ai
klasyczny Dwarkesh: tj. czy jak damy modelowi całą wiedzę o fizyce do Einsteina, ale ani kroku dalej, czy wymyśli teorię względności?
To create an Einstein in a data center, we don't just need a system that knows all the answers, but rather one that can ask questions nobody else has thought of or dared to ask. One that writes 'What if everyone is wrong about this?' when all textbooks, experts, and common knowledge suggest otherwise.
Thomas Wolf • 🔭 The Einstein AI model
czy ja potrafię tak zadawać pytania? raczej nie. albo momentami, w pojedynczych sytuacjach, jak np. u filomatów na interview
Ilya Sutskever discussed the boundary between understanding and awareness, as what comes next, in his latest test of time talk at NeurIPS 2024. To understand is to predict things accurately. To be self-aware is to be able to predict accurately with an understanding of what it is and what its environment is.
Nathan Lambert • The rise of reasoning machines
czym jest świadomość?
This perspective misses the most crucial aspect of science: the skill to ask the right questions and to challenge even what one has learned. A real science breakthrough is Copernicus proposing, against all the knowledge of his days -in ML terms we would say “despite all his training dataset”-, that the earth may orbit the sun rather than the other... See more
Thomas Wolf • 🔭 The Einstein AI model
istota nauki
1*bvMhd_xpVxfJYoKXYp5hug.png (2208×1244)
miro.medium.com
Yann LeCun’s cake on the RL, Supervised & Self-Supervised Learning
This all goes back to Ilya’s provocation for why next-token prediction is enough on the Dwarkesh Podcast:
Predicting the next token well means that you understand the underlying reality that led to the creation of that token.
The rise of reasoning machines
In that compute-constrained world, deployment and development trade off directly. That starts as a lab-internal effect, where deployment and product improvements eat into (compute) resources otherwise earmarked for development.
hard cap on what can be done & recursive self improvement not necessarily happening this decade
I believe an Einstein-level breakthrough in Go would involve something more fundamental, like inventing the rules of Go itself—exploring the vast landscape of possible game rules to create something genuinely more compelling than existing games.
Thomas Wolf • 🔭 The Einstein AI model
think: the glass bead game, inveting the rules of the game of the human culture itself; move 37 in go is not an Einstein level breakthrough