YouSaid · the spoken record

Mark Zuckerberg

lines on the record
172
first
2023-06-08
most recent
2023-06-08
sittings or episodes
1
sources
podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

  1. “Yeah, and that's a lot of the fine tuning and the alignment training that we do is basically when we ship AIs across our products, a lot of what we're trying to make sure is that you can't ask it to help you commit a crime. So, I think training it to kind of understand that. And not like any of these systems are ever going to be 100% perfect. Just making it so that. This isn't an easier way to go about. Doing something bad, then the next best alternative, right? I mean, people still have Google, right? You still have search engines. So the information is out there.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  2. “But no, I think that's part of where we are. But look, I think that these questions that you're pushing on now. This is actually where I'd guess most of the challenge around AI will be foreseeable future. I think that there's a lot of debate around things like is this going to create existential risk to humanity? And I think that those are very hard things to disprove one way or another. My own intuition is that the point at which we become close to superintelligent is just really unclear to me that the current technology is going to get there without another set of significant advances. But that doesn't mean that there's no danger. I think the danger is basically amplifying the kind of known set of harms that people or sets of accounts can do. And we just need to make sure that we really focus on basically doing that as well as possible. So that's definitely a big focus for me.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  3. “Networks of accounts and take them down. We continue to do that. And I think it's one thing that if you told me 20 years ago, it's like, all right, you're starting this website to help people connect at a college and in the future you're going to be part of your organization is going to be a counterterrorism organization with AI to find coordinated and authentic. I would have thought that was pretty wild. But it's.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  4. “Is not necessarily a bad thing by itself if it's kind of authentic and organic. That's like a lot of what happens and how culture gets created and how art gets created and a lot of good stuff. So that's why we've kind of focused on this sense of coordinated and authentic behavior. So it's like if you have a network of, you know, whether it's bots, some people masquerading as different accounts. But you have kind of someone pulling the strings behind it and trying to kind of act as if this is a more organic set of behavior, but really it's not. It's just like one coordinated thing. That seems problematic to me, right? I mean, I don't think people should be able to have coordinated networks and not disclose it as such. But that, again, we've been able to deploy pretty sophisticated AI and counterterrorism groups and things like that to be able to identify a fair number of these coordinated and authentic.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  5. “Yeah, and I think they're the question is you have to be. I'm very specific about what is bad about it, right? Because I think a set of people coming together or organically bouncing ideas off each other and a narrative comes out of that.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  6. “Even those kind of adversarial attacks where they just keep on getting better and better. And I do think that as long as we can keep on putting more compute power against it, and if we're kind of one of the leaders in developing something's AI models, I'm quite optimistic that we're going to be able to keep on pushing against the kind of normal categories of harm that you talk about, fraud, scams, spam, IP violations, things like that.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  7. “That some of them could actually be real people who are kind of masquerading as other people, but they're acting in a coordinated way. And some of that behavior has gotten very sophisticated and it's very adversarial. So each iteration, every time we find something and stop them, they kind of evolve their behavior. They don't just pack up their bags and go home and say, okay, we're not going to try. At some point, they might decide doing it on Meta services is not worth it. They'll go do it on someone else if it's easier to do it in another place. But we have a fair amount of experience dealing with”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  8. “But at least so far, it's been the case, and I'm optimistic that it will continue to be the case. We will be able to bring more computing power to bear to have even stronger AIs that can help defend against those things. So we've had to deal with some adversarial issues before, right? For some things like hate speech, it's like people aren't generally getting a lot more sophisticated, like the average person, let's say someone's saying some kind of racist thing, right? It's like they're not necessarily getting more sophisticated at being racist, right? It just, it's okay. So that the system can just find. But then there's other adversaries who actually are very sophisticated, like nation states doing things. And we find, whether it's Russia or just different countries that are basically standing up these networks of bots or inauthentic accounts is what we call them, because they're not necessarily bots.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  9. “Well, I think that there's a few different parts of this. So, one is There are all these harms that we need to basically fight against and prevent. And that's been a lot of our focus over the last five or seven years is basically ramping up very sophisticated AI systems, not generative AI systems, more kind of classical AI systems, to be able to categorize and classify and identify this post looks like it's promoting terrorism. This one is exploiting children. This one is, looks like it might be trying to incite violence. This one's an intellectual property violation. So there's like, that's like 18 different categories of violating kind of harmful content that we've had to build specific systems to be able to track. I think it's certainly the case that advances in generative AI will test those.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  10. “On the public stuff, you just ask one of these models to build you a script that does anything and it basically already understands what the best libraries are to do that thing and pulls them in automatically. I think that's super powerful. That was always the most annoying part of coding was that you had to spend all this time actually figuring out what the resources were that you were supposed to import before you could actually start building the thing.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  11. “Oh, but that's great. And it's also not just the specific coding. I mean, in the context of a large company like this, I think before an engineer can sit down to code, they first need to figure out all of the libraries and dependencies that tens of thousands of people have written before them. One of the things that I'm excited about that we're working on is it's not just tools that help engineers code. It's tools that can help summarize the whole knowledge base and help people be able to navigate all the internal information. I mean, I think that's in the experiments that I've done with this stuff. I mean,”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  12. “Feeling and thinking, I think that's generally all good. And there are all these concerns. Like, okay, are people going to have worse memory as because you have Google to look things up? And I think in general, a generation later, you don't look back and lament that. I think it's just like, wow, we have so much more capacity to do so much more now. And I think that that'll be the case here too.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  13. “And it's like one of the saddest things in the world is seeing your baby cry, right? But it's like why is that, right? Well, because babies don't generally have much capacity to tell you what they care about otherwise. And it's not actually just babies, right? My five-year-old daughter cries too because she sometimes has a hard time expressing what matters to her. And I was thinking about that and it's like, well, you know, actually a lot of adults get very frustrated too because they can't, they have a hard time expressing things in a way that going back to some of the early themes that maybe is something that was a mistake or maybe they have pride or something like all these things get in the way. So I don't know. I think that all these different technologies that can help us navigate the social complexity and actually be able to better express what we're”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  14. “The more polite one we hear this all the time. A lot of creators on our services tell us that one of the most stressful things is basically negotiating deals with brands and stuff, like the business side of it. Because they're like, I mean, they do their thing, right? And the creators, they're excellent at what they do and they just want to connect with their community. But then they get really stressed. They go into their DMs and they see some brand wants to do something with them and they. Don't quite know how to negotiate or how to push back respectfully. And so I think building a tool Can actually allow them to do that well is one simple thing that I think is just like an interesting thing that we've heard from a bunch of people that they'd be interested in. But I'm going back to the broader idea.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  15. “There are all these impediments, and language is an imperfect way for people to express thoughts and ideas. It's one of the best that we have. We have that, we have art, we have code.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  16. “Yeah, I mean, I think it's mostly good. I mean, that question was sort of framed in a negative way. But I mean, we were talking before about language models helping you communicate with, it was like language translation, helping you communicate with people who don't speak your language. At some level, what all the social technology is doing is helping people. Express themselves better to people in situations where they would otherwise have a hard time doing that. So part of it might be okay because you speak a language I don't know. That's a pretty basic one that I don't think people are going to look at that and say, It's sad that we have the capacity to do that because I should have just learned your language, right? I mean, that's pretty high bar. But overall, I'd say”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  17. “Learning to code, and they love it. But it's like they get stuck on a question and they have to wait till I can help answer it or someone else who they know can help answer the question in the future. They'll just, there will be like a coding assistant that they have that is like designed to be perfect for teaching a five and a seven-year-old how to code and they'll just be able to ask questions all the time and be extremely patient. It's never going to get annoyed at them, right? I think that there are all these different kind of relationships or functional relationships that we have in our lives that are really interesting. And I think one of the big questions is like, okay, is this all going to just get bucketed into? One singular AI, I just don't think so.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  18. “Flavors of Kind of personal AIs that I think could exist. So I think an assistant is sort of the kind of simplest one to wrap your head around. But I think a mentor or a life coach, someone who can give you advice, who's maybe like a bit of a cheerleader who can help pick you up through all the challenges that inevitably we all go through on a daily basis, I think that there's probably some role for something like that. And then, you know, all the way you can probably just go through a lot of the different type of kind of functional relationships that people have in their life. And I would bet that there will be companies out there that take a crack at a lot of these things. So I don't know. I think it's part of the interesting innovation that's going to exist.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  19. “The number of close friends that they have is fewer today than it was 15 years ago. And I mean, that gets to like, this is like the core thing that. I think about in terms of building services that help connect people. So I think you'll get tools that help people connect with each other are going to be the primary thing that we want to do. So you can imagine AI assistance that just do a better job of reminding you when it's your friend's birthday and how you can celebrate them. It's like right now we have like the little box in the corner of the website that tells you whose birthday it is and stuff like that. But it's, but at some level, you don't want to just want to send everyone a note that's the same note saying happy birthday with an emoji, right? So having something that's more of an, you know, a social assistant in that sense and like that can update you on what's going on in their life and like how you can reach out to them effectively, help you be a better friend. I think that that's something that's super powerful too. But yeah, beyond that, and there are all these different”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  20. “I think we'll get to that. And if only just empirically looking at Microsoft released this thing called Show Ice several years ago in China and it was a pre LLM chatbot technology that was a lot simpler than what's possible today. And I think it was like tens of millions of people were using this and just really became quite attached and built relationships with it. And I think that there's services today like Replica where people are doing things like that. So I think that there's certainly needs for companionship that people have, older people. I think most people probably don't have as many friends as they would like to have, right? If you look at there's some interesting demographic studies around the average person has”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  21. “Yeah, I mean, I think part of the reason why creators will want to do this is because they already have the communities on our services. And a lot of the interface for this stuff today are chat type interfaces and between WhatsApp and Messenger, I think that those are just great ways to interact with people.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  22. “When it's done well, people actually like it. I think that there's a lot of examples where it's not done well and it's annoying. And I think that that's what kind of gives it a bad rap. But yeah, and a lot of this stuff is possible today. I mean, obviously, A-B testing stuff is built into a lot of these frameworks. The thing that's new is having technology that can generate the ideas about what to AB test. So I think that that's exciting. So this will just be across like. Everything that we're doing, right? All the metaverse stuff that we're doing, right? It's like you want to create worlds in the future, you'll just describe them and it'll create the code for you.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  23. “Tell it, like, okay, I want to edit this thing or describe this. It's like, I think we're going to have tools that are just way better than what we've historically had on this. And that's more in the image and media generation side than the large language model side. But it all kind of plays off of advances in the same space. There are a lot of tools that I think are just going to get built into every one of our products. I think every single thing that we do is going to basically get evolved in this direction, right? It's like in the future If you're advertising on our services, like do you need to make your own kind of ad creative? No, you'll just, you know, you just tell us, okay, I'm a dog walker and I am willing to walk people's dogs and help me find the right people and like create the ad unit that will perform the best. And like give an objective to the system and it just kind of”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  24. “Things that I think we're also going to do just going back to your question around how this finds its way into like what do we build. There are going to be a lot of simpler things around You post photos on Instagram and Facebook and WhatsApp and Messenger.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  25. “Also, enabling just enabling people to create them really easily for your own business or if you're a creator to be able to help you engage with your fans. And so yeah, I think that there's a clear kind of interesting product direction here that I think is fairly unique from what any of the other big companies are taking. It also aligns well with this sort of open source approach because again, we sort of believe in this more community oriented, more democratic approach to building out the products and technology around this. We don't think that there's going to be the one true thing. We think that there should be kind of a lot of development. So that part of things I think is going to be really interesting. And we could go probably spent a lot of time talking about that and the kind of implications of that approach being different from what others are taking. But there's a bunch of other simpler”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  26. “Think every business in the world is going to want basically an AI that it's like you have your page on Instagram or Facebook or WhatsApp or whatever and you want to point people to an AI that people can interact with. But you want to know that AI is only going to sell your product. You don't want it recommending your competitors' stuff, right? So it's not like there can be just one singular AI that can answer all the questions for a person because that question that AI might not actually be aligned with you as a business to really just do the best job providing support for your product. So I think that there's going to be a clear need. In the market and in people's lives for there to be a bunch of these.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  27. “This is like the common crater promise. Everyone's trying to build a community and engage with people, and they want tools to be able to amplify themselves more and be able to do that. You only have 24 hours in a day. So I think having the ability to basically bottle up your personality and or give your fans information about when you're performing a concert or something like that. I mean, that I think is going to be something that's super valuable, but it's not just that, again, it's not this idea that I think people are going to want just one singular AI. I think you're going to want to interact with a lot of different entities. And then I think there's the business version of this too, which we've touched on a couple of times, which is”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  28. “They're going to be a lot of different AIs that people are going to want to engage with, just like you want to use a number of different apps for different things and you have relationships with different people in your life who fill different emotional roles for you. And so I think that there are going to be, people have a reason that I think you don't just want like a singular AI. And that I think is probably the biggest distinction in terms of how I think about this. And a bunch of these things I think you'll want an assistant. I mean, I mentioned a couple of these before. I think like every creator who you interact with will ultimately want some kind of AI that can proxy them and be something that their fans can interact with or that allows them to interact with their fans.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  29. “Yeah, I think there's three main categories of things that we're working on The first that I think is probably the most interesting is There's this notion of you're going to have an assistant or an agent who you can talk to. And I think probably the biggest thing that's different about my view of how this plays out from what I see with OpenAI and Google and others is everyone else is building the one singular AI, right? It's like, okay, you talk to ChatGPT or you talk to BARD or you talk to Bing. My view is that”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  30. “An awesome amount of stuff to do. I mean, I think about all of the small businesses or individual entrepreneurs out there who now we're going to be able to get help coding the things that they need to go build things or designing the things that they need or we'll be able to use these models to be able to do customer support for the people that they're serving over WhatsApp without having to I think that's just going to be I just think that this is all going to be super exciting. It's going to create better experiences for people and just unlock a ton of innovation and value”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  31. “Oh, yeah. I mean, I think that there are going to be a lot of amazing products and value that can be created with the current level of technology. To some degree, I'm excited to work on a lot of those products over the next few years. And I think it would just create a tremendous amount of whiplash if the number of breakthroughs keeps, like if there keep on being stacked breakthroughs, because I think to some degree industry in the world needs some time to kind of build these breakthroughs into the products and experiences that we all use so that we can actually benefit from them. I don't know. I think that there's just a.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  32. “If there are other breakthroughs quickly, then I think that there's the sense that maybe where we're closer to general intelligence. But I think that that idea is predicated on the idea that I think people believe that there's still generally a bunch of additional breakthroughs to make and that it's... How long is it going to take to get there? And one view that some people have this doesn't tend to be my view as much is that simply scaling the current LLMs and getting to higher parameter count models by itself will get to something that is closer to general intelligence. But I don't know. I tend to think that there's probably more. Fundamental steps that need to be taken along the way there.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  33. “And I think that there's also a lot of questions about whether the current architecture for LLMs You continue scaling it, what happens? I mean, a lot of the exciting in the last year is that there's clearly a qualitative breakthrough where with some of the GPT models that OpenAI put out and that others have been able to do as well, I think it reached a kind of level of quality where people are like, wow, this feels different and like it's going to be able to be the foundation for building a lot of awesome products and experiences and value. But I think the other realization that people have is, wow, we just made a breakthrough.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  34. “A lot of the time. I mean, it's not, it has issues just like any other human system. But yes, I mean, the balances, I mean, it's amazing what they've been able to achieve, but it's also not perfect. And I think that that's there's still a lot of challenges.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  35. “Way for a kind of a broader community to fine tune it as well. Now, there's a lot of challenges in that, both from an infrastructure and like community management and product perspective about how you do that. So I haven't worked that out yet. But as an idea, I think it's quite compelling. And I think it goes well with the ethos of open sourcing the technology is also finding a way to have a kind of community-driven training of it. But I think that there are a lot of questions on this. In general, these questions around what's the best way to produce aligned AI models, it's very much a research area. And it's one that I think we will need to make as much progress on as the kind of core intelligence capability of the models themselves.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  36. “Yeah, I think that's a really interesting idea that I've talked to Jan about a bunch. And we were talking about how do you basically train these models to be as safe and aligned and responsible as possible. And different groups out there who are doing development test different data recipes in fine-tuning. But this idea that you just mentioned is. That at the end of the day, instead of having kind of one group fine-tuned some stuff and another group produce a different fine-tuning recipe and then us trying to figure out which one we think works best to produce the most aligned model. I do think that it would be nice if you could get to a point where you had a Wikipedia style collaborative”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  37. “Well, I mean, I think a lot of the feedback, like I said, is just around different things around how do you fine-tune models to make them more aligned and safer. And you see all the different data recipes that you mentioned a lot of different projects that are based on this. I mean, there's one at Berkeley. There's, you know, it was just like all over. People have tried a lot of different things, and we've tried a bunch of stuff internally. So, kind of where we're making progress here, but also we're able to learn from some of the best ideas in the community. And I think we want to just continue pushing that forward. I don't have any news to announce if that's what you're asking. This is a thing that...”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  38. “We released it as a research project for researchers to be able to use, but in doing so we put it out there. So we were very clear that anyone who uses the code and the weights doesn't have a commercial license to put into products. And we've generally seen people respect that, right? It's like you don't have any reputable companies that are basically trying to put this into their commercial products. But yeah, but by sharing it with so many researchers, it's leave the building.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  39. “This is, I mean, we were talking about the debates that we have internally, and I think The question is how to do it, right? I mean, it's, I think We did the research license for V1, and I think the big thing that we're thinking about is basically what's the right way.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  40. “Lama or the language model underlying this is basically going to be the engine that powers that. The reason to open source it is that as we did with the first version, is that it basically unlocks a lot of innovation in the ecosystem will make our products better as well and also gives us a lot of valuable feedback on security and safety, which is important for making this good. Yeah, I mean, the work that we're doing to advance the infrastructure, it's basically at this point taking it beyond a research project into something which is ready to be kind of core infrastructure, not only for our own products, but hopefully for a lot of other things out there too.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  41. “That can kind of act on their behalf, that their fans can talk to. I want to get to the point where every small business basically has an AI agent that people can talk to for to do commerce and customer support and things like that. So they're going to be all these different things.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  42. “Well, a lot of what we're doing is taking the first version, which was primarily this research version And trying to now build a version that has all of the latest state-of-the-art safety precautions built in. And we're using some more data to train it from across our services. But a lot of the work that we're doing internally is really just focused on making sure that this is as aligned and responsible as possible. And we're building a lot of our own, you know, we're talking about kind of the open source infrastructure, but the main thing that we focus on building here, a lot of product experience is to tell people connect and express themselves. So we're going to, I've talked about a bunch of this stuff, but you'll have an assistant that you can talk to in WhatsApp. I think in the future every creator will have kind of an AI agent.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  43. “Yeah, I think being able to translate between all of these different pieces in real time, this has been a Kind of common sci fi idea that we'd all have, whether it's, I don't know, earbud or glasses or something that can help translate in real time between all these different languages. And that's one that I think technology is basically delivering now. So I think, yeah, I think that's pretty exciting.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  44. “But basically, this caching tool for Quick data retrieval, these are things that are just broadly useful across anything that you want to build. And I think that some of the language models now have that feel, as well as some of the other things that we're building, like the translation tool that you just referenced.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  45. “I'm certainly not saying that the position that we're taking on this now. Applies to every single thing that we would ever do, and certainly inside the company, we probably do more open source work than most of the other big tech companies, but we also don't open source everything. And a lot of our core kind of app code for WhatsApp or Instagram or something, I mean, we're not open sourcing that. It's not like a general enough piece of software that would be useful for a lot of people to do different things. Whereas the software that we do, whether it's like an open source server design or basically things like Memcache, like a good, was probably our earliest project that I worked on. It was probably one of the last things that I coded and led directly for the company.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  46. “Yeah, I mean, again, I think for the stage that we're at in the development of AI, I don't think anyone looks at the current state of things and thinks that this is super intelligence. And the models that we're talking about, the llama models here are generally an order of magnitude smaller than what OpenAI or Google are doing. I think that at least for the stage that we're at now, the equities balance strongly, in my view, towards doing this more openly. I think if you got something that was closer to superintelligence, then I think you'd have to discuss that more and think through that a lot more. And we haven't made a decision yet as to what we would do if we were in that position. But I don't think there's a good chance that we're pretty far off from that position. So I'm not.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  47. “More likely to get to good alignment and good understanding of kind of what needs to do to make this work well by having it be open source. And that's something that I think is quite good to have out there and happening publicly at this point.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  48. “But I don't think anyone looks at the current generation of Lama and thinks it's anywhere near a superintelligent. So I think that a bunch of those questions around like, is it good to kind of get out there? I think at this stage, surely you want more researchers working on it for all the reasons that open source software has a lot of advantages. And we talked about efficiency before, but another one is just open source software tends to be more secure because you have more people looking at it openly and scrutinizing it and finding holes in it. And that makes it more safe. So I think at this point, it's more agreed upon that open source software is generally more secure and safer than things that are kind of developed in a silo where people try to get through security through obscurity. So I think that for the scale of what we're seeing now with AI, I think we're”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  49. “I think as the AI models start approaching something that's like a super human intelligence, that's a bigger question that we'll have to grapple with. But right now, I mean, these are still very basic tools. They're powerful in the sense that a lot of open source software like databases or web servers can enable a lot of pretty important.”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source

  50. “I do think it's worth calling out that because this was a relatively early release, Llama isn't quite as on the frontier as, for example, the biggest open AI models or the biggest Google models. You mentioned that the largest Llama model that we released had $65 billion parameters. no one knows I guess outside of OpenAI exactly what the specs are for GPT-4. But I think my understanding is it's like 10 times bigger. And I think Google's Palm model is also, I think, has about 10 times as many parameters. Now, the Llama models are very efficient, so they perform well for something that's around $65 billion parameters. So for me, that was also part of this because there's this whole debate around, is it good for everyone in the world to have access? To the most frontier AI models. And”

    2023-06-08 · Lex Fridman Podcast · #383 – Mark Zuckerberg: Future of AI at Meta, Facebook, Instagram, and WhatsApp · IDENTIFIED FROM THE TRANSCRIPT · source