Meta's Vision for AI: Personal Assistants and Open Models

what we're doing with all of this intelligence uh he also says here's what Meta is building to sort of be part of this AI future that includes a personal AI agent for everyone uh powerful coding video design and more tools for everyone, ways to launch businesses without teams or funding, a PH level tutor or coach for any subject, contributions to science and so forth. He also discusses some of the risks, job loss, data center pro push back, cyber security and bioweapons, uh the US not keeping pace with China and he makes a few policy commitments. Jason Meta will continue releasing open- source models. They today dropped the openweight model Muse Glimmer and they're making plans for an openweight version of Muse Spark 1.2. He also proposes independent boards to review model safety criteria and he calls for closer cooperation between labs and the government that doesn't slow down the release of new models. So that's kind of a quick overview. So what what do you think? I mean, I think most people is this PR, is this advertising for the next wave of meta AI? Or do you take this as a sincere thing that Zuck is worried about the future and wanted to be part of the conversation? >> There's a lot going on here. I think he's learned a lot being u a technologist who moved fast and broke things and got himself into hot water. And one of the things he learned is he's historically a terrible communicator. he's unlikable uh because he's rich. He maybe, you know, I don't want to diagnose anybody, but maybe doesn't make eye contact or doesn't speak in a way that is engaging to people. >> And so, I think he understands those faults. And then also, I think because he was such a young founder, he was getting spun by different people telling him how to run his business and how to communicate. So, you take those two combinations together, not being a great natural communicator, uh, but great at building a business, obviously, and then people spinning you around that I think he's finally saying, I'm going to control the message here. I'm going to state my actual beliefs in a way like Elon does and let the chips fall where they may. I'm going to go direct. Right? And so, now we're seeing him get vocal and go direct because he just paid this huge fine, right? and the entirety of the second movie that's coming out about him. >> Yes. >> You put all that together as just a little prologue to this. I think he now wants to communicate directly >> as a man, as a founder, you know, founder authority. I'm going to control the message. I'm going to actually think from first principles what I believe and I'm going to state it and I'm going to be more thoughtful and preempt things that typically I would just build products, not explain myself and then when they break things um I'll just pretend they didn't that didn't happen and not say anything and then get pulled in front of Congress. Right. In corporate communications I think he's made a lot of mistakes over the years. I think he would agree with that too. So here we have a manifesto of AI and he touches everything and I thought it was actually very well written and very thoughtful but he covers everything. This is not one position. This is literally like a dozen. So you have to break down each one. So I think one place uh to start here is he's for open models and he's not for regulating open models. uh he believes that the best way to deal with these frontier models with dangerous things is not to lock them up or regulate them. It's to give everybody access to them. Sounds familiar. It's literally the manifesto of Open AI. >> Open AI's manifesto was this technology is too powerful for one person to control it. So, he's taking that position. That's also extremely convenient, isn't it? Since he's behind, when you're behind, you go open. When you're ahead, you go closed. Yeah. The Facebook network, Instagram, those are black boxes. You can't index them. There's no API for them. There's no grabbing the social graph and doing interesting things with it as the developer. They dabbled in that for a couple years and shut it all down. So, when you're ahead and you've got the franchise, you lock it down. iPhone as an example. When you're behind, Android, as an example, you open up. Okay? So, he's for opening. And that's a a really good strategic move for him. So then he talks about the dangers and there's a part where he talks about uh chemicals and that kind of stuff and I think I screenshotted it or you can just do a quick control find when he talks about materials and there's a section where he talks about materials and the dangers of them. So essentially he says listen trying to regulate the models is not going to be as effectively as trying to limit access to chemicals and precursors. So we should do that. In other words, how how you make a fertilizer bomb, god forbid, like Oklahoma City, we are going to have a hard time making an open-source model not be able to do that. We'll have an easier time going after the actual sale of fertilizer. He doesn't, I think, say it that crisply, but let's hear it. >> Yeah. He says, uh, labs should continue working to reduce biological and chemical risks in our models. Uh, first we should focus on limiting the physical production and distribution of harmful materials. I expect it will be easier to regulate and control physical components than the spread of knowledge. So this is an important area of policy focus. >> So this is really important. This is the first time I've heard this idea. >> I think it'll be the first time you've heard it. We are bringing up the dangers of AI. >> Yes. >> What does Daario say when he talks about the dangers of AI? Oh my god, this stuff is really dangerous. We had to stop it. We can't release it. He's getting involved in drama with the uh administration. What does Zuck say? He actually thought this through. You're you're not going to stop information. Information wants to be free. And AI makes it impossible to hold back this information. You're not going to stop access to open source models coming out of China or other countries or America or people making their own. Now, cats out of the bag. You can make your own model. So if we are concerned about biological weapons, the materials and the tools to make those, he didn't mention tools, but I'll add that are where we should try to limit things. If we are actually concerned about that, that's actually a really good policy statement. I haven't heard from another person. >> I feel like I feel like we already do. I mean, unless people are building bombs or whatever out of everyday common household items, we do try to already make like you I I couldn't go buy like a bunch of different hazardous chemicals today, at least not without some sort of a license or a permit. >> You would be surprised. You can buy a lot of the precursors. No founder I've