AI R&D Is Approaching an Explosive Feedback Loop | Tom Davidson

There's two fields where this feedback loop is strong enough that you could get a kind of accelerating fume. RSI and it's AI hardware and AI software just happen to be the two fields where kind of the diminishing returns really aren't very strong. The scariest thing is that you know maybe even without the hardware

loop that's like by itself enough to explode. We are back. We are live with Tom Davidson who is a senior research fellow at Fore. We also just had Will McKascal from Forethought on like last week which was pretty sick. Um, forethought is I think my my single favorite AGI research org that is focused on postagi futures.

>> There's so few people doing this and Fore is really doing it quite well. I posted about this once and God forthought a bunch of followers on Twitter which they well deserve. Tom, welcome. >> Yeah, pleasure to be here. >> So, the big news of the day was PZ Research uh released this what was it

called? Uh, Taste Val. >> Yeah. >> Yeah. which yeah measures the experimental research taste of AI systems against human experts and they find that with their specific definition of research taste uh >> AI double at this every 3 months and have exceeded human baselines. >> Yeah. So >> and it's interesting because initially

it's not doubling very fast and then it's just like what is it since 2025? >> Is that since RL >> since GBT 5.2 since 4.5 and GBT 5.2 So I I I don't know if people have hypothesized why it's suddenly started going more more more steeply. I guess quite plausibly because they've started

actually doing these types of RR on these types of environments that are that are relevant for this task. >> I wonder if it has something to do with like coding automation over the same cuz like this end of 2025 moment was the exact same time that models started to get like really good at code.