Enjoying this issue?
Get tomorrow's AI & engineering digest in your inbox — hand-picked, summarized, and always spam-free.
TLDR
OpenAI, Anthropic, and Elon Musk are advocating to 'pace the frontier' of AI development, which the speaker argues is a pretext to regulate and centralize AI, threatening local AI and open-source models. The speaker contends that these companies have repeatedly claimed AGI is near without demonstrating it, and that the real motive is control and regulatory capture.
Key points
OpenAI and Anthropic are advocating to 'pace the frontier' of AI development.
The speaker claims that these companies have repeatedly claimed AGI is near but have not demonstrated it.
The speaker argues that the push for regulation is a response to competition from open-source models like Qwen 3.8 27B.
The speaker warns that future regulation could criminalize running advanced local AI models.
The speaker notes that Chinese AI development has accelerated despite chip export restrictions.
Tools mentioned
Techniques
- airgapping
- distillation
Stop scrolling. Start reading smarter.
Receive the day's most important AI & engineering updates in one concise email. No spam.
Transcript (captions)
We need to pace the frontier is what we have Daario and Sam Alman out there saying. And they're proposing a new way to do this, which is incredibly creative and probably impact you. If you are
watching this channel for probably the reason you're watching this channel, you like local AI. So, if you have not been able to catch up and you've enjoyed your weekend, that's because you haven't been
paying attention to the ridiculousness that's going on in the anthropic and open AI camps again. And this time they've been joined by Elon Musk. So they want to pace the frontier. This is
very very likely in response to incredible showings recently from open-source models as far as their capabilities. Serious competition. But it's way deeper than that. There's
there's a line that's been crossed when you say something over and over and it doesn't manifest and you hope that that person on the other end doesn't notice. And this is a
game of confidence that people play. I I've seen a lot of other creators play it. I've seen a lot of other people professionally play this. They're going to say stuff that's leading out in
front. They're going to sell products that don't quite exist 100%, but they know that their engineers can get them there. If you work in tech, if you've worked for small companies or even
medium-sized companies, I'm sure you've seen this. There's a a bunch of actual commitments that get made around that. It's a way of operating that is very common, very very common. And we see
definitely statements around things like AGI. When was the first time you heard somebody say, "We're on the cusp of AGI?" Well, it's been years here for me,
several years. And even we're at AGI years ago, like two years ago. We're we're clearly not at AGI. We clearly were not at AGI then. There there's a huge problem here. Where is any of these
people that have said these statements walking back these statements ever? Never happens, right? Can you can you point that out? Please point that out in the comments below. Also, make sure you
hit like and subscribe. Thank you to all of our channel members. Also, where is the humility? And we've hit AGI being said so many times to you, to me, to everybody.
And yet, we're clearly not at AGI today, right now, this moment. Show me it. Challenge to you, anybody, I any CEO out there, show us AGI right now, today. Put your model where your
mouth is. We have AGI. Show us AGI. You don't have to give us AGI. Show us AGI right now. A lot of money riding on the line. Economies riding on the line even in some cases like the US economy.
China's economy not necessarily doing that hot either. Kind of riding on the line. If you start thinking about this, it starts to make a lot of sense just from
a pragmatic realist viewpoint. I'm kind of a little bit of a pragmatic realist myself. And what's there on the other side of it? Not, hey, by the way, we don't have AGI. Not that. Instead,
we have to pace the frontier. So, Open AI anthropic talking about pacing the frontier. We have to pace the frontier because it's becoming clear there's going to be an incident. And
boy, you're going to look back at it and think, "Hey, we should have paced the frontier earlier." I mean, you've got Rune I don't know if you're an ex person or not, but Runes out there saying stuff
like that on X. It's like, what? So, you see this and you got to think to yourself, what is going on here, right? This this just doesn't make sense. But maybe it really does because
I have a I have a feeling that there's only so far we're ever going to get with just straight up LLMs. And we've gotten a lot of additional mileage with aentic harnesses being
really capable, really helping push things, but yet it's still not AGI. Certainly not super intelligence. Are we getting closer to RSI? Well,
maybe. We have people like OpenAI saying, "Hey, by the way, maybe we don't want to IPO. Maybe it's just not the right time to do an IPO. We're on the cusp. Might just be more advantageous
for us as a business to kind of set over here on this side and just run it as a private business." Like, okay, we know you guys are working on stuff that's ahead of what you're giving your
customers. You're saying you need to paste the frontier. So like describe for your customers. I'm not a customer. I use this local models, local rig. Go back in the
channel history. You can find out how to build all this stuff. I show you how to run the software, how to put it together, how to keep it maintained just recently, how to make things happen with
it, how to have a good time, fun. It's a guy in his garage building garage AGI. But, you know, at the end of the day, that was a joke. And I can humbly admit Garage AGI is probably not happening.
But, but like, what does this mean for your customers? Like, if you're an OpenAI customer and you're expecting that next model release, if you're an anthropic customer
and you're expecting the next claude, like, do they not get it? Do they not get access to it? Are you pacing what you give them? I mean, I'm not even going to get into the fact that just
like this last week, we had Sabine talking about, "Hey, by the way, people are trying to essentially pay for influence. By the way, not sponsored. Nobody pays me.
Not sponsored. Nobody pays me. Like, you watch. I've got channel members. You guys pay me. Thank you for all the buy me coffees. Thanks. Thanks for the Patreons. Thanks for the channel
members. Thanks for everybody that likes, shares this out there. I think this is an important message also because them pacing is something that they want to have basically control.
You think that control stops in the United States? No. No. It's really good competition. They're facing the capabilities of Quinn 3.827B
at FB16. It is something that anybody can do and it's great. It gives you a lot of additional capabilities. A real competitor don't have to have a subscription. Are they are they are they
threatened by the Chinese? No, they're saying we need to slow it down because the Chinese are distilling and catching up really close to the edge. Okay, I can I'm not disputing that there
is not distilling happening and that the Chinese are not distilling and that that does not give them a significant amount of competitive like it's a huge catch-up. We also though have seen them
develop math and release things that have pushed forward open-source AI that really has moved things along and I think we will continue to see this. There's a necessity that the Chinese
have being chip constrained that we now are already see seeing like manifest. I mean Nvidia wasn't able to sell to them. Chip exports backfired huge and they've learned to build in rapid time very good
technologies to compete with it. Robotics the robotics is possibly even ahead of where we're at right now. Robotics might be the first application that your
average normal person gets in their household. Possibly more important than most other things. Picking bad battles, listening to bad advice, and doing the wrong thing
at the wrong time has huge ramifications. Just like if pausing the frontier, by the way, we also need to pause any open source models. You get AGI in your
garage, buddy. 20 years in the clink. That's the policies they want. That's the control they want. That's exactly the kind of dangerous thought to you that they want to have
happen. Yeah. Because that works out for them very well. That maintains control for a little bit longer. That centralizes control. Who gets to see it? We're going to have verifiers.
We're going to have verifiers. Do you feel like it's going too fast? Let me just ask that question. I don't. And if we look at the instances where we've got escapes, certainly if it was
not supposed to escape, you could airgap it. This is not a foreign concept in tech. You can airgap something. It like unless it's using some sort of soundwaves
radio. Maybe there's lasers that it could use. is also for some intererometry. But like seriously, you're going to have to get out there a little bit to escape from an air gap. I
shouldn't be this worked up on a on a Sunday. It should be all happy space. Instead, I'm all worked up space. Every every month, two months, 3 months, 3 months max, we're going to try again to
regulate local AI and all AI so we control it and centralize it in the United States. every two one to three months we're going to try another time to get this to happen. Sam Alman Dario,
the leaders of the frontier making decisions for you. You can have access, you can't have access and a lot of people would say you don't have to worry about it. You got Quinn 3.827B,
not a Frontier model, not going to be regulated. That's not the point here. We're not at a point where it's dangerous unless instructed or given the opportunity
maybe recklessly to be dangerous like humans involved very much there. My machine isn't going to turn itself on and be like, "Hey, by the way, coming for you tomorrow." It's just not going
to do it. Turn this machine on, sat down at the terminal. number of times that it's done something without a cron job or me asking it to just on its own. It took the initiative
and decided I'm going to actually send the human a message without a cron job telling me to do it without him prompting me somehow to do it zero because that's the reality. Show us the
agi. It's actually like a killer demonstration. They should totally be jumping all over demonstrating it. People would buy the Everliving bleep out of that. They would be buying the
credits and tokens like crazy. Exactly what they want, right? Unless they're just not there, right? It's a good time to talk about this with other people. But, you know,
the real reality is it's always a good time to talk about this with other people. The real reality of this is what local does. I call it free cuz like that that's very addressable to people. This
is like free AI. Like not to you, but like normal people. Not to say you're not normal, but you know. You know what I mean? People that are not into AI at all. It's free. Paid
subscription free. You should talk to them. Find out what they think about it. Most of them are not interested at all. Most of them have already had some bad
experiences and they're like, "That stuff's stupid." Really? Like they're going to need to be shown AGI or a robot. That's why robotics is probably really the big thing that breaks
through. I like I like optimism. I like seeing things cool that people make. I like seeing new technology that makes things better for humans.
But that could all come to an end. Instead, you could be looking at 20 years for having your own local AI if it crosses some arbitrary threshold yet to be named. Do
They pace the frontier here. Like, do you think the Chinese are going to say, "Yes, of course. Let's pace the frontier." Mathematicians, stop coming up with new
formulas. Stop trying to come up with new techniques. Stop it. Do Do you think this is going to happen? No. The real reality is that's not going
to happen in the United States either. That's not going to happen in anthropic and open AI. They're going to continue progressing privately. So maybe it could just be this is a big
moment where they're like, hey, you know what? We need to actually kind of not spend the compute resources on the customers as much as we need to
spend the comput resources internally. I don't know. This seems like I mean there there's a point in time coming here where data centers get built. This is like 2027. Data centers are built.
They have to move into data centers. Big bucks got to change hands soon, right? Like seriously, where's the AGI? Because boy, people will be people will be there
for that. People will be there for that. But you know what? what we see right now. If you're a customer, you I not a customer, but if you are a
customer, you got to start thinking to yourself like, man, what does this mean for me? Are you going to give me the next model? It's a really good question. I mean,
that's a really, really good question, right? I know people that spend a lot of money daily. Like, a lot of money daily. Well, I think it's ridiculous. I think it should not be every one to three
months that we have to have some sort of a moment like this. But definitely talk to your friends, talk to those that will listen. Not a bad time to make calls. Again, your representatives
are going to make it or break it for whether this is required. And there is bigger push. always bigger push because regulatory capture paching at least for a little while and then
it's going to have to be outlawed. So the best open source wouldn't be inside the United States. It's going to be hard to enforce, but it potentially could be something they try to enforce. You going
to comply with that? Sound off in the comments below and let me know. While you're down there, hit like, subscribe. If you're interested in getting up and running with your own local AI, you can
check out one of the two playlists that we got here, software and hardware. And have a great Sunday.