Compute access is Anthropic’s practical risk, while safety risks are existential

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 19:22

   Yeah, I just like that the S1, it looks like that matches up with what semi-analysis is going to say. Okay, one last thing here. It's in the risk factors. The risk factors are very interesting. Way up top is the compute part of it. If we are unable to secure adequate compute or if we fail to direct our compute effectively across our workloads, we may be unable to train models that remain competitive with those of our competitors or adequately serve our customers. You know, Dan, we've, I think over the past week or so, just seen these up and downs with the neoclouds and the market, you know, even questioning companies like Meta, it seems like we've been, you know.

2. 20:00

   eager to raise more money to build more compute. Just talk a little bit about the, does the AI game just boil down to like, do investors trust you enough to buy the compute you need to survive? I think that's a piece of it. I think that's a good way to sum it up. I have no concerns that Anthropix is going to be able to buy the compute, right? Because, you know, for starters, it's one of the biggest dogs on the block.

3. 20:23

   They are going to be at the front of the line. It's not going to be, you know, a new startup. It's going to be them, particularly if they go public and have all this extra cash on hand. But look, I mean, it's not the most interesting risk factor in there. But no, I think it's probably the most practical risk factor in there. Yeah, I mean, the most interesting risk factor is like our product models can cause you harm, which you just you kind of knew that was going to be in there.

4. 20:43

   Oh, it's beyond cause you harm. It's end humanity. Destroy the planet. It's an existential threat to humanity. I mean, which, by the way, I love when you read it, and I'm going to paraphrase this, but it's kind of like, you know, that's a concern internally and externally. We could end humanity as we know it, which could impact our financial performance. Yeah, no shit. That could impact your financial performance.

5. 21:02

   Yeah. Yeah, I'll read it. Concerns have been raised both within our organization and by other researchers, regulators, policymakers, and industry participants that sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy.

6. 21:27

   Dan, in the history of financial reporting, has there ever been a sentence like that in S1? No, because I don't think Oppenheimer went public. So no, I don't. No, no, no, nothing like this. And there was a lot, I will say, there's been a lot of commentary over the last, what it's been, two months since the Coxon thing came out about, well, you know, they're going to have to resolve this issue before they go. Not possible. Well, whether possible or not, clearly not. They're not even really trying to. And that's not even a knock on them. I don't think they can. But no, this is going to be in there. And again, whether you invest in the IPO or whether you are invested in the IPO or not is not going to impact or change the fact that Anthropic could not destroy the world, which will, again, render the rest of your portfolio meaningless. How does Anthropic's