Casey Newton

There's been a lot of chatter over the past day about whether platforms should slow down the release of large language models. I'm persuaded that we should:

https://www.platformer.news/p/the-ai-industry-really-should-slow

The AI industry really should slow down a little

This year has given us a bounty of innovations. We could use some time to absorb them

Platformer
Mar 31, 2023 at 12:25amIvory for Mac
Show threadCasey NewtonMar 31, 2023
Btw it’s a good night to subscribe to Hard Fork https://www.nytimes.com/column/hard-fork
Hard Fork

Each week, journalists Kevin Roose and Casey Newton explore and make sense of the rapidly changing world of tech.

Show threadTaco DaveMar 31, 2023
@caseynewton booooo I was hoping the trans people and allies at NYT were hard forking and making their own news org.
Show threadJon NemargutMar 31, 2023
@caseynewton Y’all are all in on this Bard Fork rebrand
Show threadMarkham NolanMar 31, 2023
@caseynewton loved the user responses in the last episode. Some amazing use cases.
Show threadTom WatsonMar 31, 2023
@caseynewton huge get! Congratulations.
Show threadrabbleMar 31, 2023
@caseynewton I’m really enjoying the podcast. Nice mix of conversation and edited news with focus.
Show threadCasey NewtonMar 31, 2023
@rabble thank you!
Show threadStefanMar 31, 2023
@caseynewton companies probably will not care but they need to see that a bad outcome will also destroy their company at a speed they can't recover from.
Show threadJP EtcheberMar 31, 2023
I asked Chat GPT to write a rebuttal.
Show threadOmid MnzMar 31, 2023
@numbertheory it's talking as if it's one of us: "We should be thoughtful..."! WE? :))
Show threadJP EtcheberMar 31, 2023
@omidmnz "We're all trying to find out who the rogue AI is. It could be any one of us!"
Show threadJP EtcheberMar 31, 2023
@caseynewton I feel like it is a Pandora's Box situation. The only way we can really control it is through educating the public to be more discerning and not give into biases so quickly. Yes, it's going to be a disaster.
Show threadjquilty3Mar 31, 2023
@caseynewton can we pause with the emojis too?
Show threadMathias af JochnickMar 31, 2023
@caseynewton I definitely hear what you’re saying but i think it’s too late? People are doing all sorts of crazy things with these leaked and serviced models all over the world, there’s *so* much happening, i think that you might not be able to get the beast back in the cage.
Show threadCasey NewtonMar 31, 2023
@2happy1sad this sort of preemptive defeatism is not going to serve anyone well
Show threadPeter F. EbbinghausMar 31, 2023

@caseynewton @DAIR put it well today, too

https://www.dair-institute.org/blog/letter-statement-March2023

Show threadmarkApr 3, 2023
@caseynewton If you've not yet read this paper, you should probably have a look, given your thesis. I think it's important to the discussion of what it takes to hit AGI. Bottom line: it may be surprisingly easy at this stage. https://gwern.net/scaling-hypothesis
The Scaling Hypothesis

On GPT-3: meta-learning, scaling, implications, and deep theory. The scaling hypothesis: neural nets absorb data & compute, generalizing and becoming more Bayesian as problems get harder, manifesting new abilities even at trivial-by-global-standards-scale. The deep learning revolution has begun as foretold.