Clive Chan

@elonmusk lfg looking forward to the race

all natural organic chip fabrication

A few people asked me to make an image of how much irrigated farmland would use the same water required for ALL ChatGPT usage, including every part of the process. I did a botec and my best guess right now is that inference uses about as much water as training, and power generation uses ~5x as much water as the data centers themselves, so it looks something like this. The water cost of manufacturing chips is marginal compared to how much water they use over their lifetimes.

@tszzl @jekbradbury with periodic appearances of chief villain @shamezhemani

Landman except it follows @jekbradbury as he wanders through the central plains of america looking for handfuls of GPU racks here and there on unnameable neo-clouds owned by frontiersmen to run a few more instances of Claude code

@S_OhEigeartaigh no, it’s precisely because he’s not quiet about it that he’s awesome!

Jeff Dean continues to quietly be one of the most decent people in AI.

Codex is so sweet 🥹

@tszzl #keep5p4

5.4 is my personal 4o honestly it just gets me

@IsaacKing314 The steelman (which never was seriously adopted internally) is that camera based cars already drive better than humans, so slowing down AV rollout costs lives. LIDAR does not save enough additional lives to justify slowing down AV rollout via higher costs / harder supply chain.

I'm generally hesitant to accuse successful people of being terrible at their own field, but I have long thought that Tesla's reasoning for this made very little sense. Yes it's true that humans manage to drive without lidar, but humans would drive more safely if they did have lidar! Even if Tesla had succeeded at human-level driving AI, there are fundamental limits on sensors that can't see in the dark. The obvious next step would be to make your cars even safer via superhuman perception, so why would you not just include those sensors from the beginning?

this is incredibly cool

FA4 is out!!!!

@EpochAIResearch woah, 25% vs 55% is not statistically significant?

On Tiers 1–3 GPT-5.4 Pro solved 52% of the non-held-out set and 42% of the held-out set. On Tier 4, GPT-5.4 Pro solved 25% of the non-held-out set and 55% of the held-out set. Neither of these differences is statistically significant.

GPT-5.4 set a new record on FrontierMath, our benchmark of extremely challenging math problems! We had pre-release access to evaluate the model. On Tiers 1–3, GPT-5.4 Pro scored 50%. On Tier 4 it scored 38%. See thread for commentary and additional experiments.

one more thing the good faith debate has been super helpful for me (thanks y’all), but i’ve learned that liveblogging my thoughts and getting undue attention neither makes positive change nor helps me form my own opinion. expect me to post less on this and chat more 1on1 cheers

GPT-5.4 is a big step up in computer use and economically valuable tasks (e.g., GDPval). We see no wall, and expect AI capabilities to continue to increase dramatically this year.

deleting my post now that dario’s retracted the internal statements in his letter 😃

thanks Dario 🙏 anthropic.com/news/where-sta…

@jeremyphoward @BlackHC * under my post

@jeremyphoward @BlackHC thought it was a good meme, but fair enough good to see his letter, lemme comment it under mine

@dlbydq @goog372121 i think the gov still has many tools effectively make ant suffer, and i’d rather that not happen. i think the admin is super happy to shoot down anything it sees as woke signaling / trying to override its power, but is amenable to being guided in ways that it can frame as winning

@itsclivetime @goog372121 ...in what way? it's still going to get thrown out the moment it hurts courts. I thought the point of doing anything prior to that point was to try and give RL signal to the administration on norms around rule of law

@ayedtay Damn what a banger

"with all this thinking the answer is gonna be amazing" the thinking:

@goog372121 This is weirdly public to have a conversation like this. Happy to DM too

@goog372121 I’ve genuinely had conversations about leaving but I’ve opted to wait for more information. And I’ve also been told that leaving is giving up and that staying helps influence events from the inside. What do you think?