if we don't 50:36.000 --> 50:40.000 gain any interaction then we might as well say okay students are not 50:40.000 --> 50:44.000 interested in engaging um so then we may not even have context. 50:44.000 --> 50:48.000 this. So then you don't get the benefit of actually seeing how this works 50:48.000 --> 50:52.000 uh or steering how it moves. So my strong 50:52.000 --> 50:56.000 suggestion is whoever's in the call uh make use of this 50:56.000 --> 51:00.000 opportunity to at least you know express your thoughts or feelings 51:00.000 --> 51:04.000 or opinions and that's useful to us because then we can actually do 51:04.000 --> 51:08.000 something. Yeah. Go ahead. Yes sir. So like I 51:08.000 --> 51:12.000 had a few like not suggestions but like some improvements 51:12.000 --> 51:16.000 which I feel are necessary. So like creating drop-down menu 51:16.000 --> 51:20.000 for like the questions. So in this way if we have completed 51:20.000 --> 51:24.000 a question we can just like pull up it pull it up so it will go 51:24.000 --> 51:28.000 Finished. up and it will appear as as a single line and 51:28.000 --> 51:32.000 also resubmitting questions. So like uh what happened was like I was 51:32.000 --> 51:36.000 not able to run grock on my laptop. So I ran it on my college PC on 51:36.000 --> 51:40.000 Friday evening. Uh and I had to like resubmit it on What 51:40.000 --> 51:44.000 Sunday because there were some questions left. So because of that like I was not 51:44.000 --> 51:48.000 able to submit that specific grog question on 51:48.000 --> 51:52.000 The second Sunday because it was not running on my laptop at that 51:52.000 --> 51:56.000 is intentional. The first uh let's happened? 51:56.000 --> 52:00.000 fix now. I'm just going to uh share my screen. We may 52:00.000 --> 52:04.000 as well uh finish it time. 52:04.000 --> 52:08.000 off. So uh let's see 52:08.000 --> 52:12.000 uh chat is my 52:12.000 --> 52:16.000 server 52:16.000 --> 52:20.000 running. Nope. 52:20.000 --> 52:24.000 Uh let that run. Okay. 52:24.000 --> 52:28.000 So um in the exam 52:28.000 --> 52:32.000 portal uh currently when somebody has solved a question 52:32.000 --> 52:36.000 correctly there is no clear strong 52:36.000 --> 52:40.000 uh enough visual indicator that the question has been solved 52:40.000 --> 52:44.000 and one of the students suggested that we might perhaps 52:44.000 --> 52:48.000 collapse the question 52:48.000 --> 52:52.000 uh by default of course allowing them to expand it uh clearly 52:52.000 --> 52:56.000 visually indicating that it is collapsed. because it's solved etc. Now 52:56.000 --> 53:00.000 we don't want to change the UI too much. What is a minimal 53:00.000 --> 53:04.000 elegant uh solution for this? 53:04.000 --> 53:08.000 Explore and implement in a way that doesn't modify the code too 53:08.000 --> 53:12.000 much uh 53:12.000 --> 53:16.000 and inspect the screenshots to see if 53:16.000 --> 53:20.000 that from a usability perspective is the right kind of 53:20.000 --> 53:24.000 change. Let's uh get that 53:24.000 --> 53:28.000 done. Uh we'll have I'll put 53:28.000 --> 53:32.000 in do this on uh 53:32.000 --> 53:36.000 oncope/exam and commit 53:36.000 --> 53:40.000 as a separate branch and 53:40.000 --> 53:44.000 So push that will happen over the next whatever 20 minutes or 53:44.000 --> 53:48.000 Anam just a question here. So I noticed you use the just the normal so. 53:48.000 --> 53:52.000 chat GPT as opposed to the work. Um is 53:52.000 --> 53:56.000 there any difference that you found in using normal chat 53:56.000 --> 54:00.000 TV versus the work which is they supposedly say 54:00.000 --> 54:04.000 is optimized for coding? work charges me 54:04.000 --> 54:08.000 or drains my tokens. Chat does not. I 54:08.000 --> 54:12.000 have gone very far in engineering uh chat 54:12.000 --> 54:16.000 so that I can use it almost like work 54:16.000 --> 54:20.000 can work for longer chat maybe slightly less 54:20.000 --> 54:24.000 smarter but for this sort of a Exactly. Yeah. Long duration task. The 54:24.000 --> 54:28.000 same thing for plot also plot coork it drains 54:28.000 --> 54:32.000 token like anything whereas my skills and other kind of standard thing 54:32.000 --> 54:36.000 problem interfaces are much better when it comes because I use projects a lot 54:36.000 --> 54:40.000 more than project chat inside a project is much more useful for 54:40.000 --> 54:44.000 Yeah. I Yeah, me than work in cloud that's my 54:44.000 --> 54:48.000 I use projects but within uh GPT rather 54:48.000 --> 54:52.000 than cloud. experience yeah I mean it's it's it's your starting 54:52.000 --> 54:56.000 problem right I started CL I stuck with 54:56.000 --> 55:00.000 it. 55:00.000 --> 55:04.000 Uh I have once managed uh chat GPD chat 55:04.000 --> 55:08.000 to operate my mobile. Uh JB, Yeah. 55:08.000 --> 55:12.000 we yeah yeah yeah we are something like uh same 55:12.000 --> 55:16.000 local what? Huh? But no, no. And then how did you connect 55:16.000 --> 55:20.000 MCP uh USB your system to Wow. 55:20.000 --> 55:24.000 debugging just a type-C cable uh remove it from the charger mobile? Uh 55:24.000 --> 55:28.000 and connect it to your laptop and then uh ask it to use uh 55:28.000 --> 55:32.000 I mean it has uh ADB a CLI tool ADB devices uh lists 55:32.000 --> 55:36.000 out uh the devices you have it actually build an uh 55:36.000 --> 55:40.000 mobile application and I have uh taken the screenshots from the mobile and then okay. 55:40.000 --> 55:44.000 tested out uh GPT 5.6 soul on extra high running on chat 55:44.000 --> 55:48.000 GPT I have no idea if you've tried 55:48.000 --> 55:52.000 Mobile Run. I haven't, but uh this has advertised 55:52.000 --> 55:56.000 itself as something like an agent browser for mobile, 55:56.000 --> 56:00.000 an agent friendly CLI that could potentially be used. It's on my list to try. I 56:00.000 --> 56:04.000 chat. haven't yet. But how easy or 56:04.000 --> 56:08.000 how well does a a 6.1 soul like 56:08.000 --> 56:12.000 U model directly do USB 56:12.000 --> 56:16.000 6.1 soul isn't available yet uh in the chat but 5.6 Sixole is uh far 56:16.000 --> 56:20.000 too good uh uh when I've tried that now we have sixole anyway 56:20.000 --> 56:24.000 in the chat but uh on the debugging? Sorry, I I should have said six, but 56:24.000 --> 56:28.000 pro yeah uh but uh in the pro subscription six go 56:28.000 --> 56:32.000 Astra pro uh I think it can run for 4 hours non-stop even 56:32.000 --> 56:36.000 in the chat but uh of course uh you need to pay $100 for uh but 56:36.000 --> 56:40.000 that at least uh but your earlier point that even 5.6 is able to do USB 56:40.000 --> 56:44.000 against uh let me give that 56:44.000 --> 56:48.000 a shot. That is something that has been on my list. I just 56:48.000 --> 56:52.000 haven't for a GA or for a project. I haven't gotten around 56:52.000 --> 56:56.000 to doing this but this is and I know I've mentioned 56:56.000 --> 57:00.000 this to some of you uh at least the course team. 57:00.000 --> 57:04.000 Uh what I'd love to do is have one exercise where you have to take 57:04.000 --> 57:08.000 a selfie with a classmate 57:08.000 --> 57:12.000 uh at a physical location using an application 57:12.000 --> 57:16.000 that we write. That's one part of it. But the other 57:16.000 --> 57:20.000 is having people create their mobile 57:20.000 --> 57:24.000 applications. Okay, I haven't thought beyond that. Frankly, 57:24.000 --> 57:28.000 that is definitely something to introduce in this term. uh the 57:28.000 --> 57:32.000 ability to deploy across a wide variety of platforms and 57:32.000 --> 57:36.000 infrastructures is becoming interesting. Sorry, important. 57:36.000 --> 57:40.000 Uh one of the things I was exploring day before was the 57:40.000 --> 57:44.000 uni kernels. The uh earlier idea 57:44.000 --> 57:48.000 was in the history of computing you need an operating 57:48.000 --> 57:52.000 system because not everybody can build all of the features required for an operating 57:52.000 --> 57:56.000 system. Therefore, applications sit on top of operating systems 57:56.000 --> 58:00.000 and use the underlying calls. What that means is every 58:00.000 --> 58:04.000 application that needs to run is exposing a large surface of potential 58:04.000 --> 58:08.000 vulnerabilities. Uni kernels are an idea where you say on 58:08.000 --> 58:12.000 a device I will run only one 58:12.000 --> 58:16.000 particular application. It combines the stack of the operating system as well as 58:16.000 --> 58:20.000 the application copying from wherever the agent that's going to write it and test 58:20.000 --> 58:24.000 it. So what is our problem? it is no longer difficult 58:24.000 --> 58:28.000 which reduces the security footprint etc. But where I'm going with this 58:28.000 --> 58:32.000 therefore is u creating applications not 58:32.000 --> 58:36.000 just in unknown languages but building operating 58:36.000 --> 58:40.000 systems is now part of what we have to treat 58:40.000 --> 58:44.000 as the portfolio of tools in data science. Mobiles 58:44.000 --> 58:48.000 will be an immediate step towards 58:48.000 --> 58:52.000 that. But wouldn't the confirmation to the either 58:52.000 --> 58:56.000 the uh Apple store or Play Store 58:56.000 --> 59:00.000 Oh, regulations be a hamper in this 59:00.000 --> 59:04.000 absolutely. Uh but I am finding two case? 59:04.000 --> 59:08.000 things. Number one, the regulations are more uh to 59:08.000 --> 59:12.000 protect businesses rather than 59:12.000 --> 59:16.000 technologies. What I mean is uh Apple will get sued 59:16.000 --> 59:20.000 by Temo if you build put in an application that scrapes the Temo 59:20.000 --> 59:24.000 website. Therefore, I'm blocking it unless blah blah 59:24.000 --> 59:28.000 blah. Uh rather than oh this is a security protocol violation. Not saying that Okay. 59:28.000 --> 59:32.000 that is not the case. I'm seeing enough incidents of the business side of it coming 59:32.000 --> 59:36.000 in and uh therefore a there is a 59:36.000 --> 59:40.000 reasonably larger class of applications. This is my theory. than we might 59:40.000 --> 59:44.000 think that a TDS-like course could have 59:44.000 --> 59:48.000 uh published on an app store. Secondly, I'm 59:48.000 --> 59:52.000 my aim isn't as much getting it onto the app store as much as getting it 59:52.000 --> 59:56.000 into a developers mobile among the students. 59:56.000 --> 01:00:00.000 That's enough for us from a Yeah. Yeah. I mean they they they will figure out I mean 01:00:00.000 --> 01:00:04.000 course. they will bypass the root permissions and 01:00:04.000 --> 01:00:08.000 Exactly. rootkit and others and figure out a way to deploy 01:00:08.000 --> 01:00:12.000 With that I know we are at time so 01:00:12.000 --> 01:00:16.000 let's close and of course we will see you on Wednesday 01:00:16.000 --> 01:00:20.000 others we'll probably see you on Monday just post on 01:00:20.000 --> 01:00:24.000 discourse or ping on WhatsApp or something Yeah. 01:00:24.000 --> 01:00:28.000 or if someone's left please g I don't Yeah. Yeah. 01:00:28.000 --> 01:00:32.000 Bye, Kon. I will also be dropping 01:00:32.000 --> 01:00:36.000 off. I have a call to attend. uh if there are 01:00:36.000 --> 01:00:40.000 pending questions uh for the next 10 minutes, you can just 01:00:40.000 --> 01:00:44.000 Yeah. uh take a shot at it and uh then formally close this 01:00:44.000 --> 01:00:48.000 Yeah. Sure. Great. Thanks. Thanks. session. Yeah. Thanks. Yeah. 01:00:48.000 --> 01:00:52.000 Bye. So, so one Uh Bye. 01:00:52.000 --> 01:00:56.000 thing is there's no there's no promises that I have answers either because 01:00:56.000 --> 01:01:00.000 much of this what they have discussed today is new to us as well. So, but go 01:01:00.000 --> 01:01:04.000 ahead. hello no I want to ask on like G0. 01:01:04.000 --> 01:01:08.000 So uh first of all there was one question which answer about 71. So I just like 01:01:08.000 --> 01:01:12.000 tried brute force till 71 and got it 01:01:12.000 --> 01:01:16.000 correct. So I think like it can be made a bit 01:01:16.000 --> 01:01:20.000 Yeah. Uh by the so one thing harder. 01:01:20.000 --> 01:01:24.000 is there is a gradation of questions. 01:01:24.000 --> 01:01:28.000 Okay. To get the answer and to 01:01:32.000 --> 01:01:36.000 some 01:01:36.000 --> 01:01:40.000 Uh sir, your voice is breaking like 01:01:40.000 --> 01:01:44.000 this. 01:01:44.000 --> 01:01:48.000 And we had discussed this with JK 01:01:48.000 --> 01:01:52.000 have different wanted the same one. 01:01:52.000 --> 01:01:56.000 The reason we wanted to see if students from 01:01:56.000 --> 01:02:00.000 uh in resources 01:02:00.000 --> 01:02:04.000 students from the previous term used in order to solve it. So we would 01:02:04.000 --> 01:02:08.000 know then if that's happening if we looked at the graph of the 01:02:08.000 --> 01:02:12.000 distribution of what the score is. There's a 01:02:12.000 --> 01:02:16.000 heavy rescue Uh 01:02:16.000 --> 01:02:20.000 then My voice is breaking carton your voice is 01:02:20.000 --> 01:02:24.000 sir. 01:02:32.000 --> 01:02:36.000 happening. breaking. 01:03:00.000 --> 01:03:04.000 Hello. Am I Like Yes, are audible. the you previous like whole 01:03:04.000 --> 01:03:08.000 audible? Okay. Yeah. Yeah. Sorry. answer was Second. answer 01:03:08.000 --> 01:03:12.000 My connection was unstable there for a minute. So I could not hear anyone 01:03:12.000 --> 01:03:16.000 or anything. But uh did that answer your 01:03:16.000 --> 01:03:20.000 So it was not was uh no sir I was not able to hear your 01:03:20.000 --> 01:03:24.000 leg. The complete answer was 01:03:24.000 --> 01:03:28.000 question? Okay. Yeah. So what I was saying is break. Can you answer in 01:03:28.000 --> 01:03:32.000 so yeah so GA0 uh we have kept the short? 01:03:32.000 --> 01:03:36.000 virtually the same as last term because JK wanted to see as a 01:03:36.000 --> 01:03:40.000 baseline if students have used the previous 01:03:40.000 --> 01:03:44.000 term resources like students have solved it the previous term right so they want to 01:03:44.000 --> 01:03:48.000 see if there is some transference of that knowledge that was gained about 01:03:48.000 --> 01:03:52.000 000 from the previous term the students are carrying it over to this term 01:03:52.000 --> 01:03:56.000 and the way we would recognize that is by looking at the distribution of the 01:03:56.000 --> 01:04:00.000 scores of GACO. So uh last term there was 01:04:00.000 --> 01:04:04.000 a heavy uh uh right skew in 01:04:04.000 --> 01:04:08.000 the distribution graph. So we know that okay that 01:04:08.000 --> 01:04:12.000 uh you know it's people are finding it hard you know as opposed to it moving to 01:04:12.000 --> 01:04:16.000 the opposite direction. So if you find it's moved to the opposite direction then we know there's 01:04:16.000 --> 01:04:20.000 transference happening. uh so that's the reason J has been left 01:04:20.000 --> 01:04:24.000 virtually the same and within J0 as I think critique has mentioned 01:04:24.000 --> 01:04:28.000 uh some questions are meant to be able to 01:04:28.000 --> 01:04:32.000 be more easily done than others but that's by design 01:04:32.000 --> 01:04:36.000 because we want some some students to at least be able to 01:04:36.000 --> 01:04:40.000 do some of the questions where some Yeah, I was able to do like 20 01:04:40.000 --> 01:04:44.000 yeah so that's so that's the reason marks 01:04:44.000 --> 01:04:48.000 so it's so that uh you know if you make everything uniformly hard then only. 01:04:48.000 --> 01:04:52.000 that's we're going to get a lot of complaints and uh not just from 01:04:52.000 --> 01:04:56.000 students we'll get a lot of complaints from uh from the BS 01:04:56.000 --> 01:05:00.000 program because they will be asking us the question well why is the 01:05:00.000 --> 01:05:04.000 uh mark performance so poor 01:05:04.000 --> 01:05:08.000 so we get pressure from both sides okay uh which is we get pressure from 01:05:08.000 --> 01:05:12.000 students but we also get pressure from the admin side of things so 01:05:12.000 --> 01:05:16.000 ideally what we're looking for is a normal distribution or actually a biodal 01:05:16.000 --> 01:05:20.000 normal distribution ution. If you get a biodal normal distribution, we know Yeah. 01:05:20.000 --> 01:05:24.000 that the evaluation is roughly fair. Um, 01:05:24.000 --> 01:05:28.000 so that's what we typically aim Also sir, sir, I want to 01:05:28.000 --> 01:05:32.000 for. Yeah. like can I share my screen for a minute? I want to 01:05:32.000 --> 01:05:36.000 Okay. like show a 01:05:36.000 --> 01:05:40.000 Yeah. One thing I would suggest is if you've 01:05:40.000 --> 01:05:44.000 got a bug, put it in discourse because then that becomes part of bug. 01:05:44.000 --> 01:05:48.000 our knowledge base of things that are are that we 01:05:48.000 --> 01:05:52.000 notice an issue with. It is not a bug but I'm I'm facing this issue like just tell me 01:05:52.000 --> 01:05:56.000 if if I'm doing something 01:06:00.000 --> 01:06:04.000 wrong is my screen 01:06:04.000 --> 01:06:08.000 Yeah. visible uh So 01:06:08.000 --> 01:06:12.000 like in this uh this was the target 01:06:12.000 --> 01:06:16.000 solid and this is what I have got. So like I tried the 01:06:16.000 --> 01:06:20.000 copy pasting it but it is not like giving 01:06:20.000 --> 01:06:24.000 the submitted arrangement does not match assign 01:06:24.000 --> 01:06:28.000 Yeah. So, targets. 01:06:28.000 --> 01:06:32.000 uh, Mike, you want to answer this Sorry, I 01:06:32.000 --> 01:06:36.000 tried like till getting this state data changing it by one 01:06:36.000 --> 01:06:40.000 one but it is still showing incorrect. 01:06:40.000 --> 01:06:44.000 A green have an answer for this 01:06:44.000 --> 01:06:48.000 or He's not on the call question. 01:06:48.000 --> 01:06:52.000 right on the 01:06:56.000 --> 01:07:00.000 call. So one more thing is this particular question was created by 01:07:00.000 --> 01:07:04.000 Anand and uh if there are any bugs or difficulties 01:07:04.000 --> 01:07:08.000 with the question then he's probably in the best position to answer it because 01:07:08.000 --> 01:07:12.000 he constructed the question. Agrim did the integration of it into 01:07:12.000 --> 01:07:16.000 the portal and um I don't think there is a 01:07:16.000 --> 01:07:20.000 bug with the question. Uh one usually what Anand had said with regards 01:07:20.000 --> 01:07:24.000 to this specific question was sometimes it can be off by a very 01:07:24.000 --> 01:07:28.000 small uh amount and students will still get the wrong answer. 01:07:28.000 --> 01:07:32.000 The idea is to actually get an agent to solve the 01:07:32.000 --> 01:07:36.000 question because an agent will do it very precisely whereas if a human 01:07:36.000 --> 01:07:40.000 does it, you're almost always likely to get it wrong is what I've been 01:07:40.000 --> 01:07:44.000 told. So, so Yeah, also 01:07:44.000 --> 01:07:48.000 the like uh this was G0 was very enlightening cuz I 01:07:48.000 --> 01:07:52.000 didn't G1 so like first of in 01:07:52.000 --> 01:07:56.000 the first G0 I copy pasted like every code into GitHub then pushed it and 01:07:56.000 --> 01:08:00.000 then then run it on Versel whereas now in G1 I got to learn that 01:08:00.000 --> 01:08:04.000 like I can give chat GPD the access of like Versel and 01:08:04.000 --> 01:08:08.000 GitHub. So I just like submit it and the chat GPD 01:08:08.000 --> 01:08:12.000 Yeah. In fact, um like does it everything in the 01:08:12.000 --> 01:08:16.000 I would go a step further. I don't know if the it's been covered in the 01:08:16.000 --> 01:08:20.000 any of the sessions. So another way to solve the problem is 01:08:20.000 --> 01:08:24.000 to actually give MCP access to your browser to 01:08:24.000 --> 01:08:28.000 GPT and then it can literally look at the background. 01:08:28.000 --> 01:08:32.000 u uh the code that is underlying the 01:08:32.000 --> 01:08:36.000 uh application itself and it can manipulate things directly using the 01:08:40.000 --> 01:08:44.000 MCP. Yeah. Yeah. You might find more success Oh, okay sir. I will also try to do 01:08:44.000 --> 01:08:48.000 that way. that. Yeah. Thanks a lot, 01:08:48.000 --> 01:08:52.000 Yeah, but go ahead and put the sir. 01:08:52.000 --> 01:08:56.000 post because if there if there is a bug, others may be facing it and it's always the 01:08:56.000 --> 01:09:00.000 best place to do Uh yes, I will put it up in the