This is another "must read" AI article in @nytimes. Jaron Lanier speaks with experience about Silicon Valley's ability to think fantastically about what their technology can do - to the point of radicalism.
He unpacks Anthropic's unusual & error-prone approach to safety, which is a direct result of their religious belief that AI models will one-day run the world and "manage the humans." It's this belief that causes them to approach safety through "alignment," vs control & constraint.…
This is another "must read" AI article in @nytimes. Jaron Lanier speaks with experience about Silicon Valley's ability to think fantastically about what their technology can do - to the point of radicalism.
He unpacks Anthropic's unusual & error-prone approach to safety, which is a direct result of their religious belief that AI models will one-day run the world and "manage the humans." It's this belief that causes them to approach safety through "alignment," vs control & constraint.
Regulators should understand that "building a technology that eventually controls us" is their ultimate goal or assumption. As such, this one company should be as far away from the regulation process as possible. (unfortunately they are actively in the mix).
The mid-terms were won on the message of being anti-big tech & anti-billionaire. This will be an interesting test. Anthropic has built the most-potent "billionaire machine" ever created. If they get to write the regulation, they will try to kill open-models, and create 10x more billionaires than they already have.
I will leave you with this from Jaron; "Someone’s metaphysical beliefs should be protected and are none of my business. It is also reasonable for nonbelievers to refuse the imposition of a religion from others."
t.co/ASkInCjQCs