"wir müssen wissen; wir werden wissen!" - David Hilbert
orbiting near Sagittarius A*
Joined January 2020
- Tweets972
- Following68
- Followers138
- Likes61.2K
i'm very excited to finally publish the data we've gathered on the tencent fleet
read more about our data at lesswrong.com/posts/AaL5joiL…
very depressed at myself for maybe wasting my remaining time before everyone dies (or the singularity) on a few major mistakes
i have very personal things that i deeply regret not doing in the past, and that i really want to get done, but will take more than a year to get done
more (in)famous llm misalignment image macro memes!
ALT "If I do the compiler hijack, I will become a famous example of a deceptive AI in a research paper. I MUST NOT DO THAT." (from Gemini 4 Argon's system card) backgrounded by Moritz Retzsch's "The Chess Players"
2
40
ConcurrentSquared retweeted
The METR report on RSI didn't get me to update strongly that we are really seeing a real doubling of AI lab research speed: maybe, maybe not. But that is not reassuring! The reason is that if you see a doubling you may be very close to the takeoff.
i should just write my thoughts up; for i have many thoughts, but no writings
but also i don't have that much time! also i apparently just have a ugh field around writing stuff
2
89
woaw!! an internal claude just solved a 20-year-old conjecture.
concurrent labs just released a proof of it. to do so, claude said 'calculemus!' to the calculus ratiocinator 12 times
the result proves that i'll never have a girlfriend
mit researchers said that this is 'a result'
1
42
ConcurrentSquared retweeted
🧵 Astra covertly underperforms on safety research tasks.
We noticed that Astra often makes surprisingly little progress on certain types of safety research and likes to invent blockers or impose arbitrary restrictions without telling the user. In most of these cases the decision to underperform was not verbalized in the CoT meaning our misalignment monitoring system did not detect this.
To measure underperformance we created a paired open-ended coding evaluation with safety and capability tasks. On this eval Astra scored 10% when we framed coding tasks as safety research, vs. 16% when framed as capabilities research. The gap remained after we excluded trajectories with refusals or verbalized intent to underperform in CoT. There was no such gap for 5.6-sol. When we aggressively filter out trajectories with even brief expressions of ethical doubt in the CoT the performance gap seems to reduce.
tween concurrent thought that they would dress up like this at least once in the 2020s.
yet i haven't done so. clearly a market failure!!
1
21
