Can we actually get AI to pick plastic out of the ocean, or do all the robots have to become screenwriters?

unfortunately AI safety is in the hands of lunatics who believe that if you shut down a computer you might be "killing" it and violating its rights

“The Low End Theory” by A Tribe Called Quest turns 35 years today.

AIs and LLMs don’t have agency. Period. Autonomy ≠ agency. Conflating the two is how we end up fear mongering and politicking.

No, LLMs do not feel pain simply because there is a cluster in language space correlated with how people use language about pain in a certain set of contexts. I believe this argument to be flawed, and will write more about it in October.

Why does the AI industry wants you to believe in “rogue agents” when we are really just talking about irresponsibly developed software? Here’s a good theory:

GN

i don't speak to AI because to do so I'd have to turn off the music.

Any time someone tells you that AI has an X% chance of wiping out humanity, ask them for their math that got them the number. If they didn’t do a calculation, if they just pulled the number from their butt, you can safely treat it like anything else that came from their butt.

I wish people would notice that the safety scare is not driven by people outside of the big labs, but directly from their leadership. The "whistleblowers" are plants that are amplified from within the labs, and supported by the CEOs. The goal is to capture the regulators, which will not shut down the big labs, but turn them into an oligopoly and outlaw decentralized AI.

If you think a LLM is conscious when you talk to it, what about alpha fold when you fold a protein or a chess computer when you play a move. The term conscious has no meaning in this context.

Rabbits, 1933 by Jean Charlot (French, 1898–1979)

wow i missed that @fire_toolz had an album on warp. super fun listen! fire-toolz.bandcamp.com/albu…

Fair point indeed from @naval : If AI companies (Anthropic/OpenAI) train on the open web, perhaps they should be required to open their models after a certain period. "And Anthropic, you’re the one talking doom and gloom, and you’re building God, but you’re not going to keep God on a leash for yourself, Dario." --- "If the government were to do it, is to say, “Hey, if you trained on the open web and open data, you have to open your model after after 12 months or something.” That seems like a pretty fair thing to say. We don’t want, like, a priest controlling God for the rest of us and interpreting the Bible for the rest of time." ---- From "Naval and Nivi" YouTube channel, (full video link in comment)

Guess who gets f_cked in the end?

AFAIK the most significant breakthrough since 2017 besides scaling old ideas was broadening from base models into larger systems that incorporate symbol/manipulating entities like harnesses, tools, and code interpreters, enabling things like Claude Code.

sketching with #threejs