Our Real Atlas Builds and Use Cases

Brian (0:00)

Hey, what's going on, everybody?

SPEAKER_2 (0:01)

Welcome to The Daily AI Show.

Brian (0:03)

We're live.

SPEAKER_2 (0:03)

It's 10 a.m. on the East Coast. It is September 8th, 2026, which doesn't mean anything to anybody else other than my buddy just turned 50 today.

Brian (0:13)

So happy birthday to my buddy Shane, who I've known since college. So 30 years now or something. So I hope everybody's having a great day with us today. We have Anne, Andy, Beth, and Brian. And yeah, we're with you for most of the time today, and he's got to duck out a little early, but you got us. We'll see if anybody else pops in the door. How are you guys doing? Yesterday, we had a fun show.

We really got into the idea that the paper on the alien brain that came out of OpenAI, and I feel like we actually spent some good time on it. In fact, I've gone even further and used that as the premise for the conundrum, because if you think about it, that paper really presents a easy conundrum, which is, okay, AI is happening. Let's just assume we can't pump the brakes.

AI is happening, right? And it's increasingly happening, as we know. It's speeding up. It's not just plotting along at some pace that we can plot out. It's exponentially increasing, so we know that, too. So the conundrum really is just the basis that comes out of that paper from Jacob, who wrote the paper, the chief AI scientist, which is just to say, okay, we're in a really uneasy place here, like we talked about on the show yesterday. Do we create more AI because we're saying, the only way us as humans can actually maybe track what frontier, pushing frontier model AI is doing is by having a better AI, what's the word for it? Investigator, it's not the word for it, but anyway, you know what I mean. Or do you say, well, that's crazy, because that's just a dog chasing its tail. You can never get ahead of that if you are constantly building new AI checkers for new AI models, at what point is the human even really in the loop anymore, other than just to say, I guess so, that looks right. I guess I believe you. I created this AI to go look at this AI and do, now do we need a third AI in a very simplified way of thinking about this to check the first checker?

SPEAKER_2 (2:21)

Where the hell does that stop?

Brian (2:22)

It's just like, it spirals out of control. It says one side is saying, like, why would we do that? The other side is saying, hey, if you are the belief that we can't slow down AI, meaning that, as I was saying in the comments yesterday, it's like the nuclear race. Yeah, yeah, yeah, let's all draw down, you first. That's right, it's the three Spider-Mans pointing at each other, as I was saying to Cisco yesterday in the comments. And so if you believe that that's the case, that I think unless there's some major event, some major break out of the box, get into some real serious places event, that would equally affect China as it does Russia, as it does the US, as it does Canada and all. And everybody worldwide said, whoa, whoa, whoa, whoa, we can't pause, pause, we can't have this because it had a worldwide impact that went across borders, geopolitical borders or country borders.

If that doesn't happen, then we're still on this ever, ever increasing pace roller coaster going on an ever steeper hill. So the conundrum is just based off that paper, which is to say like, well, which one? And both have very good arguments for them. Both of them have very good blockers to them. And why you would say no, that doesn't sound like the right move at all. But which one is it? So anyway, I just want to bring it up because that will actually be the conundrum episode. As a matter of fact, I even, you know, in Google Notebook LN, now Gemini Notebook, you can default for the host, the two AI co-hosts, right?

That do the actual conundrum part after I do my intro. Normally, I just set that as the default, whatever the normal length is. I don't do short, I just do the middle one. This actually one I haven't listened to yet. I actually defaulted it to long, which probably means it'll be closer, like maybe a 25-minute conversation if I had to guess, which I actually think is valid. When I do my intro and it goes out on Saturday, I'll actually say that, hey, I purposely made this a longer episode because I think this is the type of discussion that we really need to be digging into more. So anyway, that's coming up. I just wanted to let people know that that conundrum episode comes out on Saturday but is actually directly off the conversation we had on the show yesterday. So if you missed it, please go back and take a look. But what I was gonna say to you guys, and you know, Karl's with us now too, and Anne's hanging out with us today, we really didn't touch Astra a bit. Andy, you were saying at the time you still didn't have access. Has that changed at all?