Remark: I consider this a failure of Robin Hanson's predictions in the AI-Foom debate.

On review: It seems like most of the actual problems with SB 1047 as written are in the definition of a derivative model. In particular, the fact that the model remains derivative under unlimited additional training.

This is a mistake. We can and should fix it.

Are there any seed oil apostates?

Plenty of people left & denounced paleo, low-carb, keto, low-fat, vegetarianism, veganism, carnivore.

Anyone out there stopped eating seed oils, felt really shitty after a while, and it all got better adding soybean oil back in?

1/2 This is how civilization ends.

'Due to the dwindling supply of quality machine shops and very poor quality aftermarket parts as well as a lack of people interested in learning the engine building and machine shop trades, I regret to inform our many followers and current as…

Big Tech (especially Meta) pushed a conspiracy theory that tiny AI safety nonprofits were doing all the lobbying… all while they quietly outspent ALL of civil society perhaps “5 to 1, 10 to 1”

“While they have publicly been supportive of AI regulation, in closed-door…

Bryan Chau (Brian Chau)'s argument against AI foom:
* Science has diminishing returns
* AI doesn't help much with science today

Ok but chess AI vastly dominates humans, yet:
* Learning to play better chess has diminishing returns
* AI didn't help much with chess in 1960

astonishing story i just learned about: in 1883 an antiquities dealer, Moses Wilhelm Shapira, tried to sell some Paleo-Hebrew manuscripts, seemingly containing some variation of Deuteronomy, to the British Museum, but then third-party experts examined them and said they were…

Greenpeace and its allies just won a court case in the Philippines blocking Golden Rice from reaching millions of vitamin A-deficient young children. If this isn't a crime against humanity, I don't know what is. Avoidable deaths could total 100,000s. More: spectator.co.uk/article/childr…

'AI killing literally everyone just a Sci-Fi'

Saying that something is just Sci-Fi is no different than saying it didn't happen (yet).

Whatever approach to alignment we wind up using, it should be scale free and not specific to human-scale systems. Because we need to test it on systems smaller than human-scale and need to scale it up past human-scale systems.

Your periodic reminder that you can't trust AI companies, EVEN when they do public commitments to a major government.

At what point do you as a scientist start asking “is this stuff ontologically evil”

Alas, it's not fully general stupidity. But good on showing the problem with the people who get reasonable-sounding responses in particular contexts and believe they've developed something that Generally Works Fine.

'The question we should be asking,' one imagines the other tribes solemnly pontificating, 'is not 'What if the aliens kill us?' but 'What if the Aztecs get aliens first?''

This is a very revealing quote from Billy Bragg. If you're independent-minded (paulgraham.com/mod.html), you often find yourself on the same side of an issue as people you dislike. So if he's shocked by that idea, it shows he's not.

Are you fucking kidding me?

Underrated logical fallacy: ergo decedo, or traitorous critic fallacy; the critic's criticism is used as evidence that the critic is loyal to an out-group, therefore can be dismissed. Related to shaming tactics generally. A huge factor in bad discourse.

en.m.wikipedia.org/wiki/Ergo_dece…

if you think you know what a proof is try this puzzle from Timothy Gowers