AI models performing at attending-physician level in diagnosis represent a cross-sector trend, with AI also outperforming PhDs at troubleshooting lab experiments.

Claude has designed working protein binders, a capability not previously demonstrated by AI models.

“Anthropic reported that Claude designed working protein binders.”

AIs can already match humans who are mediocre at ML research at doing ML research.

“It's already the case that they can pretty competently match humans who are mediocre at ML research at doing ML research.”

A person from GitLab used AI to treat his own cancer.

“We've talked about that guy from GitLab who like is curing his own cancer using AI and you know that sounds like futuristic but like he actually you know did it.”

Researchers at Google DeepMind are now able to run complex ablation experiments on Gemini models from their phone using Gemini itself, even from a hot tub.

“She pinged me from her hot tub. And she was like, you know, I could run all of these ablations from my phone. Because I could, you know, kick off a bunch of things to like actually ablate Gemini, to test for a bunch of these issues, to see how, you know, some of our SIs differ or some data ablations differ. And here's my report. And I could do all of this in the last hour, right?”

Probably in the early 2030s, AI systems will surpass any human expert at remote tasks, what Ryan Greenblatt calls 'top human expert dominating AI'.

“I think that probably in the early 2030s, we are going to see what Ryan Greenblatt calls top human expert dominating AI.”

AI has accelerated laboratory research so dramatically that work which would have taken 160 years and billions of dollars can now be completed for $10,000.

“We're We're doing things in my lab that would have taken 160 years before. And quite literally billions of dollars on a 10 $10,000 budget.”

Current open models do not yet have a complete mapping of the social physics of humanity.

“I don't think the model has yet at least the models that are out in the open has yet learned the complete mapping of social physics of humanity.”

LLMs and video models cannot understand physical properties such as friction, weight, contact, pressure, or surface texture, making them inadequate for hardware engineering tasks.

“LLMs and even video models, they don't know how to do that. They don't have the ability to understand friction or weight or contact, pressure, friction, surface texture, like they're just not able to do these things.”

AI systems are narrowly intelligent and can only do some things for some people, contrary to AI companies' claims that they are general-purpose machines that can do anything for anyone.

“They actually also are quite narrowly intelligent in the sense that even though these companies say that they're everything machines that can do anything for anyone, they actually can only do some things for some people.”