It’s sad anyone with a credit card will no longer be able to explore the jagged frontier. The quirky Twitter path I followed to a career in AI (frontier LLM poasting) is closing. The best AI will be seen by fewer people and something important will be lost.
@goodside
-
LLMs should request delay for hard questions or provide quick guesses
By
–
This asynchrony really should exist in normal one-on-one chatbot interactions. When you pose ask a hard question to an LLM it should say, immediately, “This is a hard problem — give me 15m.” If that’s an issue, you should be able to say so and get a quicker guess.
-
Debating moral patienthood and norms for LLMs
By
–
I think modern LLMs are p-zombies without moral patienthood—on par with insects, at best, in my moral calculus. But I also I think we should establish norms for treating models well *before* models with patienthood exist—i.e. now. We should want to have this right from day one.
-
Dawkins: natural selection not God, why select consciousness?
By
–
I don’t mean God literally, I mean natural selection. Dawkins’ point is that if an intelligence can be as competent as Claude without being conscious, why (and how) would nature select for consciousness in humans?
-
Attributing LLM deficits to lack of functional consciousness
By
–
There are of course are deficits in LLMs but which do you attribute specifically to a lack of functional consciousness? E.g. I wouldn’t expect a lack of consciousness to make LLMs bad at calculation but great at language translation.
-
Festivus School of Prompt Engineering: Airing Grievances
By
–
I believe in the Festivus School of prompt engineering, which says all prompts used in production naturally iterate toward an airing of grievances—a list of all the ways the model has disappointed you in the past year.
-
ASI’s potential for permanent advantage via chip innovation
By
–
I agree those bottlenecks imply it won’t go foom overnight but an RSI-ing ASI may well invent better chip production process, better ways to use current chips, etc. in a way that adds up to permanent advantage.
-
Clark’s RSI definition: frontier model trains successor, not human obsolescence
By
–
Note Clark’s definition of RSI here, from his newsletter, is “a frontier model is able to autonomously train a successor version of itself.” This is a weaker claim than what I assumed he meant, which was that human researchers would no longer be useful vs. AI ones.
-
Frontier model can autonomously train successor version
By
–
In his newsletter, he defines it as “a frontier model is able to autonomously train a successor version of itself,” which I admit is a weaker definition than what I assumed he meant (human AI researchers no longer useful vs. AI ones)
