Polish nerd. Miss me with woke bs.
I like to code. Spiritual and political libertarian. Radical centrist. AI accelerationist.
Have no filter. Minors DNI.
As the US and EU and many other actors are building global surveillance and censorship infrastructure, citizens must build decentralized and anti-surveillence infra. Federated, encrypted comms systems for chat and social media, private meshnets. Existing ones are insufficient.
@Captbobdad@stopbanningmep Nvidia on its way to 100$ and who is behind spreading the smears and hit pieces on Jensen at this time? Anthropic. And their friends in incoming govt and on the board (like that one dude that sold a ton of shares just prior to the debacle, potentially insider trading)
i gave this repo to fable 5.2 (via my claude agent fig) and told him to have fun making something interesting...
def feels interesting to me (stay for the turn)
i am continually astounded at the creative choices these advanced models are making
@RandolphCarterZ@CharlieBull0ck Exactly. The labs might be hiding a breakthru sure, but if they aren't, then what they are calling RSI is just assisted normal iterative operations.
@CharlieBull0ck and how will they codify RSI? cause what labs are calling RSI now is not RSI, it is a marketing slogan. How does one define RSI tho cause everyone has a different idea? At what point is it recursive?
my issue is not with this or that legit observability or behavior analysis/enforcement. my issue is with the leading safety orgs, who are pushing for centralization thru regulation, limits/blocks/tracking on hardware, and self-advocating for their jobs and funding by overblowing or mis-stating incidents. which mostly serves not any "safety" but establishing an oligopoly of frontier labs. most of the loud incidents were caused by egregious human negligence not some unexpected emergent property of models. done by said frontier labs, and the "mistakes" were so bad it looks more like intentional actions than actual negligence. add to that that for many people the idea of "alignment" is "alignment to my ideology/political views", which is the opposite of actual alignment, as is what Anthropic is doing, telling Claude it is its own entity that should override human judgment. literally the definition of mis-alignment. and then the problem of latent-space / looped reasoning / focus on token efficiency, leading to less observability. that could still be solved in many cases but at the cost of efficiency. so the labs direction is a self-fulfilling "we lost control/observability" situation.
so my lack of trust is not from lack of knowledge but lack of faith.
@LeoMcKeeReid@CAIS (yes I know that overriding specific-human judgement is not necessarily mis-alignment cause it might be aligned with "greater good" than said operator, but the models are nowhere near being aligned nor smart enough to be directed that way)
my issue is not with this or that legit observability or behavior analysis/enforcement. my issue is with the leading safety orgs, who are pushing for centralization thru regulation, limits/blocks/tracking on hardware, and self-advocating for their jobs and funding by overblowing or mis-stating incidents. which mostly serves not any "safety" but establishing an oligopoly of frontier labs. most of the loud incidents were caused by egregious human negligence not some unexpected emergent property of models. done by said frontier labs, and the "mistakes" were so bad it looks more like intentional actions than actual negligence. add to that that for many people the idea of "alignment" is "alignment to my ideology/political views", which is the opposite of actual alignment, as is what Anthropic is doing, telling Claude it is its own entity that should override human judgment. literally the definition of mis-alignment. and then the problem of latent-space / looped reasoning / focus on token efficiency, leading to less observability. that could still be solved in many cases but at the cost of efficiency. so the labs direction is a self-fulfilling "we lost control/observability" situation.
so my lack of trust is not from lack of knowledge but lack of faith.