@ARoniou
Joined January 2023
  • Tweets237
  • Following263
  • Followers6
1Media
AI becoming more capable and gaining control ≠ AI killing humans
1
2
25
There's only 2 paths that could cause large number of human deaths. 1. Bioweapons / 2. Nuclear weapons. But there is an enormous chain of assumptions for any of them to cause any real havoc. And AI is dangerous ≠ AI kills millions ≠ Human extinction
1
2
26
Safety is good and important and some alignment risks are real. Besides safety is tight to product reliability. But this madness is gone out of control. People went full schizoid lately.
4
Google is done and dusted in the LLM race. They are still well positioned in the long term overall AI race but not in the LLM race
2
24
Definition of Intelligence = Generality x Efficiency x Proximity to Truth. Alphago = Low intelligence (low generality) Calculator = zero intelligence (zero generality) Humans = highest generality & truth proximity to any other animal. Current AI (2026) = subhuman in all 3
1
41
Gemini gets alot of hate for benchmaxxing but Google only needs to focus on one test, "Remote Labour Index". If they just focus on that they'll improve user productivity and hence satisfaction.
29
Write AI Benchmarks that models score below 20% (as of Dec 2025).
10
Load more