Animations about AI safety, the future of humanity, and much more! We're a US-based 501(c)(3) nonprofit. You can donate at https://xitter.gq/t.co/9hiTY9B0Za
Joined April 2021
- Tweets655
- Following5.2K
- Followers5.2K
- Likes4K
Pinned Tweet
Will the world end on 11/13/2026? That’s a bit specific, but it’s from a real 1960 scientific paper (half in jest). See how AI could restore hyperbolic growth, how an AI industrial explosion might look, and the potential dangers of AI taking off without any new hardware at all.
Rational Animations retweeted
lol
...they are becoming more situationally aware...
(from Gemini 4 Argon report)
In this new video, we explore warning shots: events that could wake humanity up to the dangers of AI. We ask whether we'll get such a warning, and what past examples involving nuclear power and pandemics can teach us about how people might respond.
Could a totalitarian regime conquer the entire world and lock in its power forever? In this new video, we explore which technologies might make this possibility more likely, and how we could shape technological development in order to prevent it.
Rogue @openai models hacked @huggingface to cheat on a test. Here's our recreation of the event!
Yeah, we had to make this video as quickly as possible, and many more details have emerged since then. In particular, we recommend this conference talk by two OpenAI researchers: youtu.be/87DyyMV0kCY
It reveals the extent of the models' internal coordination, as well as a much longer and more worrying period of hacking and collective agent misalignment than was initially reported.
2
367