Jascha Sohl-Dickstein
jascha.sohldickstein.com
Jascha Sohl-Dickstein
@jascha.sohldickstein.com
Recently a principal scientist at Google DeepMind. Joining Anthropic. Most (in)famous for inventing diffusion models. AI + physics + neuroscience + dynamical systems.
There are similarly loads of ways that the system dynamics resulting from billions of interacting, individually pro-social, AIs can go wrong and weird.
January 4, 2025 at 10:09 PM
It helps if participants in government (/corporations/economies/...) are good faith. But there are loads of ways that well-intentioned smart people can achieve terrible group outcomes.
January 4, 2025 at 10:09 PM
Gradient estimators, not gradient resonators
February 18, 2024 at 7:45 PM
version of the meta-loss. the current best version of the algorithm: openreview.net/forum?id=Vhb...
February 16, 2024 at 1:17 AM
We would avoid it if we could! The challenge is that when you descend the meta-loss by gradient decent, you converge into the perilous region. It's hard to know how to exclude it. Our best approaches so far use stochastic finite difference gradient resonators (variants of ES) to descend a smoothed
February 16, 2024 at 1:15 AM
The boundary between trainable and untrainable configurations is *fractal*! (and beautiful!)

For details, and more pretty videos, see:

blog post: sohl-dickstein.github.io/2024/02/12/f...

paper: arxiv.org/abs/2402.06184
Neural network training makes beautiful fractals
This blog is intended to be a place to share ideas and results that are too weird, incomplete, or off-topic to turn into an academic paper, but that I think may be important. Let me know what you thin...
sohl-dickstein.github.io
February 12, 2024 at 5:39 AM
All the appointments are filled. Will see how the meetings go, and evaluate doing this again. I'm looking forward to finding out what people are interested in!
January 10, 2024 at 1:48 AM
I created 11 meeting slots for this first round. I'll reply to this message when/if they've all filled up.

calendar.app.google/ZdnWzeYw3qyR...
January 10, 2024 at 1:07 AM
Paper here: arxiv.org/abs/2311.02462

(Imagen is a Level 3 "Expert" Narrow AI)
November 7, 2023 at 4:04 AM
These Levels of AGI provide a rough framework to quantify the performance, generality, and autonomy of AGI models and their precursors. We hope they help compare models, assess risk, and measure progress along the path to AGI.

(AlphaGo is a Level 4 "Virtuouso" Narrow AI)
November 7, 2023 at 4:03 AM
AI can enable awesome (as in inspiring of awe) good in the world. We have amazing leverage as the people building it. We should use that leverage carefully.
September 11, 2023 at 3:28 AM
A theme is that we should worry about a *diversity* of risks. If we recognize e.g. only specific present harms, or only AGI misalignment risk, we will find our efforts overwhelmed by other types of AI-enabled disruption, and we won't be able to fix the problem we care about.
September 11, 2023 at 3:28 AM
My top fears include targeted manipulation of humans, autonomous weapons, massive job loss, AI-enabled surveillance and subjugation, widespread failure of societal mechanisms, extreme concentration of power, and loss of human control.
September 11, 2023 at 3:27 AM