Harrison Ritz
@hritz.bsky.social
cybernetic cognitive control 🤖
computational cognitive neuroscience 🧠
postdoc princeton neuro 🍕
he/him 🇨🇦 harrisonritz.github.io
computational cognitive neuroscience 🧠
postdoc princeton neuro 🍕
he/him 🇨🇦 harrisonritz.github.io
Ya I think I’m taking it back — it’s one thing to take your action with some noise, but spending the time to learn a chaotic flow field just to explore doesn’t really make sense.
IIUC, you don’t add a bit of chaos to your policy (like noise in random exploration), you’re learning a chaotic policy
IIUC, you don’t add a bit of chaos to your policy (like noise in random exploration), you’re learning a chaotic policy
sense.like
November 7, 2025 at 3:54 AM
Ya I think I’m taking it back — it’s one thing to take your action with some noise, but spending the time to learn a chaotic flow field just to explore doesn’t really make sense.
IIUC, you don’t add a bit of chaos to your policy (like noise in random exploration), you’re learning a chaotic policy
IIUC, you don’t add a bit of chaos to your policy (like noise in random exploration), you’re learning a chaotic policy
Weird to think about this part of the architecture being learned…
November 7, 2025 at 3:09 AM
Weird to think about this part of the architecture being learned…
Ya a big part of their prestige comes from the postdoc/phds they attract, who won’t get a job in this budget
I think that hiring a couple ERCs from each CERC-candidate lab (+ top Canadians) would be better than hiring the whales themselves, certainly as a long-term strategy
I think that hiring a couple ERCs from each CERC-candidate lab (+ top Canadians) would be better than hiring the whales themselves, certainly as a long-term strategy
November 7, 2025 at 2:59 AM
Ya a big part of their prestige comes from the postdoc/phds they attract, who won’t get a job in this budget
I think that hiring a couple ERCs from each CERC-candidate lab (+ top Canadians) would be better than hiring the whales themselves, certainly as a long-term strategy
I think that hiring a couple ERCs from each CERC-candidate lab (+ top Canadians) would be better than hiring the whales themselves, certainly as a long-term strategy
Cool! Do you think it also helps with exploration? Add a little weirdness that pushes you to try new states
November 7, 2025 at 2:41 AM
Cool! Do you think it also helps with exploration? Add a little weirdness that pushes you to try new states
Also don’t like ‘intelligence’, but one argument I’ve heard is that everything is correlated for some reason.
Idk how we’re going to recognize super intelligence by ‘gud PC1’, but the positive manifold is kind of wild
Idk how we’re going to recognize super intelligence by ‘gud PC1’, but the positive manifold is kind of wild
November 7, 2025 at 2:26 AM
Also don’t like ‘intelligence’, but one argument I’ve heard is that everything is correlated for some reason.
Idk how we’re going to recognize super intelligence by ‘gud PC1’, but the positive manifold is kind of wild
Idk how we’re going to recognize super intelligence by ‘gud PC1’, but the positive manifold is kind of wild