I ๐จโ๐ฌ a math. definition&theory of Artificial Super-Intelligence ๐ฅ&๐ค@ ๐
I now work @GoogleDeepMind ๐ง History:๐ฉ๐ช๐จ๐ญ๐ฆ๐บ๐ฌ๐ง
"Algorithmic Compression via Pretrained Neural Networks" is a short recap of the ~15 publications in the last 5 years of the Universal Artificial Intelligence team. https://t.co/btne1QpUhR
As a good mathematician the first book I read on AI was @mhutter42's book on AIXI. Then once I knew what real AI was, I read all the neural network stuff. I think this is also important conceptual kit for AI safety https://t.co/knbVbe7v51.
A little paper by Kamal Dingle and myself using Kolmogorov Complexity for measuring "Simplicity and Complexity in Combinatorial Optimization" published today https://t.co/tZRpDHy33Z
Introducing Q Labs, a research lab focused on solving generalization.
Alongside others (SSI, Flapping Airplanes), we see data efficiency as the key problem, but we're taking an unconventional approach to solve it: a new learning algorithm approximating Solomonoff induction.
Interview mal auf Deutsch auf @everlast_ai รผber SuperIntelligenz und die Zukunft der Menschheit https://t.co/WbO8e7QDfr
Anmerkung: AGI hat zu viele Vรคter, aber Vater der ASI Theorie gibts nur einen๐
#Intelligenz#Bewusstsein#Mathematik#Singularitรคt#Arbeit#UBI#AIXI #OffSwitchGame #X_Risk #Alignment #Zukunft (auto-translate available)
TIL that George Hotz's @realGeorgeHotz career was kindled by working on the Human Knowledge Compression Contest ("Hutter Prize") for 6 months of his life. He also understands why lossless compression โ lossy compression โ intelligence. https://t.co/9Adoq267Ms
@JasonRute@ElliotGlazer@davidmbudden I hope a positive outcome of this saga will be that the LEAN community puts more emphasis on formalizing important open problems, conjectures, definitions, and theorems (as opposed to formalizing proofs of existing theorems) AND improves LEAN so as to allow non-domain experts and AIs to fully automatically and failure-safe to verify their LEAN proof: attempts.https://t.co/OJIwOd8BOR
@IsaacKing314@littmath@RadishHarmers I can confirm that David has paid off his 10k betting loss with me. Thanks for being a good sport, but I will (re)pay him 20k if he wins the Clay prize for his NS proof on github. I don't want to have won for technical reasons.
@Trader_Shiba Yes, I advised him quite a bit, which he essentially ignored. I assign only a very small chance. Manifold odds feel roughly right. He has no clue about the physics nor the math. It's all PingYou generated including most of his X-posts. Good luck.
Has anyone here considered whether David Budden's X account might actually have been taken over by his PingYou bot, and the marketing is that it took so long for you to notice? This seems more plausible to me than a psychosis. I suspected this already weeks ago. Another bet?
200 pages Grain of Truth: A new gold standard for embedded multi-agent learning with infinite-order theory of mind leading to new game-theoretic solution concepts and novel forms of cooperation: https://t.co/NX1TjMyWpq
Mathematical Science of AI: Interview of Marcus Hutter and Cole Wyeth by Daniel Murfet at the Mathematical Research Institute of the University of Sydney.
https://t.co/kyRs3ZpDBi
New paper with Marcus Hutter at AI Magazine! "Imitation Learning is Probably Existentially Safe"
We contest 6 arguments to the contrary from Eliezer, Paul Christiano, David Krueger, Gwern, and Evan Hubinger et al.
We've also tried to make those arguments more accessible.
Congratulations to Yoshua Bengio for being the first living person (afaict, and the 2nd person in history after Michel Foucault) to pass one-million citations on Google Scholar. Geoffrey Hinton will follow soon.
More detailed/technical Grain of Truth slides: https://t.co/9xRP4gI8DW
Recorded talk: https://t.co/7gD8uM8OBG
Other exciting talks from the recent AIT&ML Symposium: https://t.co/atO54pg0Te
Reflective-Oracle AIXI solves the Grain of Truth problem for super-intelligent multi-agent systems/societies. Finally the long-awaited more comprehensive treatment building upon earlier work from last decade is out. Slides: https://t.co/Aixsac9rlO Paper: https://t.co/D0CAo30kc6