Lightnews — Scholar-powered news

Eryk Salvaggio @eryk.bsky.social · 9h

I’ve just learned that John Searle, whose Chinese Room thought experiment is often used to challenge ideas of “understanding” in LLMs, died at age 93 on Sunday. www.theguardian.com/world/2025/o...

John Searle obituary

American philosopher whose Chinese Room thought experiment rebuts the idea that computers can think as humans do

www.theguardian.com

2 17 45

Eryk Salvaggio @eryk.bsky.social · 1d

Ah, understood!

Eryk Salvaggio @eryk.bsky.social · 1d

Sorry, what “position” are you referring to here?

1

Reposted by Eryk Salvaggio

Eryk Salvaggio @eryk.bsky.social · 1d

The "you don't understand how (AI/LLMs/Diffusion Models/NNs) work" posture in any debate is often just "we don't agree on how to interpret what (AI/LLMs/Diffusion Models/NNs) are doing"

6 8 47

Eryk Salvaggio @eryk.bsky.social · 1d

Agree! And others!

1

Eryk Salvaggio @eryk.bsky.social · 1d

Gotcha, and agree, I'd missed what you meant by "capacities" of the model.

1 3

Reposted by Eryk Salvaggio

Clemens Vasters 🇪🇺🇩🇪 @clemens.vasters.com · 1d

Techmeme front page:

2 110 290

Eryk Salvaggio @eryk.bsky.social · 1d

I think we agree on principle, but I'm thinking more that the data is translated into numerical representations (weights, biases) and the NN is the process through which that translation occurs. A handwritten "9" is not an image recognition network in itself.

2 3

Eryk Salvaggio @eryk.bsky.social · 1d

Point is, if you think NNs are most efficient formula for optimization, you may see things like education, the arts, and politics as moving through training layers too. The end result of which will be the full optimization of society. So you throw random stuff out and see what people do with it.

1 5

Eryk Salvaggio @eryk.bsky.social · 1d

I think it's pretty common for people to see the world through the lenses they work with (once you do cybernetics, everything is "just" a feedback loop; once you do pattern-finding, everything is "just" pattern-finding).

1 4

Eryk Salvaggio @eryk.bsky.social · 1d

Striking parallel between how NNs are trained and how VCs imagine the world is built: start with noise, then remove noise through "optimization." There are assumptions not just that the machine "learns" as a human does but that the world will "learn" to "optimize AI" just as an NN "learns."

2 12

Eryk Salvaggio @eryk.bsky.social · 1d

The extent to which biases in data influence biases in NN weights is pretty hard to ignore tbh, I still want to dig into any good faith arguments by those who disagree or challenge this. Most agree but seem to find it inconvenient, so don’t focus on it!

2 5

Eryk Salvaggio @eryk.bsky.social · 1d

Yes, and my point is that shifting this disagreement from “debating interpretations” to “the other party is ignorant of basic architecture” can be counterproductive.

1 4

Eryk Salvaggio @eryk.bsky.social · 1d

The "you don't understand how (AI/LLMs/Diffusion Models/NNs) work" posture in any debate is often just "we don't agree on how to interpret what (AI/LLMs/Diffusion Models/NNs) are doing"

6 8 47

Eryk Salvaggio @eryk.bsky.social · 1d

Totally agree, and it's unfortunate that prompt-based AI music is literally forced to be navigated by describing narrowly constrained genres -- especially when "genre collision" is already so well established as to be its own genre.

1 1

Reposted by Eryk Salvaggio

Mike Sowden @mikeachim.bsky.social · 1d

Damn. This is amazing. £325 per week, paid monthly, for 3 years - and the result was a profit for the Irish economy:
www.citizensinformation.ie/en/employmen...

Post from Threads user rodneyowl: "Ireland has declared the Basic Income for Artists scheme permanent. This will be officially announced in tomorrow’s budget. Details to follow. Congratulations to all who fought for it and the present and future artists of all sorts in Ireland. That includes me 👌We’re just comin to the end of a 3 year pilot scheme. It’s been a roaring success. For every €1 paid out to the 2000 participants, the government got €1.46 back. Can’t argue with that. Other countries are already taking note."

130 5.2K 15K

Eryk Salvaggio @eryk.bsky.social · 1d

I feel like that would create a kind of explosion of new experiments, no?

2

Eryk Salvaggio @eryk.bsky.social · 1d

I think you have it. Seems like rising populist conservatism alongside consolidation of tech & entertainment industry at a moment where tech is also shifting its money into an endless GPU expansion … is a bad-timing situation that has stifled support for the arts by industry & philanthropy alike. :/

1 1

Eryk Salvaggio @eryk.bsky.social · 1d

This makes sense! But we keep hearing the contrary re: 1968, for example. I suspect there may really be something to be said this time around about the entertainment industry consolidating production power, lots written on radio in the 21st century, streaming services etc may be on the same track?

1

Eryk Salvaggio @eryk.bsky.social · 1d

But certainly there are also fallow periods. It’s just curious to me that technological and social upheaval seems to have exhausted everyone when everyone seems to suggest punk and rock came from the same conditions.

1 1

Eryk Salvaggio @eryk.bsky.social · 1d

I understand it, but it’s worth noting: since the popular adoption of generative AI, culture feels like it’s been at a standstill. It might be my filters, but I rarely feel inspired by discussions of books, or catch great indie films, or come across unique and exciting (new) records anymore.

8 16

Reposted by Eryk Salvaggio

Anna Kornbluh @annakornbluh.bsky.social · 2d

"the solution is still the one that can succeed: to build a new cultural order, a new civilization. To do so, academics must embrace an unusual new role: as knowledge workers, they must seize the means of knowledge production."

www.publicbooks.org/academics-mu...

Academics Must Seize the Means of Knowledge Production - Public Books

Trumpism has canceled the knowledge society.

www.publicbooks.org

2 21 65

Reposted by Eryk Salvaggio

Tech Policy Press @techpolicypress.bsky.social · 2d

While there is value to foresight and anticipating risks, writes Tech Policy Press fellow
@eryk.bsky.social, could the language used by AI risk communities to describe the technology contribute to the very problems it aims to curb?

The AI Safety Debate Needs AI Skeptics | TechPolicy.Press

The language used by AI risk communities to describe the technology may contribute to the very problems it aims to curb, Eryk Salvaggio writes.

www.techpolicy.press

3 5

Eryk Salvaggio @eryk.bsky.social · 2d

Any organization considering the risks of AI should make room for those who believe the greatest risk is believing in it. In my latest for @techpolicypress.bsky.social, I propose that AI skeptics play a crucial role in discussing “AI safety.” www.techpolicy.press/the-ai-safet...

The AI Safety Debate Needs AI Skeptics | TechPolicy.Press

The language used by AI risk communities to describe the technology may contribute to the very problems it aims to curb, Eryk Salvaggio writes.

www.techpolicy.press

6 18

Reposted by Eryk Salvaggio

Eryk Salvaggio @eryk.bsky.social · 3d

Wrote more about Benjamin Bratton's paranoid imagination of a media studies cabal holding back AI progress in Europe. mail.cyberneticforests.com/is-the-media...

Is the Media Studies Cabal in the Room With Us Right Now?

I read The Stack in 2020 as a grad student in ANU's Applied Cybernetics program. I give it credit for directing my attention to the interaction between layers of digital and physical infrastructures. ...

mail.cyberneticforests.com

8 24 84