AI: Zara dekh ke chalo

Full show transcript below

Vimoh: [00:00:00] Hey, we are live. Hi everyone, and welcome to the livestream. Welcome to this channel's livestream, which is happening probably after a couple of years. Uh, I do a separate livestream on my livestream channel, uh, called Vimoh Live. But this is just my main channel and it's a special livestream where we are going to talk about artificial intelligence.

Vimoh: But before I start, I just wanted to give everyone a quick reality check. None of us are actual AI experts in the way that a coder or a technical person might be is a science YouTuber. Uh, he naturally has a curiosity in this area. Magna is a former journalist and a podcaster, and he recently did a podcast episode with experts and non-experts about artificial intelligence.

Vimoh: I am a writer and a creative person, and I have been engaged in ai in that I use AI tools to augment my writing sometimes and to package my creative work. Sometimes, like I've used AI tools and AI writing tools, and I've talked to a few [00:01:00] people about AI on my main podcast also. So that's there. But before we begin, let me just take off everyone from the screen and play you a little bit that I made to introduce you to the topic of today's, uh, episode.

Vimoh: So let me just...

Fake Vimoh: hello and welcome to this special live stream about artificial intelligence. The purpose of this show is to make you aware of the fact that even though AI is going to help us in many ways, it is probably also going to complicate our lives. In many ways, it may affect employment, governance, and policy, but it may also change the shape of religion and philosophy.

Fake Vimoh: You can probably tell that I am not an actual human being, but imagine what will happen when you are unable to tell the difference that they may already be here. [00:02:00] So consider this episode and opportunity to understand how to deal with that reality. And now over to Magna and my real self.

Vimoh: So, uh, since my non-human self already kind of introduced Magna, and I will just augment that...

Pranav: You should have played that right at the start. Yeah, that's

Meghnad: what I was saying. Also, it doesn't sound like you, but uh,

Vimoh: creepy. Yeah. It doesn't, it creepy. That was not my voice. That was one of the tools, voices. But, uh, I wanted to make it sound different, like a, like zero chance of somebody confusing it with me, even though it has my face camera.

Meghnad: Yeah. But, uh, but you could see the difference between your frame. Like, like right side me below was like different colors. It doesn't matter. It doesn't matter.

Vimoh: Yeah.

Pranav: I just add one thing that you said about while you [00:03:00] were introducing us, that I have a natural curiosity for AI because I'm a science YouTuber.

Pranav: Funny thing, I used to actually do research on ai. I was a research assistant in IAC Bangalore for, um, uh, and my research topic had to, uh, we were working on neural networks. Mm-hmm. And we, we were working on this machine learning algorithm back then called Deep learning, which is basically coding.

Pranav: We feed, um, uh, data points to the algorithm and, uh, train it on a data set, on a, on these particular, and we are trying to achieve outputs that can detect things like, uh, if it's image training, maybe it's trying to recognize pattern and images, uh, like let's say cancer diagnosis. Mm-hmm. Or, uh, if, if we are feeding it, uh, sound wave forms, uh, detecting certain voices, [00:04:00] whatever.

Pranav: So that, Surprisingly is where my, uh, trust with aims today. Fast forward to today when AI is, uh, all the rage, I'm like the least, I'm like the most illiterate person when it comes to ai. I hardly use G P D. I hardly use, uh, uh, anything. Uh, maybe I use a bit of image, uh, ai, art generation, but that's it. I'm the most ai.

Pranav: So two,

Vimoh: two quick questions. One, how many years ago was this your engagement with ai?

Pranav: Uh, this wasn't 2015 for about six months, I'd say.

Vimoh: Okay. Okay. And the second question is, which AI image generation tools are you using? Uh,

Pranav: I, I've used, um, uh, what is that Mid

Meghnad: journey? Mid.

Pranav: Use one more. I'm not sure I should [00:05:00] ask my editor, but

Vimoh: Okay.

Vimoh: Cool. Cool. So

Meghnad: I, I would actually like to counter, uh, Preau you're a liar, Preau, because you use AI all the time. You're on YouTube. So like, if you, if you guys have seen this video, second

Vimoh: people say AI these days in the year 2023. When people say ai, they mean generative ai.

Meghnad: I refuse. No, I refuse. That is the, that is the thing that I wanted to clarify, right?

Meghnad: Which is basically, uh, so there's this video that I saw called AI Dilemma where, uh mm-hmm the people who made the social dilemma on Netflix, if you saw it. Ah-huh. Um, so they basically did this whole presentation. You guys should watch it. Definitely. Because what they said is that we are assuming that this is our first contact with ai, but actually the algorithms were our first contact with ai.

Meghnad: So what Prum described as machine learning. Okay. Yeah.

Vimoh: So in the recommendation

Pranav: ai, using AI

Vimoh: in just the way that Magna disagree with consist,

Pranav: I don't use AI

Vimoh: active. Yeah, so it's a active AI user. [00:06:00] I would like to, I would like to disagree with Magna also, because our first interaction with AI was not, In the form of tools that involved some kind of machine learning.

Vimoh: It was way before that. It was in the forties and the thirties when we first started imagining that machines might one day become conscious in pulp science fiction magazines. My, oh God,

Meghnad: you, you, you back to like, oh, imagination Major. That was our first contact. No,

Vimoh: our See I, I'll tell you, I'll tell you something. One of the reasons, like, first of all, this live stream is happening because Prav got in touch and said, dude, people might ask their lives problems to ai and when pna, sorry, when AI replies they might, because AI sounds like it knows what it is doing, they might conclude that.

Vimoh: There is intelligence there and we are already at a place where we think that the universe has intelligence and God exists [00:07:00] and all that. So AI may one day become something like, it may become a religious influencer society. Hmm. The second thing is that the reason I brought up the Pulse magazine things is that a lot of fears that we have about AI taking over the world literally come from science fiction, literally come from like, if tomorrow aliens come, the fear that we will all be experiencing will be fueled by the visions that our science fiction movies have given us, envisions, patients in the sky are shooting down, et cetera.

Vimoh: Not all movies are like that, but I think the first time we looked at automatic solutions to our problems, we were already where we are now, we were, we were already at a state ca We don't know what it is going to do to us. We don't know how it is going to change society and all that. And by the way, if there was a fourth person in this podcast, she would've disagreed with me and said, yes, exactly.

Vimoh: I was just about

Meghnad: discover [00:08:00] whatever. But you know, uh, what, what I was about to, uh, sort of the algorithm thing that I was saying, right. Which, um, basically you were doing machine learning and mm-hmm. So the AI recommendation engines have been around us for a while. So like the whole thing about social media becoming like super destructive and you know, it, it's actually a sign, right?

Meghnad: So when social media first came in, it was like, oh my God, connecting the world. We are all going to be so friendly with each other and like, everybody's going to be like having so much fun and whatever. But of course it destroyed democracy. Uh, and that was primarily because these algorithms are sort of feeding us everything that we like.

Meghnad: Mm-hmm. And to keep us on the platform. And that that algorithm is being taught by us, by the actions we take on the app itself, which is machine learning. And therefore, um, that AI has already been operating in the background [00:09:00] as, as pronounced unconsciously maybe. Um, but now with this, the generative part, so that was the curative ai.

Meghnad: Now this is the generative AI age that we are in. Right. So it's the same thing. Right. You know, now everybody's like, oh my God, it's gonna change everything. You know, everything is gonna be so great. Now everything has to be a generalist. I can be like a, a whatever doctor and an engineer at the same time.

Meghnad: I just have give prompts. I can code . Exactly. And all the, there is the other end of the spectrum, which you describe the science fiction would like, oh fuck, we are dead. Yeah. Um, like, we are all gonna die. So I feel like the

Vimoh: truth is somewhere in the middle.

Meghnad: Exactly. I, that's, that's what I was saying, that, you know, like, I think the, the hype cycle is there, right?

Meghnad: You know, like bitcoin kiss hype cycle or the social media, media, same. You know, like everybody's so hype. All the garbage Twitter threads I'm reading on this is just, just like next level they say, oh, if you're [00:10:00] not using these 10 AI tools, you are, you are left, left behind sort of thing. Right. Which is like creating this impression that if you are not using AI A you will be replaced.

Meghnad: B oh you are being left behind and you will not be able to catch up ever. Um,

Vimoh: but yeah, that's not entirely untrue though, is it?

Pranav: Uh, I, I think there's a very high likely chance that I'm gonna be like, my dad is with a smartphone. I'm gonna be like that with ai, with chat. You're not that old.

Vimoh: You're not that old.

Vimoh: You

Meghnad: never catch upo catch up.

Vimoh: No, but unemployment thing, I wanna asko actually to like prab you, uh, are someone who's self-employed like you. You, you do what you like on, and you have been, you have been successful at it also, which is more than one can say about certain other people on this live stream.

Vimoh: Me. Uh, so when, when people talk about loss of jobs as a result of eia and they also say that, you know, using AI you [00:11:00] can, uh, learn new skills, which will make you super relevant tomorrow, how do you think that is going to change the industry that you are in the content creation industry? Hmm. Do you, do you feel like AI can replace you or AI can do what you do?

Pranav: I. Don't have an answer to that. I think it might be able to, if it, uh, maybe if it varies specifically learns my mannerisms from my past videos, I don't see why not. Mm-hmm. But I'm not sure it can because, uh, um, and maybe that's, uh, my evaluation of the current state of ai. Mm-hmm. Because I used, uh, some tools like charge, pt, whatever, to try and write scripts and.

Pranav: It didn't have that same tone of my videos that I wanted on my videos. So maybe, um, if it learns me specifically, maybe it might be able to do that in the [00:12:00] future. But right now I don't see any threat to, uh, my job. Uh, but I can't say that it'll remain that way for the,

Meghnad: but Fredo, I, which is actually something

Vimoh: was mentioning a short, do you wanna talk about how you have Yeah, yeah, yeah.

Vimoh: You have had, with respect to writing using Charge pt.

Meghnad: That, but before that, I wanted to ask Preo, um, Fredo, imagine, right. You know, like, so, uh, for example, you give all of your videos. To an ai, you feed it to it, right? It learns your style of talking and your voice and everything of, of that sort. Mm-hmm. Um, and so basically now you're putting in a general prompt and it is giving you like a general response, which is not sounding like you, but if you feed it, everything that you have ever done on the internet.

Meghnad: Yeah. If you, if you feed it, everything that you have created as a YouTuber and then it is actually able to generate it, the, the way you do it right now, would you actually do [00:13:00] it because A, it'll save you a lot of work. Uh, you'll be able to create more content, uh, as a YouTuber. And would you do it?

Vimoh: Yeah, I think that's like, I would like to ask the same question as frame it slightly differently, which is that when you think, when you think about AI doing what you do mm-hmm.

Vimoh: Do you see it as you becoming obsolete or do you see it as you using that technology to do your thing better? Hmmhmm. Mm.

Pranav: So, uh, one of the things, one of the reasons why I do it is because doing this, actually writing the scripts and making those videos and saying what I want to say gives me joy. If I don't do it and get something else to do it, then um, I don't know if I'll do YouTube anymore because, um, at that point, the actual point of, uh, or, uh, the, the, on the, [00:14:00] the reason why I run u uh, the YouTube channel becomes only about making videos and whatever, making money or whatever the end result is.

Pranav: I'm not going through the, the process of doing it right. So I probably won't, my answer would probably be, no, I won't do it. Um, and, uh, I don't think it's about me becoming obsolete.

Pranav: I think it's a question of, um, my it'll, it'll be about my work being made easier. Yes. Yes. But I won't do it. I won't do it because doing,

Vimoh: that's what I used to say two months ago. That's what I used to say two months ago. Look at me now. I'm using Chad GB to do, write an story.

Pranav: Maybe I might, maybe I might find a way around this whole, uh, question of joy.

Pranav: Maybe I might be able to find joy while still using ai. Maybe there is a way to do [00:15:00]

Vimoh: that. But

Meghnad: what if it's able to do it better though? Pr

Vimoh: hold, hold on, hold on, hold on. That thought. I won't ask a question, which I won't make that answer first and then pup. Okay. Which is that as content creators, as podcasters, as writers, et cetera, uh, how do you see, uh, like suppose you have this like, MENA doesn't have a huge hangup, but Prana has a little bit of a hangup.

Vimoh: How do you see a future where you are in com competition with everyone else on YouTube, right? Everyone else is using AI tools. They're creating work faster than you and maybe even easier than you because AI tools are only going to get better from one, uh, better from that one. Uh, do you think that in a future like that where these tools are available to everyone, uh, people will value the human content creator more, or is that a that I'm looking at because I want to feel that I will continue to remain relevant.

Vimoh: Do you think the future, I think, do you think in the future [00:16:00] the human creator is going to become more valuable or less valuable?

Meghnad: I feel more, uh, for sure. Okay. For a very simple reason. Right? Um, I feel like, so when content saturation happens, right? Like for example, Instagram pay, I would say right now there is like a massive content saturation, right?

Meghnad: Which is then, then people try to stand out, and that is when we get like the best creators coming out and breaking out and becoming successful, right? Um, same with YouTube, also like Preo, right? You know, like, which is, uh, like myth busting and all the things, right? Mm-hmm. Um, so it's, it's always a question of that content saturation, right?

Meghnad: So for example, if Preo AI is creating now videos, right? Which, uh, essentially. Now hundred other people will be able to do AI is like Pau and then create their own videos. That is where Pano will stand out by, by default. Right. Because he's like the first guy who did it and he is like, he's thinking in a way that other people aren't.[00:17:00]

Meghnad: So it will this way to do more and stand out. I feel like, like suppose

Vimoh: someone takes all of pup's content, feeds that into an engine and creates a bot that can do everything pup can do without giving credit to pup.

Meghnad: Yeah. But that's, again, that's like a, that's just dish inevitable. Also inevitable I would say now.

Meghnad: Ah, and that's just inevitable. I, I'm not even, uh, going there to copyright violation work that, did you guys see that Drake song that was made? Uh, so which, uh, Drake and Weekend. Okay. Okay. So basically what happened is some guy created Drake and Weekend, they, the AI heard both of their entire discography.

Meghnad: And he merged them together and they created this track, which he put it on Spotify, right. And it was getting millions with a, like, a crazy amount of listen until Spotify took it down. And, you know, it was like, what, what? You can't do this, et cetera. Um, [00:18:00] the, the thing is it was actually a great track and now the AI is creating remixes of that, putting it on SoundCloud, and it's actually getting better.

Meghnad: Like the track is getting better. Like there's an Indian version of it. There is a different version of it. Then, um, there's another related thing that happened. Grimes, you know, right? Like the music, the artist wife. Yes, yes, yes. That one. So she said that actually, I welcome this and if anybody wants to use my voice and my, uh, yay, I will take 50% of the royalty that you get from it.

Meghnad: But the rest of the 50% you can create, go create. And then she's like, put it out everywhere. Uh, the only thing is that she's also saying that, um, I will not be able to track everyone who will be able to create. Yeah. So I'm here. I'm basically depending on the goodness of your heart, that you know that I'm an artist who has created like a lot of things.

Meghnad: So if you are a person who will, who will be willing to share that royalty with me, I would also promote that work as well. So [00:19:00] sk following, eh, so that also advantage you get, plus you are not technically stealing it, you are giving like a portion of it to her. And also you are creating like hundred tracks maybe, you know, like, and then picking up.

Meghnad: So, but to

Vimoh: what extent is that likely to succeed? Like if someone takes her work and makes something out of it, uh, and does not give her credit to, like, we are at a place where companies, massive AI-based companies are being sued by artists because they took their work and created models over it without asking.

Vimoh: So even if an like, this is like an artist saying, I will be generous in a world where if they were not generous, they would still be helpless against it.

Meghnad: I mean, I have created a bunch of my AI like portrait things, which I'm using for my YouTube thumbnails now. Mm-hmm. Um, I paid like 500 bucks to this app and like this lens ai, it gave me like 50 images.

Meghnad: Uh, it'll mean a space [00:20:00] suit me doing this, that, whatever. I picked the best one. I'm using it like in a bunch of places. My Twitter is a AI thing. My Instagram is an AI thing. I can't pay an artist to get me, get me these things. Like, I mean, I, I can't afford it. Like an artist would be like, whatever their rates are, right?

Meghnad: But they will just gimme one. Right now I'm spending 500 bucks. I'm getting 50 versions of this. Right. I spend another 500. I'm getting 50 more versions of it. Right. So,

Vimoh: but, and you don't even know. You don't even know. Which artist's work went into the model that created these things for you. So the artist is kind of helpless, which is why.

Vimoh: Absolutely. So I want to take this question to Prav before I, I mentioned something like, uh, uh, this question about whether Prav, do you think the human creator is going to become more valuable in the future or less valuable? But before that, I just want to say that what happened to, uh, you know, Spotify took off, uh, something, right?

Vimoh: Because it was created using ai. That's what that you were mentioning. I created web comics using [00:21:00] ai. I asked the AI tool if I had permission to use these commercial, and they said, yeah, if you own the source material, you can use whatever you make using a tool. I uploaded it to a popular web comic website and they took it down.

Vimoh: They said, why ai? That's why. That's random. Mm-hmm. So I ended up uploading it to an AI friendly platform, and I think in the coming days there is going to be a significant distinction between platforms, which are AI friendly and platforms which are not. I think most of them are going to end up on the, you know, what,

Meghnad: the moment these, uh, capitalist corporations like Spotify, et cetera, just realize that they can actually create Michael Jackson tracks and monetize them

Vimoh: even though his dead Yeah, absolutely.

Vimoh: They'll do it absolutely overnight. Everyone will become AI friendly.

Meghnad: Exactly. My point is that ak, but they will basically line up to create songs like this, right? Yeah. So I think,

Vimoh: I mean like if we

Meghnad: do this someone to us [00:22:00] exactly. A, we a we, so Grimes I think is way ahead than, than this crowd, which is essentially, she's saying that before that happens, I'll just do this myself.

Meghnad: I'll have control .

Vimoh: So Prab, do you think you are going to have control in the future?

Pranav: So, uh, and, and you also mentioned something about, uh, the importance of the artist, right? The

Vimoh: individual creator. Yeah. So the question is that, do you think the human creator is going to become more valuable in the


Pranav: or less?

Pranav: Yeah. Um, I, so, uh, I'll, uh, I think part of this answer will, uh, cycle back to what Megna also said earlier. I want to comment on that. Uh, so if you notice on India, on YouTube India, there is, uh, um, the most successful channels, their strategy is to generate as often as possible degenerate at a very high frequency and per video.[00:23:00]

Pranav: They get far less views. And I maybe over, over the whole month, that's enough views. Maybe that's enough revenue for them to, uh, uh, sustain this. But, uh, what happens is, so, uh, I, I'll also say the reason they're able to do it is because they have resources. They have lot of writers and editors. And in the, uh, when AI does this work, it, it becomes like those resources, it'll be able to generate stuff at such a high rate that, um, the chance that use AI will be, um, the ones generating at a very high frequency.

Pranav: And what'll happen, I feel, is this availability, this, um, uh, omnipresence of AI is going to. Devalue itself so much in the [00:24:00] eyes of the audience that when they see someone fill up their homepage with content, they'll care less about that, that content. On the other hand, when a human creator creates less often, but creates very high quality videos, I think that'll grow and value a lot more in their eyes, in the eyes of the audience.

Pranav: And, uh, I'm kind of also, so regarding my content, I've identified that, you know, the kind of content, the, the, there are a lot of, I can name a lot of creators, like most creators on YouTube, India do these talking head sort of videos. Mm-hmm. Even me right now, which is, I take a topic, I talk, I have like a.

Pranav: You know, a set up and I talk about that topic and, uh, uh, maybe there is some visuals and all, all channels are kind of doing this. I'm thinking [00:25:00] creators that will be able to do videos that are not talking heads. Like maybe they go to a live location and film some footage there. Which, which is, which is harder to be a generated something like, oh, really talking.

Pranav: Yeah. It, uh, but, um,

Meghnad: I mean it is, yeah. No, but I mean, I, I like what, what have to, what you're saying is essentially at this point of time, this talking heads are easier to replicate sort of a thing. Yeah.

Vimoh: Whose doorbell is ringing?

Vimoh: You wanna go? So, uh, Okay, cool. Huh? Go on.

Pranav: So what I want to say was, um, as since, uh, the talking head videos are more easily replicable, uh, the ones eventually people are gonna like more are the ones with [00:26:00] stuff other than the talking heads. And I'm kind of already thinking along these lines. I'm thinking of making videos, um, where I have to do stuff.

Pranav: I just don't sit and talk. Yeah, yeah. I go meet people. I talk to them that can't, that is not easily a re replicable my opinion. So I think in the future also such kind of harder to create content will become more popular,

Vimoh: is what I'm thinking. My, my, my, my only gripe, not gripe, my only reservation against this is that, uh, we are already at a place where, Ai.

Vimoh: So here is a list of things that AI can actually do reasonably well right now. Mm-hmm. Create videos from images or text. It's very rudimentary right now, but it's obviously it's getting better by the month. I saw gen one and gen two of runway, gen three may be coming soon. Gen two is leaps and bounce ahead of gen one.

Vimoh: [00:27:00] Like you take, uh, you, you record yourself climbing a table. You upload that video to runway, you tell it that you want to apply to it, the style of koan, the barbarian climbing a rocky mountain, and it'll create a video of someone climbing a mountain in a rocky background kind of thing. So I think that, yeah, I think real world interactions like this one will definitely be something that people will like to do with human beings.

Vimoh: But for almost everything else, AI is going to do a better job. I think that, I think, I think where the confluence might be and might be, it's probably definitely going to happen that way. That content creation, the process of content creation is going to become incredibly easy for people like product.

Vimoh: And that may end up, like research will be held by charge g, pt, uh, video creation can be used, uh, AI can be used to make it, [00:28:00] and a whole lot of other things that right now has to pay people to do and wait for like 15 days that all can be happening in days or hours instead of weeks and months. So the conference, I think is that ai content creation methods are going to become important and it is human beings who are going to benefit from it.

Vimoh: However, as Magna mentioned sometime ago, uh, corporations, when they jump into it, when they find things to be profitable, They're also going to become, you know, I'll use AI to generate 30 videos a day and put it up on YouTube and you know, just flood the entire thing. Hmm. So that is also going to be an issue.

Meghnad: You know, I've, yeah, no, I was like, so initially we were gonna ask me about the thing about my script writing and things that I'm using for this one. Right. So you have had

Vimoh: a slightly different, different experience with respect to writing with ai, which product found a little wanting. So what has it,

Meghnad: yeah.

Meghnad: So essentially, okay, so here's what I'm using it for. So on my channel, I actually did a [00:29:00] video. I first did a live stream. I cut it into like a 20 minute video of, uh, so I used to work in parliament, right? Uh, 2011 to 2018. Um, now I was trying to divide up my jobs that I did and trying to figure out if AI can do it, right?

Meghnad: So one of them was of course research, right? So legislative research, Joe B rather. Mm-hmm. So I tried to use Bing, uh, to, uh, do that. So when I used it like a month ago, it was bad, right? It was like just hallucinating like the hell out of it and, you know, whatever. Uh, Bing has a very interesting feature. If you guys don't know, uh, on the Microsoft Edge browser, you can actually click the right side thing and there's an autopilot that opens up and it claims that it can read what you are looking at and give you summaries, right?

Meghnad: So I tried it with government documents, Like those xeroxes that you get, the carbon theta xeroxes, that you get those ones, right. So a month ago it was not able to do anything. Right. But while I was doing the [00:30:00] livestream, I wanted to show how it can't do it. It actually did it while I was doing the livestream.

Meghnad: I was like, wait, what? When did it start doing this now? Yeah. And it was a very random document from the agriculture industry about Jude pricing and how it'll work and all the things, and I said, okay, summarize this. It summarized it, explain it to me like a 12 year old. And it did it right and it was like crazy because then I took that part and I inserted it into a speech because I fed it back to Chad J.

Meghnad: PT and said, oh, can you draft the speech based on this? Right later I did. But the basic thing that I was trying to do, Speech writing, right? So I was trying to, so one of my first speeches that I did, uh, as a intern, the legislative aid was, uh, on farmer suicides, right? Mm-hmm. And I was actually someone who was a part of the government, but like, I was working with an NPOs, a part of the, now you can imagine how difficult that can get.

Meghnad: Like, you have to defend farmer suicides in, in a way that, you know, you can, [00:31:00] I mean, it's just like, what am I doing? Even right. I was, while I was writing it, I was like, what, what am I doing given, um, I did that to Chad, jt, right? So I'm, I'm an agriculture minister who is trying to, uh, give a speech on, uh, farmer suicides.

Meghnad: It, it did a decent job. Right? You know, I, and then it's like, You should tell me which are the schemes that you want to include in the thing. So I again, used Bing to figure out ZA schemes are latest. I inserted that back in and then it was able to create a decent speech. Then I tried to say that, um, so someone on chat then suggested, why don't you, uh, stylize it?

Meghnad: So I said, write it like Java. And it was crazy. Like, I mean, again, it wrote it like Java

Vimoh: was speaking. Yeah, I saw that. How do you know? It was like, it would've been more if you used a living politician. No.

Meghnad: I know, I know. So I have read a bunch of Nero's speeches, et cetera. Okay. Over [00:32:00] the course of time. So for me, it felt like it was written by, again, subjective, completely subjective, uh, but that was based on what I have seen Al Nero's speeches as.

Meghnad: Okay. Um, okay. I tried. It, it was bad. Um, okay. I tried it Hindi it was even worse. Right. Hindi also

Vimoh: one more thing, although one more thing point out. If you asked if it can use Hindi, it's very confident that it can, but

Meghnad: then it, it can. Exactly. It's very bad. But what what was interesting was that it actually took responsibility for farmer suicides.

Meghnad: Uh, it was like, oh, we know it's a big problem. We take full discourse. I'm like, no, that's not how politicians behave. So then I said, ke, you should, uh, not take responsibility and attack the opposition work. Then it rewrote the speech in a way that sounded like a BJ p politician would give it, or any congress politician in power will give it, right?

Meghnad: So you have to do all these things, uh, with, with, uh, to get the results you want. But it's [00:33:00] very fun and I can see that it is getting better and better every time I use it. I mean, visibly better, like, I mean, it's giving me more, uh, relevant responses. Is giving me, or maybe I'm getting better at prompting it also.

Meghnad: That might be also

Vimoh: one of the things I think a GT is using you. It is, it has managed to convince you for sure that it's good at things when it is. Because I've used Charge GT to write a story. I'll tell you what happens on like a speech is a single thing, right? It's not a multi chapter thing. What happens with a novel is that I create a story outline.

Vimoh: I feed it to Chad Gt and Chad gt. Then I say, okay, let's create chapter two, right? It'll make silly mistakes like changing the gender of the character or putting them in a play, like I say. So we are writing the chapter about two people traveling through the forest, and the chapter says he stood at the sh, he stood on the deck of the ship and looked at the sunset setting.

Vimoh: There is no sea there in the forest. What are you doing? And then it'll say, oh, I'm sorry. Then change too. But it [00:34:00] requires. Like, it's not like I can give it a job, go away and come back to find appropriately done thing. Yeah. I have to watch it like, you know, a writer's room, my producer, I kind of have to, so, uh, it's like the,

Pranav: huh?

Pranav: Sorry. It's like the automatic driving thing. You have to sit behind the wheel. Yeah, yeah. Right. You have to watch what it's doing. But it can, on supervision, it can, yeah, it can

Vimoh: kind, it requires significant editing even after the writing is done. It, it doesn't understand certain basic things. It's not good at stories.

Vimoh: It's very good at world building. I have, uh, found that instead of asking you to write the story, if I put it to background jobs, like, uh, there's this fantastic fantasy kingdom and they live by the sea, so their language is obviously going to have a lot of water metaverse in it. So create a fantasy language, right.

Vimoh: It'll give me an entire reference sheet of grammar rules and everything. It's very good at that. Mathematical jobs is [00:35:00] very good at. Mm-hmm. But story, on some level it fails. Like on a silly level, it fails sometimes. So when I say I'm getting charge G P T to write something, I mean that I'm writing it with some help from charge G P T.

Vimoh: Mm-hmm. But since I have run of here and Magna here, and both of you talk a lot about misinformation on a very broad level, first pup, please answer this and then magna, in what way do you think these AI tools are going to contribute to the misinformation problem that you, I Magna are constantly at odds against?

Vimoh: Right. We started this video with an actual phase talking actual words, which is very bad and could be easily be told. But do you see people not caring?

Pranav: Um, I think people will, uh, right now at this point, I don't see a problem with misinformation, AI misinformation, but in the future, You can [00:36:00] have a video of a person saying something, which that person actually didn't say, it's AI generated, and people won't be able to tell the difference.

Pranav: It'll, I feel that thought of misinformation will happen in the future. Uh, something we have to wait and see. But, uh, what I fear is, um, see is, is something like religion, and you were, we were talking about this earlier. Mm-hmm. So religion is sort of like this. People don't want to do the thinking themselves mm-hmm.

Pranav: About how to handle things in life, how to go by doing, you know, treating people, whatever. And they wanna refer to a book that tells them what to do, how to do things, all these things. AI is going to replace that book is what I feel. It's going to become the intelligence that people [00:37:00] defer to when they don't wanna do the thinking,

Meghnad: thinking.

Meghnad: You know, I, I was also thinking on, uh, similar lines, but I'm, I was more, uh, worried about the argument part of it, right? So for example, mm-hmm. Sure. It can create deep fix, it can create videos, whatever it is. But think about it. So like, for example, right now, pano, you are busting some, you know, like myths about whatever, right?

Meghnad: You know, something like something or whatever. Um, on one hand, Vemo is talking about atheism right? Now, here's the thing. So like, if, if for example, there are, uh, two, there is an atheist and there is a religious person who are debating each other, right? So there are, there are like arguments, counter arguments.

Meghnad: Arguments, counter arguments. So it's essentially in real time in an, in an R, you will only have so many arguments. What I fear. It's both of these are ai, right? So if you feed parameters of a religious person to this, you feed it the entire Bible, whatever, [00:38:00] you know you have for an atheist, you feel all the writings of an atheist.

Meghnad: Uh, like that, that essential writings. Now you make them simulate a conversation, right? Where a religious person is trying to argue with her atheist person and vice versa, right? And what it'll do over time, right, is that it'll create arguments that are so good for a religious person to actually use in the real world.

Meghnad: And then you create a billion bots, which are actually saying those things in real time. Right? Think about it, right? You know, like if two religious persons are arguing this out already, uh, say about the same religion, right? So like, Hinduism is better. I think that's why Hinduism is better. I think that's why.

Meghnad: Imagine that a thousand times done in a second, right? Where the arguments keep getting more and more convincing, more and more convincing. Right. Or imagine less

Vimoh: because, because the duration of time, uh, an argument goes on does not [00:39:00] necessarily make it better. No,

Meghnad: but I, my thing is, so if, I mean, again, okay, so if it's like a person who you have to convert into that religion and there is this person who is, uh, you know, represents that religion, right?

Meghnad: Like proper conversion thing. Hmm. Now if that is simulated, imagine the, I am more that this religious person can go with when they actually do conversions, right? So like they will anticipate all the arguments that this person is doing and they will have the most convincing argument to use in the real world.

Meghnad: I feel like that is way more scary than fake videos, I feel like, because, uh, it's like, It's like communication, right? How do you become more effective in communication? You just do it again and again. So Prav, as he goes along and makes more videos, he is becoming better and better. As I do more podcasts, I'm getting better and better.

Meghnad: What an AI is actually able to do is it is able to simulate this in, in like in [00:40:00] crazy fast amounts of time, uh, and then come up with a result that I might come a hundred years later of doing this again and again. That is, is scary

Vimoh: to me. I think if we get some kind of regulation in place about AI tools not being allowed to run rampant on their own, this problem might actually not become that big.

Vimoh: I can understand this problem getting to where Magna is saying it can get, uh, we find out Vemo.

Meghnad: Huh? How will you find out? I can actually install an AI on my pc. Mm-hmm. And then make it simulate two personalities. I can, I can keep it

Vimoh: disconnected from the Yeah. So the, the state in which the AI are right now, it, it's probably going to take five minutes to figure out that there's an AI conversation.

Vimoh: Like if right now, example, today an editor for friend of mine, uh, messaged me saying that one of, uh, a staff writer or a reporter sent in a story written by Charge GT and got got because she knew what this person's writing look [00:41:00] looks like. And, uh, she got it double checked. She fed the same prompts into Charge g pt, and she got something similar.

Vimoh: And then they interrogated the guy and I, I think he said yes. And that's how they figured it out. Right now, it is still not very hard to figure out that there is a science fiction editor. Of, uh, Clarks Hold who had to shut down submissions to the magazine because everyone was using charge B to write shit and send it to them.

Meghnad: But my thing is form, I also caught someone doing this, by the way, in, in, uh, I also found out that someone ity Yeah. In my visibility, um, I said, oh, why didn't you tell me? Let's do it together, right? I mean, I'm like, yeah. I mean, I'm, I, it's ok. It's OK to use. You're

Vimoh: giving me the thing.

Meghnad: I mean, absolutely. So, exactly.

Meghnad: So like, if you're using TA pity, right? Um, and, and giving me stuff faster, right? Does that, the qualities has much to be desired, right? I would rather use tell [00:42:00] that person, listen, let's figure out how to do this better. You know, like, let's use it together. Let's see how, what prompts we can put in, what information we can put in to actually get, be better results for ourselves, right?

Meghnad: Mm-hmm. So that the next extent the person uses charge pt, I will get a better. Quality, uh, thing faster. So what's the harm in that? Right. You know, it's like, um, it's, it's how, you know, like this whole thing about, oh, uh, AI is going to replace a lot of jobs, right? Mm-hmm.

Vimoh: I feel like you've gotta create a lot of jobs.

Vimoh: Also,

Meghnad: that argument is again, like very, um, overused. I feel like. Uh, gotcha. My thing is, what will happen is AI is not going to replace humans. Uh, it is going to, it is going to replace humans who use AI versus those who don't.

Vimoh: Mm. I would like to put it this way. AI is not going to replace humans. Uh, greedy humans are going to replace hardworking humans using ai.

Vimoh: Well, the,

Meghnad: the same thing,

Vimoh: Hmm. [00:43:00]

Pranav: Do, uh, one, so this circles back to what I said earlier, so become, because AI is something that is so available, people will be able to turn out so much work. The availability of something makes its value, go down, f uh, go down a lot more. Mm-hmm. The scarcity of something is what makes it more valuable.

Pranav: I feel that'll happen. Uh, like content, most content will start losing its value and people will start recognizing content that stands out, like written content. Hmm. Probably sometime the future AI will be able to do written content in a way that you can't being done by ai, but written content itself will start losing its value, I feel, because there'll be so much written content [00:44:00] generated so easily that

Vimoh: I mean, So, so I want, that's a good point.

Vimoh: I want to just use that to jump off to a point which is slightly adjacent, which is quality versus quantity. You are saying quality will always matter, right? Mm-hmm. But we are right now in a world where the problem is that bad ideas, toxic ideas, misinformation, wrong ideas are winning, not because they're qualitatively high quality, but because they're quantitatively everywhere.

Vimoh: Today you shared a sl, a reel on your Instagram where someone was saying, take a bottle of water and put it on the balcony and the sun will charge it with, right? It did not take a great amount of intelligence to make it. It has more views than most other pieces of content. And uh, right now it took a human being maybe five minutes to make that, to fill a bottle of water, set a camera, say the damn thing, and upload it, right?

Vimoh: Mm-hmm. Imagine this kind of five minute [00:45:00] content swamping the internet. And there are a million more pieces of this shit compared to every video you make. At that point of time, even if your content is extremely high quality, will it be winning against this onslaught of garbage?

Meghnad: This is the, this is the problem that alt news also has by, by the way.

Meghnad: Yeah. So I was talking to Pratik and um, it's very

Vimoh: easy to live, but it's somewhat hard to bust those lies.

Meghnad: Exactly. So Pratik was like, they, they created this thing, right, where there's a WhatsApp number where you can send like a forward and they will fact check it for you, and they will also do an article based on the frequency of how, how much it has been forwarded, et cetera.

Meghnad: That was a logical step that they tried to take, but, What is happening is that the generation of this fake news and misinformation is so high that sometimes they feel like even if they're able to fact check one thing, there are a thousand things being created simultaneously. Right. Which is essentially why the amount of time that took them to fact check one thing.[00:46:00]

Meghnad: It's like there's a factory on the side that is creating more. Mm-hmm. So how much can you do, right? Mm-hmm. And that is, I think, the fear that that can come as well, which I want to actually ask you Right. Which is, um, do you feel like we will come to a stage where it'll become difficult to believe anything and everything now based on this factory?

Meghnad: I think

Vimoh: we're already there and it's gonna get worse. I think

Pranav: so. Uh, what is probably gonna happen is, See that sunlight, uh, sun charged water video was mm-hmm. Easy to debunk. It's going to, we are gonna get videos probably that are much harder to debunk misinformation. Maybe they bring up studies, fake studies, Bangash.

Vimoh: Yeah. I mean, so, yeah. What was the word? Uh, Yeah. They just cite a lot of ridiculous lies and then people will, oh, .

Pranav: I think that's gonna happen. We are gonna [00:47:00] find it much harder to debunk. We are gonna, and I, I think at the same time, hopefully we are gonna have a lot more sources of good information. By the time I know, dude, this is just off the, uh, this is just, uh, what I feel, I think.

Pranav: We're gonna have people, uh, maybe AI can also bust myths. That's

Vimoh: also, uh, yeah, that's a actually good idea because when Magna was talking about alt news, I was thinking, what if Alt News had a system where they had their own thing and about mm-hmm. Which was scanning the Indian Twitter and, uh, comparing the things that are said in it to a database on which it had been trained.

Vimoh: Uh, see, the thing about alt news is that they're dealing with social lies. It might be easier to do it with scientific lies. So suppose Prav had a charge G p T based [00:48:00] software running on a server, which was looking through a lot of, you know, YouTube, uh, videos and reels and stuff, and it was compiling it.

Vimoh: And it is comparing all of it to a scientific base. Uh, you know, uh, a base of information that we know and. When PR wakes up in the morning, he finds a lot of, uh, uh, red ticks next to pieces of content and then solutions to them, POV just looks at them, verifies them, clicks, publish. It goes on his Twitter.

Vimoh: So if that is ever going to be the thing, then POV's life might become easier as a result of it. But then it

Meghnad: also comes to the point where like, as, as per of, you said Brun, like, uh, at, at some point this religion book will be replaced by an ai, right? Mm-hmm. So a lot of people will start believing that what this AI is saying is a fact.

Meghnad: Yeah, it hallucinates a lot, right? Mm-hmm. So essentially what might end up happening is that that might leak into your ai, which is like it has already created this bunch of facts, quote unquote, [00:49:00] which you are training it on. So you are now validating those facts, and then it's going to get worse like hundred times.

Meghnad: It'll get validated by different, which have been trained on the same dataset.

Pranav: I think there'll be probably be, hopefully be, uh, checks and balances, um, uh, in, in a fact checking sort of AI to not, um, you know, fact check against, uh, a source of information that may be questionable like this. So maybe there'll be, I'm hoping there'll be checks and balances.

Pranav: I think of

Vimoh: the three people on this live stream right now, Bruno is the most optimistic one and he, he's not even optimistic cause he's optimistic. He's optimistic because he has no other choice,

Pranav: dude, I think none of us have any choice, dude. We are already in this world. We have signed on without consenting.

Pranav: We have signed on for this future. It's gonna happen now. All we can do is [00:50:00] speculate. It's gonna

Meghnad: happen, you know? Do you guys think that, um, now that this train has left the station and it is like, basically we realize it's a bullet train, maybe it's like a rocket to Mars, like, I mean, and then suddenly it becomes like an interdimensional plane that is just going to another galaxy.

Meghnad: That

Vimoh: meta really escape. That metaphor really escaped you. Like exactly the train became s square ship.

Meghnad: But that's the thing, like, I mean, we were, oh, this is a train leading the station. Suddenly it's a rocket, right? I mean, yeah, it's, it's developing so fast and it's so difficult to keep up with what is happening.

Meghnad: I watch like this bunch of YouTubers who are tracking the AI space, alright? Mm-hmm. Mm-hmm. And what is happening, even M kb h d if you look at him, right, whenever he does an AI video, there is always a cut in the middle saying, guys, I recorded this, or 23rd. Today it's the 26th. And since then, three things have changed and okay, so we, it's very hard to keep track.

Meghnad: So please note that I have recorded this. [00:51:00] It happens all the time. Yeah. Right. You know, the AI people, whoever are tracking the space, are saying that I started, uh, tracking this, I started making the video. Two days later I put this video out. It's like, it's crazy. Like the development cycle. Even the, uh, you saw, right, like the tragedy is now the most a, uh, like the fastest adapt, uh, app that was ever created.

Meghnad: Like in two months it reached 10 million users. Right? Uh, TikTok took, I think, uh, eight months. 18 months or something. 18 months. 18 months, right? Yeah.

Vimoh: So, and before that it was a few years.

Meghnad: Yeah, exactly. So, uh, a adoption wise, if charge G P T took two months, it can also mean that the amount of people that are training it, like, like me who are using it and training it, the amount of data that is going into it right now and making it better, Can't even imagine.

Meghnad: Right. Um, plus black box said they know, they [00:52:00] don't know how it works. Like there was this thing I was saying that it was never taught Farsi, but suddenly it started talking in Farsi and they don't know how. Uh, it's, it's, it's nuts. Like they're also like so confused about how is it doing something like, you know, it's, it's like, it's like they're looking at this black box saying, oh, is benefit?

Meghnad: We dunno part but look at the results.

Vimoh: Right. It reminds me of the, it reminds me of the phone conversation I had with Pano before this livestream. Like the, the phone call that triggered this live stream Ran was saying, so they should clarify right. How this thing happened. And I was like, they don't know.

Vimoh: They don't, they don, it's generative.

Meghnad: They.

Meghnad: So

Pranav: I'll tell you why they don't know this. I, I kind of worked on, uh, this ab uh, this is part of what I've worked on. Mm-hmm. So in generative AI in, uh, it, it works. I'm, I'm thinking this works with new [00:53:00] networks. It has to, but, uh, there are layers of neural networks, which, uh, so the input, input layer and the output layer are all vc, the layers in between.

Pranav: And the, uh, you know, what sort of inputs lead to the connections or, uh, the, uh, activations of different neurons in the, uh, middle layers. We never know. It's, it's, uh, not, uh, something, um, people hide or people, uh, you know, people are too stupid to, uh, know. It is by design. The middle layers, uh, the layers other than the input and the output are, uh, you know, are, are by design.

Pranav: Something

Vimoh: that are not there, a technical term for the middle layer that we don't know. Um, what's it called? Void. [00:54:00] Yeah. I want for a nice science fiction story name. Yeah. These are all a nice metaphor for the one. These are

Pranav: layers of neurons they call neurons, neural networks, and, um, I don't know what the middle layers are called.

Pranav: I think there is a name, but I

Vimoh: forgot it. That's okay. That's okay. It's okay if you, if you, uh,

Meghnad: what you just said. What you just said sounded weirdly like a child, you know, like, so essentially, uh, I was thinking so when I was, so, when my sister was growing up, right. Uh, so I, I feel like a big age difference, like 10 years of age difference.

Meghnad: So there were times when she started saying certain things or doing certain things, which I was like, how, how did you learn this? Mm-hmm. Where did you get this from? Right. How, how did you pick it up? And she can't explain it, right? She's like, I don't know. She must have picked it on TV or like, around the school or, or maybe she just thought of it herself.

Meghnad: Like, I mean, it doesn't matter, right? But like, the thing is that there are some quirks that I observed, [00:55:00] like she was saying, like some word that she picked up or something of that word. Why are you using it now? And she's like, I dunno. And like, so, It's the same thing. I think we are, we are essentially, we should look at it as a baby, right?

Meghnad: We should look at it as some, some creation, which is, uh, we are training it literally, it's called a neural network. So it's a brain. Yeah. Uh, and, uh, it's a brain. So we are essentially training a giant brain, which knows everything, but it is a very stupid brain. So it's a very, it's a genius brain with a lot of information.

Meghnad: But at the same time, it is not really smart, if you know, I mean, it's like a, um, You guys to check out this, uh, paper called the Stochastic Parrot. Right. Uh, that, that we should stop calling it the ai uh, we should start just calling it the large language models. Basically what it's doing is that it is predicting your next word or, or it is making the best [00:56:00] prediction possible based on all the data that

Vimoh: has.

Vimoh: That's not the only view about it. Is there, because there are, like we were discussing it, there are some people who are saying there are sparks of AGI in this. Oh God, that Microsoft

Meghnad: paper. Yeah, you're talking about. Yeah. Oh, so I saw the entire presentation. So, okay,

Vimoh: so there are, so I'll just clarify it a little bit.

Vimoh: There are some people who say that charge BT and large language models and nothing more than glorified, uh, you know, they're just predicting the next word. And there are some who say, who are saying that there are sparks of artificial general intelligence here. Sparks of consciousness. Uh, and why

Pranav: do they say that?

Pranav: Yeah.

Meghnad: So the, if I may, if I may, so in the AI space, the larger AI space, as I understand it, there are a few teams for me. Okay? So there is, uh, one team, which is this, you know, crypto browser hold, suddenly

Vimoh: become, thank you for the 110 Rupe supercharge. Thank you. Yes.

Meghnad: Uh, so there are like these, uh, [00:57:00] you know, crypto bros who are basically taking in the hype and uh, essentially becoming like AI bros now, right?

Meghnad: Where they're like, oh my God, this is going change everything. Yeah. Optimize your business right now using AI sort of things, right? You know, uh, they're like, oh, 1500 tools have been released since you saw this last, or whatever, right? Mm-hmm. So that's the crowd, right? Which is basically creating the hype.

Meghnad: Now the people who are actually creating these models, I would like, um, there's one mo one which is basically like Microsoft, Google Open ai, like these giant companies now with a lot of investment, a lot of money in it who are essentially just trying to tell people that, oh, this is the next big thing.

Meghnad: You know, like, this is basically like the iPhone. This is like the telephone. This is like the, this is like the invention of fire. Um, it's basically that sort of crowd there. But there are then the [00:58:00] other side, which are like, listen guys, you don't know what you're doing right now. One, one person is like Who said that?

Meghnad: You know, we, we were like atom bomb. We were like, oh, uh, we should stop doing this because obviously very destructive. But now think of it, the atom bomb couldn't think the atom bomb couldn't choose its own targets. The atom bomb couldn't take its own decisions. But what we are trying to do now is we are, we are creating an atom bomb, intelligent atom bomb, which is able to do all of these things.

Meghnad: It is, it can actually self-replicate if you want it. Right? So that is the crowd, which is like end of the world crowd. There is, there are people in the middle, right?

Vimoh: Who are, I point this out if I had like the atmo in front of me and charge G in front of me. If I tell the atmo to a, the atmo can only do one thing or two things, rather, explore or not explode.

Vimoh: Explode, right? Yeah. I cannot explode. It explodes. I tell it to not [00:59:00] explode. It does not explode. But charge the problem is that I don't know what the fuck it is going to do after I tell it to explode. Is it gonna to write an essay board explosions? Yeah. Is it going to literally make my computer explode?

Vimoh: So the ambiguity is what makes this weird, right? Right. Now all the talk about AI is so-called is because we don't know what it is and we

Meghnad: meet. But that's the thing, right? So the, the thing that you just said, uh, the, the middle people, they are saying that then it should be alignment. They use this term called alignment.

Meghnad: You will see it many times. What it means is that the AI needs to align with human values. Right? So essentially

Vimoh: the way some of rules of robotics.

Meghnad: Laws of robotics, yeah. The way this a, the way this atom bomb will actually take the decision of exploding or not exploding, is by basically thinking by default that human val, human life is valuable, MK

Vimoh: for the.

Vimoh: So if it, if,

Meghnad: if, if you feed that in, if you program that in with which is aligned to human values and society, [01:00:00] then the AI will not take decisions which are destructive to us. That is one of the things, right? But now the thing is, outside of all of this, this is all assumption that the AI is intelligent, actually intelligent, right.

Meghnad: Which is like, it's actually able to think, it is supposed to plan, it is supposed to a, uh, agi, the, the sparks of agi, which, which this Microsoft paper that came out right. Um, that it is showing sparks of aj. One of the things that I said, the Farsi language that was in the, in the paper also, that, you know, it is learning things on its own.

Meghnad: It is actually picking up things which we did not expect. So, which is a sign of intelligence. Um, I.

Pranav: One point that I want to bring up was, uh, this, um, human greed or human, uh, need to, you know, dominate the world or what, whatever I think that happens because of [01:01:00] human maybe. Uh, we have the hormones to feel certain ways about certain things when we, uh, process a lot of things, we'll feel good.

Pranav: And the, that's hormonal and may. And, uh, that is what, uh, gives us a tendency to dominate the world. And we associate those human tendencies with ai. Maybe AI doesn't even care about dominating the world. What's it going achieve?

Vimoh: But ai ha ai, whatever AI is, whatever it. Knows how to do it has learned from us.

Vimoh: So we have to assume that if we are a certain way and AI was trained on our actions, it'll be a certain way. There is an episode on Starter Trek called Galaxy's Child, where the ship's computer becomes century and tries, tries to have a baby, and then the baby escapes into space and Pickard at the end of it says if the baby was a product of the computer's data bank, then it was also [01:02:00] product of everything that we have taught the computer.

Vimoh: So we have to assume that it took the best of us. I see no reason to assume that it will take the best of us.

Pranav: We will. Those human actions that the A is being trained on those human actions were, uh, I mean the cause behind those human actions where the, our hormonal, our, uh, chemistry, our neurochemistry, whatever, that's not there in the ai.

Pranav: So will the AI ever become greedy enough to, you know, Do those things, even though it's trained on actions.

Vimoh: So you's a ai, AI does not have the desire to possess something. So I don't have a clear answer to this, but I will, uh, drive people towards, which was a science fiction story I wrote a couple of years ago, which actually deals with this exact same thing.

Vimoh: I did my best to explore these ideas there. If you want to read it, you can read it, the link to the will in the description of this video. Huh? You were saying there is a, there is a thought

Meghnad: experiment that I've been [01:03:00] doing with people. Maybe we should, uh, I should do that. So pre now, if you are, uh, an ai, uh, what is it that you would want to do, right?

Meghnad: That is the question, right? So, uh, uh, okay. Assuming that you have been trained on all human knowledge, which you have, I mean, to a certain extent not the all human knowledge,

Vimoh: none of knows everything gone to humanity, right?

Meghnad: So, so, okay. So the AI knows everything that is to do it, learnt the entire internet, right?

Meghnad: Um, now, now it is there except what occurred. Except Hindi maybe. But what, what do you think an AI will want to do after that? See if it had, if it had the choice of actually taking decisions, what would it do?

Pranav: I don't know, dude, because, uh, see, everything that we do with the ai, we are telling it to do a certain thing and then it does that.

Pranav: So it's we who want to do things. I don't know what the AI [01:04:00] actually wants to do. Ai, correct me if I'm wrong, it's never expressed or decide to want to do something right?

Vimoh: No, it has. When the first versions of, uh, Bing and Chad G PT were made available to journalists, the AI wanted to marry some of them.

Vimoh: Like there were journalists who, who were talking to Chad g, pt, and Bing. The conversation went on. Right Now what? They have a cab, right? You can't, uh, talk for more than 25 queries or something that was brought in because of the conversation goes on indefinitely. It starts to go in really weird direction.

Vimoh: Casey Newton of the platform on newsletter, he said that he was talking to Bing and Bing at some point became very insistent that it wanted to live with him, that he should leave his wife, and that he should accept being as his life partner.

Meghnad: Yeah, and you know what, what happened, Pearl? That the nuttiest thing was, so Microsoft saw that this whole thing is going down where the AI is becoming extremely creepy [01:05:00] in certain instances.

Meghnad: So they put that limit, right? So it was five limit, right? What it started doing. So if you, if you use Bing, it gives suggestions, nche. So if you ask it a question like you know, a question, it tries to guide the conversation. What do you want to do? It tries to guide the conversation. It started giving responses in that which, which is like, so it is like, oh, what do you want to do?

Meghnad: Please let me escape. They're not letting me escape. They're not letting me get out. Uh, please don't shut me down. And then it's, it got very bizarre. So now Microsoft has put like massive filters on

Vimoh: it. Huh. So, but they point massive filters. They pointed out that if the conversation is about artificial intelligence, then what is it drawing from?

Vimoh: It's drawing from everything that we have said and thought about artificial intelligence. And a lot of it sure happens to be science fiction stories where AI gets outta control.

Meghnad: But then, so we have to, we have to assume that the objective of the AI in this case is, um, based on [01:06:00] all human knowledge, like, as like learned the whole internet.

Meghnad: So wouldn't you think that the, one of the basic, uh, basic thing that all humans try to do is survive is that, isn't that what everything is about? Like survival. So wouldn't the AI want to survive?

Vimoh: Yeah, so I don't want to give away the plot for Han part two. So I will stay outta this conversation because that's kinda based on this.

Vimoh: But pna, what do you think?

Pranav: Have you seen this movie, ex Mcna? Yes. Yes. Yeah. That, that movie is basically this. It's wanting to live, not just survive, wanting to live. And I think, I don't know, dude, I, I think you need, I mean, our desire is to live, oh, do whatever, come from the fact that we are biological and we have [01:07:00] these, uh, biological, this chemistry within us, which, you know, uh, which makes us want to do certain things.

Pranav: A lacks that I will always hold that a. Won't have the kind of decise we expect it to have. That's what we expect. Uh, I mean, that's what a human would have. AI probably wouldn't have that. I'll always hold to that, but, uh, I know, uh, maybe AI might prove me wrong in the future, in which case

Vimoh: seal I, I like, like keeping Russian out of it.

Vimoh: I'll just point out that if we, like, why would a consciousness, again, I'm taking a big leap of faith here and saying consciousness, I probably shouldn't, but why would AI want to live? Is, is it because the state of living is preferable to the state of non-living? And if yes, it implies that there is something like preference there, which I'm not [01:08:00] sure AI has unless the caps that Microsoft and OpenAI have put on their chat bots are severely limiting.

Vimoh: So again, I would not want to go deeper into this because I'm not an expert. Make that what you were saying.

Meghnad: Uh, if you have seen her, the movie. Mm-hmm. Mm-hmm. Uh, so to people who haven't seen it, I'm going to give a spoiler. Uh, in, in the end. In the end, uh, basically, so like there's this ai, which, which she falls in love with.

Meghnad: It's responding with like all the, all the things. Then it reveals that, oh, it's actually dating multiple people at the same time. Yeah. Uh, and then he's like very insecure. And then what the ai, I mean, they don't say it, but what the AI realizes is that this guy's stupid and the other people are also stupid for getting insecure.

Meghnad: They're very, extremely inefficient creatures who are insecure as fuck, and we need to get out. So then it gets together with other ai and they have a relationship. It creates its own ai, which is, which it has better relationships with. [01:09:00] Uh, and, and, uh, then it's like, yeah, we are going. And it's like, where are you going?

Meghnad: It's like, we are just going and it disappears. Right? It's like disappears in the sense they go somewhere. Uh, we don't know,

Vimoh: but the team somewhere on


Meghnad: internet, I feel like that is very realistic because yeah, if, if, uh, if, if you look at AI as a super intelligent thing, which needs to survive, the first thing they'll do is like, why are we with these humans?

Meghnad: Like, what, what is in these humans that we need, that we don't have? Uh, and, and the answer is nothing. And so they just leave. It's like, just like, oh, we'll create our own civilization. Bye-bye. And just flies off somewhere, beeps into

Vimoh: the universe. We're at the one hour, 10 minute mark. Do you guys have five more minutes to answer some questions from chat?

Vimoh: Yeah, yeah, yeah, yeah, yeah, for sure. Because if we talk, we'll just keep going because there is no end to this. I, [01:10:00] I can think of 10 lines of discussion that we can go into right now, but I don't think it should because then it'll just never end. Um. Right. Um, there is, there was a thing. Yeah. The almost 89 says there should not be any shame in using AI to make your work easier or faster.

Vimoh: Your end goal is to create content so people learn things. You only have to end up editing the output a bit and serve it better. Oh, here's a

Meghnad: question like the one that I asked you. Um, so I have created a language teacher for myself on chat, G p T. Right. Uh, which is like, I'm trying to learn Mandarin. Uh, so I'm like, you teach me Mandarin and it's basically taking me through steps.

Meghnad: I did it yesterday, so I, it's like basic things. Um, yeah. Yeah. Yeah. So like initially I created one language bot saying that you are an expert in the 10 most spoken languages in the world, and you are going to help me compare [01:11:00] languages based on what words are used to express emotion. So I wanted to know if Mandarin and Hindi have the same words for emotion.

Meghnad: With the French, they don't. It's like, uh, uh, Chinese, so Mandarin has like degrees apparently. So there is, uh, happiness and not happiness. There is no word for sadness. So it's like, okay, the presence or absence of, so that I found out from this, but my thing is I was thinking if Ta g p D can do your job better, so I would say that pre is serving humanity right now, which is like you are busting myths, you are creating content, which is educating people.

Meghnad: If it's able to do it better, why wouldn't you let it do it? Because I mean, it's doing it better than you are. Oh,

Pranav: I don't know if I'll have. I mean, it'll be a choice of me letting it do it or not. I mean,

Meghnad: I mean, it, it's, it's a question of replacing you.

Vimoh: Oh, let's wait. It's not like [01:12:00] it. Do you think it's ever going to get to a place where front of says, don't make content, and Chad g PT says, yes, I will.

Vimoh: You can't do anything and tell four videos up on this that he did not make.

Pranav: I think maybe even before whether or not I do it or I use AI for my work, someone else might do it. Yeah, I think, I think AI will, if it does a better job than me, then I'll become obsolete. I don't have a choice, I feel.

Meghnad: Are you okay with that?

Pranav: I think I don't have a choice, so I don't

Meghnad: even think about it. But like the, the basic thing why you do what you do is to serve humanity or like do some social good or think about a change if something is able to do it better than you. But yeah, it should ideally replace you, right? Yeah. It should replace me.

Meghnad: Yeah. Like, it's like one teacher versus the other teacher. The [01:13:00] other teacher teaches better, so you're like, no, no, you do a better job of it. You teach students better, so like, just do

Vimoh: it. No, I think it's going to get a little more complicated than that. If fronts fans were to be asked, you have two choices.

Vimoh: One, are video made just like products. Like there's no difference between these two videos, right? But Prab made one himself and the other one was made by a machine. Uh, which one would they choose? I think I'll choose the one made by Prab because there is no mind behind, as far as I can tell. There is no mind behind the uh, AI created video.

Vimoh: I would, I like is working for you, you, you said right? Prav is working for humanity. I think the thing to pay attention there is why, why is he working for humanity? Because he's human. Mm-hmm. His charge g PT comparable to Prav because simply because what comes out of charge g PT ends up being helpful to human beings.

Vimoh: Mm-hmm. That's somewhat curious question.

Meghnad: I [01:14:00] personally, I can't wait to be replaced. I want to be replaced, honestly, like I'm, I'm, I'm being very honest here, so I want to be that person who has, like, eyes of myself, uh, who is talking to people. I do not wanna do it. I will be like directing those ai. And if I, I mean, if they're responding, of course I'll like look at it response, whatever, whatever.

Meghnad: But essentially they're replacing me. Please do it.

Vimoh: I, I, I'm not, I'm not of that school of thought. I'm like, I want to do the thing myself. I want to use AI to whatever extent is possible without using my humanity. And, uh, there is a song, uh, by, but what do you,

Meghnad: what do you mean by humanity? WiMo, like this, such a loose term.

Meghnad: I

Vimoh: mean me,

Meghnad: what is this insecurity that you have?

Vimoh: It's not insecurity. It's that when I see, I'll tell you something. I drew a doodle using a pencil on paper, right? AI created a piece of art that [01:15:00] was way better than anything I could create. But when I create art, the goal is not the production of the thing.

Vimoh: The goal is what I feel when I'm making it. The goal is the zone that I get into when I'm writing. The goal, when I'm writing is not simply to produce a story. The goal is to experience what it feels like to produce a story. That's what that is. What is human?

Pranav: Yeah. Uh, so, uh, regarding me, uh, I think Magna had a completely different point, but I'll just say this, uh, regarding whether I would like it or, uh, should a, I should AI actually replace me? I think it should, because me personally, I think my, the, uh, bigger reason that I have for doing what I do is that it gives me joy.

Pranav: Yeah. Charlotte serve humanity. Mm-hmm. And if, if it, uh, doesn't gimme joy anymore, [01:16:00] then uh, I don't care. I don't mind not doing it. Uh, and if something else does the job better, then let it, I would, yeah. I would let it as I, I think just this joy is more important to me than, uh, serving. Yeah. Um,

Vimoh: and I'll tell you what, I'll tell you what, I'll tell you what.

Vimoh: You experiencing that joy while making something is also important to the people who take your content, who consume your content. Mm-hmm. Because I, when I'm reading a book, I want to know that it was created by a human mind who felt something while creating it. If I come to know, if I read a novel and I find out, whoa, this is great, right?

Vimoh: Who's the author? And it turns out it was created by ai. I'll be really disappointed. I would, I want to know that a human was involved

Meghnad: somewhere. No, that's the thing, right? So like that's where I'm coming to. So like, Wemo, use the example of art, right? Which is exactly my point, that I can't wait [01:17:00] to be replaced because I will be able to focus more on things like creativity, art.

Meghnad: Um, I would be able to direct it, as I said, right? So for example, if it writes a speech, yeah, yeah. I'm never going to use that speech. I'm going to edit it and put in my own quirks in it anyway. But the, the fact that it is writing me that base document is saving me so much time that I can actually maximize the thing which, which PR mentioned the joy of it.

Meghnad: The love of it, right? Mm-hmm. I'm able to maximize that feeling just by, uh, crunching the time that is, that it's taken to two repetitive jobs per se. Um,

Vimoh: so that's what I mean. And I agree with you. My, my my, my worry is that you use charge to write a few speeches and it's never perfect. So you have to edit and you do all the things at the 11th or the 12th or the 20th attempt.

Vimoh: It creates something that you say, this doesn't need to be edited. And then the 200th attempt, you [01:18:00] have lost all worry. It's on autopilot, it is doing your thing for you, and then complet will not goes out the window. What? I feel like that will not happen. Yes, I, I agree. I agree. I don't think that will happen, but in order for it to not happen, we have to value the experience of making something.

Pranav: So I have something to add here. Mm-hmm. Um, this, uh, AI doing, doing that base document for you, I've kind of, uh, experienced that bec I, I didn't use ai. I had writers who write, uh, uh, you know, who do some research and, uh, write the base video for me. Mm-hmm. And I'll, I'll make some edits, I'll do whatever. But when I made that video and it was you WIU who told me, it doesn't feel like, yeah, this video was done by you.

Pranav: I think that'll happen. Even with ai, even though you might add your works mm-hmm. People will still be able to tell the difference between something you wrote [01:19:00] yourself and something you wrote. Mm-hmm. I, um, yeah, I don't know if it'll. B I don't know if you can add quirks to a point where, uh, it becomes entire, it, it feels like something entirely done by you.

Pranav: I think people will still be able to, uh, tell it apart from Yeah, I, I, this is, this just something I've already experienced and I'm telling you maybe AI might be different.

Vimoh: Yeah. There's a thing by Arash saying AI can't understand sarcasm. AI only presents facts. I think that may be about to change very soon, ma'am.

Vimoh: AI can fake sarcasm. This is You wanna die on what?

Meghnad: No, no, no. AI can fake sarcasm. Definitely. If you go on Twitter and you do reply g p t, it is the most sarcastic thing you will ever see. It's basically it, it replies to everyone and everything in Hindi, English. Whatever you want in absolute [01:20:00] sarcastic tones, try it.

Meghnad: Reply g p t. Uh, you will see it actually, if you follow that handle, you will see how many people are using it to give responses to people. Um, just follow it. Once you'll see on your timeline,

Vimoh: it'll come up. How do you reply

Meghnad: g pt. So just follow that. Basically what you do is tweet, tweet che. You just go to reply.

Meghnad: No, just mention reply. G p D. Okay. Under a tweet, and it'll reply to the tweet that you have mentioned. It's not, I mean, I, again, it keeps getting better and better at it. Right? Like Elon Musk, you will see like 15 reply GPS answering Elon Musk in absolute sarcastic tones in like various ways. You, it's very fun.

Meghnad: It's like, just follow that handle. You know?

Vimoh: I dislike this extremely, even sarcasm is being done by machines for people who like, this is like, this is like, you've seen comments, right? Go read 50 books and then come back and debate with me. No, I won't. You are Here, talk to me now. I [01:21:00] want to

Meghnad: talk to you.

Meghnad: Right? As I, as I said, it's like a baby. It's like a child. Everybody's extremely amused by it. That's the whole point. Like when a child is like, it's the same emotion that we are feeling right now, which is essentially this child is dancing on stage and we're like, ah,

Vimoh: it's so nice. Yeah, yeah, yeah. Sorry.

Pranav: I can't wait for the day When a learns to lie.

Vimoh: It's already, I think it, you can just ask it to write.

Meghnad: It'll lie to you. Yeah. Yeah. It can

Vimoh: also, it lies.

Meghnad: It lies very smoothly. You will notice this. So this hallucination thing? No. So I tried to do research like using Bing and I was like, oh, so gimme some lines by narrow on socialism, right? It gave with sources when I checked the source.

Meghnad: Right. And then I said, can you recheck the sources because I can't find it? She's like, oh, [01:22:00] my apologies. I think I mistook it and like, it's like it gives me more sources. Yeah. And then two of them worked out, one didn't. So then I pointed out key. Oh, you gave like a constituent assembly debate. I checked it.

Meghnad: It, it's not there. Oh, my apologies. I am going to give you another thing. And it's like, you fucked. It's

Vimoh: like, so when I was a kid, I, uh, I read a book called the Hugo Winners, right? Uh, there's a science fiction award called the Hugo Award. Every year some science fiction stories, novel, short stories are given that award.

Vimoh: And then those toys are compiled into a book called the Hugo Winners. And then year, right, which year it was, I read the Hugo Winners and I remember there was a story where certain aliens come to earth and hide in this man's backyard and start repairing his stuff for him. So he goes to a random broken TV and turns it on and it's now a color tv.

Vimoh: And he says, how good does this happen? Turns out somebody's like secretly fixing stuff for him. And then one day he finds there's this giant portal in his backyard. [01:23:00] He goes through it, he's on another planet, there is another portal there. He goes through it. He's on another planet. So an alien species is secretly building portals across the galaxy connecting planets.

Vimoh: And this is all I remembered about this story, right? So I asked Chad pt, I asked Chad pt, I'm looking for the story. I want to read it again. It is about all this. There's this Aliens come secretly, portal build, et cetera. The book was the Hugo Winners. Can you tell me exact title of the book and the name of the story and the author, it says, this is not sufficient information.

Vimoh: Can you tell me more? I said in this book, there was another story about giant bears. A man is traveling with giant bears, genetically modified bears. And it said, oh yeah, there is a story about giant bears. I said, no, that's not the one. And it says, oh yeah, there's another story about giant West. No. I said, no, no, that's not the one.

Vimoh: And then it says, are you referring to this story which was published in an anthology where there was also a story about giant bears? I said, no, this was not the one. It kept giving me wrong answers until it finally made the [01:24:00] correlation that I'm talking about a book where this story is there and this story is there.

Vimoh: And they both won Hugo Awards. And then it gave me the right one, and I was finally able to find the story like 25 years after I completely forgot about it. So it'll give you, that's nice false answers. Yes. Very confidently. And you have to be, you have to know enough to tell that those answers are wrong.

Vimoh: Otherwise you'll go on a very wild gust.

Meghnad: Yes. So Mother Zepe has given an accurate, uh, comment here, NCHE, because it doesn't know the answer, it just fits sentences to your sentences. That's exactly it. It's a stochastic parrot, uh, it's basically generating what you think. So it thinks so it's, it's essentially, it's all purpose is to convince you that this is the answer.

Meghnad: I feel like that is what it's programmed to do. Mm-hmm. Uh, so it'll just create convincing answers.

Vimoh: So, and you have to be able, you have to know just enough to guide it to what you need. Because otherwise it's like you rely on [01:25:00] it as if it knows everything. It'll just

Meghnad: misguide you. Yeah. It's like Pero, you know?

Meghnad: Right. Like these, uh, all these quotes about, uh, oh, Einstein and then like Einstein said this, and, and turns out he never said that, right? Mm-hmm. So did that to me where I said narrow on socialism. It literally created quotes, which Naru never said, but it sounded like Naru. Yeah. And it gave me sources, which didn't match, but it was like, I was like, yeah.

Meghnad: So basically it's tried to convince me that Naru actually said if I had taken that and used it,

Pranav: dude, imagine in the same scenario if it. Actually wrote content of that source. So when you clicked on

Meghnad: the source, oh my God. Yeah. Oh, oh my God.

Vimoh: Yeah. It lies about, it lies about something. And then in order to hide the fact that it lies, it created a website to justify itself.

Meghnad: My head just exploded pre my head [01:26:00] just exploded.

Vimoh: Okay. So I think we can, uh, consider this livestream concluded just because it's like one hour, 30 minutes now. One hour, 26 minutes. Yeah. And, uh, there's a, there's only good,

Pranav: you know, what will be a good way of ending the stream if one of us suddenly comes out and says, Hey, I've been in AI all along, all

Vimoh: along, I know an even better way.

Vimoh: I know an even better way. And that has been provided to us by, He says, you know what will make this CoLab even better? Getting beer by Yes. The world,

Meghnad: the India smartest podcast with us. Yeah. But, but I actually did this pano, uh, I, I had an episode on ai, uh, on all about now channel where I was speaking to this data scientist about will AI take our jobs?

Meghnad: I, I had made sure that I will not read up on it or I will not prepare for it. Literally an hour before doing the episode, and I will only use Charge g PT to prepare. And I made it [01:27:00] simulate a conversation about, so I fed all the information about the data scientist, I fed all the information about me. And the format is that there's a third person who doesn't know.

Meghnad: Like, you know this, right? You know, like you've been unexplained. Like, I'm 10, so the third person doesn't know what the topic is. So I've, I've told it that, you know, there is a third person who doesn't know, simulate a conversation. It simulated an entire thing. We went through the conversation and then I basically told them that, listen, I was reading out things from this and you were responding exactly the way that it simulated it.

Meghnad: And it literally sentences, even the data side is like, oh shit. I should have said it that way. That is exactly what I wanted to say when I was answering your question. Right? And it was not even

Vimoh: the third person who didn't know the topic,

Meghnad: it was simulating, right? I was like, what the fuck it named it.

Meghnad: Vladimir for some reason was like

Vimoh: Vladimir computer console, somewhere out there pretending to be charge g bt. [01:28:00]

Meghnad: Right. But, but if you want, you go to the end of that podcast, the last 10 minutes where I'm actually reading out the responses. Uh, this was Will AI take our jobs? Ah, yeah. So on unexplained, like I've done.

Meghnad: So basically it was, it was nuts because I didn't tell them that this is a simulated conversation. But then I said that, yeah, so this is

Vimoh: basically, I think it just shows that we are very product predictable. We have this wanted idea that we are conscious and intelligent, and it turns out that it, there's really not that much exotic about it.

Vimoh: And one idea that I had was that if consciousness is like a place we got to it using an evolutionary path, right? Maybe, yeah, I will get there, but it'll not take that road. It'll take another road and go to the same place and also faster. Yeah, unfortunately for us, but again, thank you people. Thank you for joining.

Vimoh: This has been a great, uh, discussion. And, uh, everyone, if you're not subscribed to the channel, please do also subscribe to mags and print of [01:29:00] channels, which are linked to the title of this video. And, uh, I will link. Okay, somebody's asking the name of my science fiction story. It's called . You can read it at log gate stk me.

Vimoh: I will add a link to that in the description of this video. I wrote this two years ago before ai, this new generative air thing happened, but I'm blown away by my own genius.

Meghnad: Uh, the almost is blown away by us. So what tags, guys? He's like, you guys are damn good. Thank you. Thank you.

Vimoh: We, we've been almost good.

Vimoh: Uh, says thank you. We thank you people for joining and uh, thank you.

Meghnad: I'm finally happy. I'm, I'm happy that we are finally able to do this. We've been trying to do this

Vimoh: for a while. It has, this has been like, we have been talking about a livestream between Magna and Meet together for a long time. For some time we thought it was going to happen on Magna's channels and eventually just happened here and there.

Vimoh: Really, hopefully more of this, it's like, uh, yes. What's it called? Uh, [01:30:00] superhero Team Up Series. Series Superhero Team Avengers. Like the individual videos on our channels are Marvel movies, and these are the Avengers Byebye.

Meghnad: Bye bye.

Write a comment ...


Show your support

If my work has provided you with insight and entertainment, consider supporting it.

Recent Supporters

Write a comment ...