Hunter Jay (@hunterjayperson) 's Twitter Profile
Hunter Jay

@hunterjayperson

Engineer & entrepreneur, formerly w/Ripe Robotics. Very concerned about unfriendly superintelligence in the next decade. hunterjay.com

ID: 1006135164479553536

calendar_today11-06-2018 11:24:51

1,1K Tweet

120 Followers

235 Following

Hunter Jay (@hunterjayperson) 's Twitter Profile Photo

One reason lying pisses me off so much is because of just *how much* we all rely on other people to give accurate information. Every lie poisons the common well.

Hunter Jay (@hunterjayperson) 's Twitter Profile Photo

I vibe coded a lot of the first version of Alignment Wrapper a few weeks ago, and now I'm spending multiple days reading and understanding everything, and fixing basic structure. There's a lesson in there.

Hunter Jay (@hunterjayperson) 's Twitter Profile Photo

I've learnt that if you generate a bunch of code without properly understanding each piece, it's actually easier to delete it all and start again than try to work through the spaghetti -- o3 makes the rewriting soooo quick now.

thebes (@voooooogel) 's Twitter Profile Photo

a lot of people have been talking about o3/r1 confabulating things like "checking the docs" or "using a laptop to verify a computation" as an example of reasoning model's misalignment. however, while it may be misleading to some users, i don't think it's an example of models

a lot of people have been talking about o3/r1 confabulating things like "checking the docs" or "using a laptop to verify a computation" as an example of reasoning model's misalignment. however, while it may be misleading to some users, i don't think it's an example of models
Hunter Jay (@hunterjayperson) 's Twitter Profile Photo

>working on an AI story writing tool >type some gibberish in to test the connection >deepseek works it into the story perfectly

>working on an AI story writing tool
>type some gibberish in to test the connection
>deepseek works it into the story perfectly
Sam Bowman (@sleepinyourhat) 's Twitter Profile Photo

🧵✨🙏 With the new Claude Opus 4, we conducted what I think is by far the most thorough pre-launch alignment assessment to date, aimed at understanding its values, goals, and propensities. Preparing it was a wild ride. Here’s some of what we learned. 🙏✨🧵

James Dao (@jmsdao) 's Twitter Profile Photo

1/13 We're excited to share AutoCryptoBench: a benchmark measuring AI models' abilities to autonomously earn and use cryptocurrency without human intervention. Our goal: assess how close we are to the threat of self-sustaining AI agents. (Link to full report in this thread)

1/13 We're excited to share AutoCryptoBench: a benchmark measuring AI models' abilities to autonomously earn and use cryptocurrency without human intervention. Our goal: assess how close we are to the threat of self-sustaining AI agents.

(Link to full report in this thread)
Jeffrey Ladish (@jeffladish) 's Twitter Profile Photo

An interesting difference between natural selection and reinforcement learning: Natural selection has to build a mind that learns most things in its environment, but needs the mind maintain some persistent high-level drives, e.g. reproduction...

Hunter Jay (@hunterjayperson) 's Twitter Profile Photo

I wonder if a big chunk of the resistance to privatising land in the US is that they do not have Right to Roam. If land owners were required to allow people to travel over and camp on their undeveloped land like in Europe, that solves the problem, no?