Ryan Greenblatt

It seems that Anthropic might have their own take-off speeds, timelines model that differ substantially from current public modelling, produces much less conservative conclusions about the level of concern, and that they are using for decision-making.

If so, I think they should either publicly write up their modelling, informally would be fine, or get third parties to review it privately.

Insofar as they mean we think we'll maybe reach 2x overall progress when our survey.

That's mostly capturing vibes and doesn't have a clear correspondence to any particular notion of uplift.

Reaches 40x, fair enough, but it seems good to clarify this.

The current state of our evidence about AI R&D acceleration from Mythos seems extremely limited and AI companies should, and can, do much better going forward.

Heading.

Appendix.

Estimating AI progress speed up from serial labor acceleration.

There's a list of bullet points here.

Suppose we had a serial labor acceleration of X, as in employees go X times faster and also increased experiment compute by X. Then AI R&D progress would go X times faster.

I mean instantaneous progress, putting aside diminishing returns to research effort.

Equivalently, the research effort per unit time would go up by x. This is also putting aside parallel compute being worse than serially faster compute, though I think this doesn't make a huge difference in practice.

So, production is some function of serial labor acceleration and experiment compute.

We're uncertain about the exact function between something more like a CES model and a Cobb-Douglas production function.

I happen to think it's more like Cobb-Douglas than CES for reasons I discuss here.

I tend to think the functional form for just AI R&D progress, like algorithmic progress, is like serial underscore labor underscore acceleration caret 0.55 asterisk compute caret 0.45.

It might be pretty different as you start growing these values by orders of magnitude, especially if it's very CES-like, but at least if we're talking about less than 30x increases to these variables, I think it's something like this.

I'm uncertain about the exact constants.

Serial underscore labor underscore acceleration caret 0.7 asterisk compute caret 0.3 and serial underscore labor underscore acceleration caret 0.3 asterisk compute caret 0.7 are somewhat plausible and make a pretty big difference to the bottom line.