Sign in

Flow Void

@flowvoid.bsky.social
527 followers 1.4K following 9K posts

All voids preserved.

PostsRepliesMedia
Flow Void @flowvoid.bsky.social · 3h
5K is over three months median rent. It would definitely make a difference for the majority of people.
000
Flow Void @flowvoid.bsky.social · 7h
That only applies to training on legally obtained works. He said training on pirated works is "irredeemably infringing".
110
Flow Void @flowvoid.bsky.social · 7h
bsky.app/profile/flow...
100
Flow Void @flowvoid.bsky.social · 8h
No, he didn't. Quite the opposite. Read to the end. bsky.app/profile/flow...
100
Flow Void @flowvoid.bsky.social · 8h
110
Flow Void @flowvoid.bsky.social · 8h
Relevant part of decision in Bartz v Anthropic:
110
Flow Void @flowvoid.bsky.social · 8h
Courts have ruled that training AI with *legally obtained* works is fair use (assuming the output does not reproduce protected works verbatim). But if pirated works are used to train AI, then it's not fair use. Which is a big problem for some LLM developers.
110
Flow Void @flowvoid.bsky.social · 10h
Posters like this generally have one thing on common: they never identify an actual candidate who they would vote for.
1211
Flow Void @flowvoid.bsky.social · 29/09/2026
95% of Americans don't live in SF, NYC, LA or Boston. So it's hard to believe those cities are driving survey results.
000
Flow Void @flowvoid.bsky.social · 29/09/2026
Another classic example is writing a story about a murderer and illustrating it with a picture of someone who is not the murderer (uncaptioned). There are no false statements, but the picture is defamatory*. * technically both these examples are not defamation but "false light", a similar concept.
001
Flow Void @flowvoid.bsky.social · 29/09/2026
Something can be true yet defamatory, if it misleads by leaving out important info. For example, "Bob criticized the town church on TV. The next day, Bob was arrested" That's possibly defamatory, if Bob was arrested for something unrelated like tax evasion. Something like that is alleged here.
100
Flow Void @flowvoid.bsky.social · 28/09/2026
Yes, they shall have the *sole power* to impeach. They can't delegate that power to someone else, like a special prosecutor, hoping to avoid responsibility. If they were required to use that power, it would say "The House shall impeach all high crimes and misdemeanors". But that's not what it says.
000
Flow Void @flowvoid.bsky.social · 28/09/2026
The Constitution says the House has the sole power to impeach, just as it has the sole power to originate revenue bills. It does not say they have any duty to use either power.
100
Flow Void @flowvoid.bsky.social · 27/09/2026
To be fair, the author agrees with you. I think she is trying to make a point about the kind of people who *refuse* to listen to country, but her article seems a little confused.
020
Flow Void @flowvoid.bsky.social · 26/09/2026
"Is everything made of particles?"
000
Flow Void @flowvoid.bsky.social · 26/09/2026
And that only matters in terms of how to avoid those fates. The solution to the first category is "try to be better". The solution to the second is "try to hide".
010
Flow Void @flowvoid.bsky.social · 26/09/2026
Regardless, we're responsible for creating the ones on this planet. But honestly, the Fermi solutions don't split hairs like this. If it helps, think of the category broadly as "avoidable mistakes were made", as opposed to unavoidable stuff like "annihilated by superior aliens"
110
Flow Void @flowvoid.bsky.social · 26/09/2026
That's under the general category of "they destroy themselves". Not necessarily by war or even intentionally, it includes things like climate change, social collapse, resource exhaustion, and anything else where they are responsible for their own destruction.
110
Flow Void @flowvoid.bsky.social · 26/09/2026
Well I definitely agree with you there. Regardless of what we call it, something astonishing and unprecedented is happening
020
Flow Void @flowvoid.bsky.social · 26/09/2026
Also, earlier you said solving quadratics is just plugging numbers into an algorithm. Is an LLM fundamentally different? If I had infinite time, I could produce the same results as ChatGPT by plugging numbers into formulas. Is that thinking or not?
100
Flow Void @flowvoid.bsky.social · 26/09/2026
If the integers were randomly generated instead of sequential, would it change anything?
100
Flow Void @flowvoid.bsky.social · 26/09/2026
I'm not claiming I fully understand either thinking or love. I am claiming that as far as I can tell, LLMs don't seem to be doing either of them.
100
Flow Void @flowvoid.bsky.social · 26/09/2026
I mean, you go to qualia when the phenomenon seems to involve qualia.
000
Flow Void @flowvoid.bsky.social · 26/09/2026
If navigating an open hypothesis space without a recipe guaranteed to converge is thinking, then solving quadratics by sequentially checking all integers is thinking...
100
Flow Void @flowvoid.bsky.social · 26/09/2026
Sure, but not being able to prove something doesn't mean we can't talk about it. We can't prove "love" for the same reason as "think", do we need to redefine love as something that can be objectively established by a neutral observer?
100
Flow Void @flowvoid.bsky.social · 25/09/2026
I think my debate with Michael boils down to whether thinking has more to do with qualia or problem-solving. I think it doesn't necessarily have anything to do with problem solving, eg when you say "I am thinking of my grandmother".
000
Flow Void @flowvoid.bsky.social · 25/09/2026
Depends, if qualia are essential to thinking then LLMs might be no closer than a rock.
300
Flow Void @flowvoid.bsky.social · 25/09/2026
I think people want to apply "thinking" to LLMs because they *want* to draw a parallel to humans. And "verbal processing" doesn't have strong enough connotation for that. But then all the folk-psychology baggage comes along for the ride. You can't get rid of it without breaking the link to humans.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
In fact, "I think" is the only direct observation of thinking. So folk-psychology is baked into that term, just like "love" and "fear'. If you want to avoid folk-psychology (and many do) then you should use different terms, like "verbal processing".
100
Flow Void @flowvoid.bsky.social · 25/09/2026
It's just one kind of problem solving. So is using the quadratic equation to solve problems in algebra, practically by definition. Geocentricism is a theory to explain observations, like a sunset. Theories are judged based on their consistency with observation. And "I think" is also an observation.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
You can't prove it, but you can have different probabilities of belief. For example I think it's more likely my dog has qualia than a rock.
200
Flow Void @flowvoid.bsky.social · 25/09/2026
Why is lack of a heuristic search over an open hypothesis space relevant to whether a Cray thinks? And the something I'm demanding is consistency with existing descriptions. If forced to choose between denying LLMs think and denying humans know whether they are thinking, I'll choose the former.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
(And something *not* included in all human reports of "not thinking")
100
Flow Void @flowvoid.bsky.social · 25/09/2026
Or to be more precise, the LLM / Cray reports do not include something included in all human reports of thinking.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
If a Cray reports every step it takes while finding pi, is that a report of actual thinking? Earlier you suggested no, the program itself lacks something essential to thinking no matter how well it explains itself. But that can also be said of LLMs, the reports are insufficient.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
In all these cases, problem solving is delegated to unconscious processing and awareness comes only after the problem is solved. But the existence of non-thinking mechanisms doesn't invalidate reports of actual thinking.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
Human problem solving does not always require thought, earlier I gave the example of walking and recovering balance after stumbling. Driving is another example, and so is playing a piano by the trained pianist in the article. People don't report "thinking" about these activities, just the opposite!
100
Flow Void @flowvoid.bsky.social · 25/09/2026
(And this is why when people want to exclude mundane "thinking" and study only certain types of problem solving, they often develop new jargon to describe it)
000
Flow Void @flowvoid.bsky.social · 25/09/2026
If your definition of "thinking" excludes activities people normally consider "thinking", then there is no point to trying to expand it to AI.
200
Flow Void @flowvoid.bsky.social · 25/09/2026
If you ask someone who had a flash of insight "Did you think about this problem before you had the insight?" they will answer yes. And I think your view of human thinking is too narrow. It's more than complex insight. Solving trig problems can be rote but we generally report "thinking" about them.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
In fact, if a human tried to follow the same procedure a Cray used to calculate pi, that would be considered thinking. So how can you draw an equivalence between an LLM and human thinking without allowing an equivalence between the ancient Cray and human thinking?
100
Flow Void @flowvoid.bsky.social · 25/09/2026
Even accepting Poincare, there is more to problem solving than flashes of insight. The drudgery is less dramatic but it's necessary and requires awareness that isn't post-hoc. Humans think when they do long division but they aren't searching undecidable spaces, why is that a relevant factor for AI?
200
Flow Void @flowvoid.bsky.social · 25/09/2026
(Or put differently, you wouldn't consider a pi-finder to be similar to an LLM even if you programmed the pi-finder to explain its steps!)
000
Flow Void @flowvoid.bsky.social · 25/09/2026
In other words, to the extent that LLMs are separated from pi-finders because of how they work in a general sense, so are humans separated from LLMs.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
There is more to solving it than explaining the steps. You made this point yourself! Unlike finding pi, the process used by LLMs isn't guaranteed to converge. That's not a step in the process. It's a significant *property* of the process. And awareness is a property, not step, in the human process.
200
Flow Void @flowvoid.bsky.social · 25/09/2026
(That said, in my limited experience LLMs generally deny self-awareness, and I wouldn't expect that to change when instructed to solve proofs)
110
Flow Void @flowvoid.bsky.social · 25/09/2026
Sure, but even if the LLM said it was self-aware while solving proofs I'm not sure I would believe it.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
I know that's not particularly convincing. Some argue that interiority is always an illusion, not just for proofs but for everything else. I don't have a good counterargument for that, so I probably can't convince you if you think interiority is meaningless.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
Well, it's true that we don't know much about interiority. Lacking measuring tools, our knowledge depends on anecdotes. But you don't really hear of people finding proofs with no awareness at all of the process. So I'm inclined to believe it's a prerequisite for this process in humans.
100
Flow Void @flowvoid.bsky.social · 25/09/2026
That's not invalid! But if you use this as a basis to argue that finding digits of pi is dissimilar to finding a proof, then one can just as well argue LLMs that find proofs are performing a different task than humans who find proofs. The process matters.
100