The people who worry about fast takeoff, RSI, AI doom, etc. forget that at inference time LLMs are just ordinary programs – there’s no learning, rewards, or self-anything, let alone desires, intentions, suffering or consciousness.
@pmddomingos
-
Critique of rushed AI doom legislation by uninformed lawmakers
By
–
There’s nothing worse than legislation made in a hurry about something the legislators don’t understand and that has never happened (e.g., AI doom).
-
Skepticism on singularity claims from OpenAI and Anthropic
By
–
At this point AI progress could stop dead in its tracks and the folks at OpenAI and Anthropic would still believe the singularity is here.
-
Chain-of-thought that doesn’t reflect internal reasoning is hallucination
By
–
Hey OpenAI, chains of thought that don't reflect the reasoning going on internally are hallucinations, not deceptions.
-
Critique of AGI imminence arguments as dumb extrapolation
By
–
All arguments that AGI is imminent boil down to “AI has solved a lot of problems recently, so it will solve all the remaining ones at the same speed.” This kind of dumb extrapolation is particularly hilarious coming from people who are supposed to know machine learning.
-
Debate over claims that frontier labs achieved AGI internally
By
–
“The frontier labs have already achieved AGI internally, they just haven’t released it yet.”
Yeah, that’s what OpenAI said in 2023. -
AI automating internship grunt work to shorten them
By
–
Internships etc. are 99% grunge work and 1% learning. AI doing away with the former just means they can be much shorter.
-
AI exploits have come from model makers, not models
By
–
Funny how with all those powerful AI models on the loose for months all the exploits so far have come from the model makers.
-
Why I use Gemini instead of ChatGPT or Claude
By
–
Why I use Gemini instead of ChatGPT or Claude:
– Much of what I do is interactive search, and it's the best one for that.
– It integrates better with everything else I already have in Google.
– It's not the leader in some areas, but that changes by the day.
– I don't trust OpenAI