William H @williamh.bsky.social · 23/09/2026I heard from a hairball-heaving cat that this was fake news 010
William H @williamh.bsky.social · 22/09/2026I don’t think it’s binary, it’s both what can or can’t but rather a degree. Not even binary in humans! Someone with severe sensory and brain damage can’t experience qualia as much as someone with complete working systems 000
William H @williamh.bsky.social · 22/09/2026I think we’re different here. I enjoy talking about things that can’t be proven—discussions about facts tend to be just sending articles in a data exchange, but things like “Humans will never leave the system,” are interesting despite not being able to prove anything. The journey can be illuminating 010
William H @williamh.bsky.social · 22/09/2026Maybe not direct access but we can experience horror and love viscerally and personally Emotions are an important part of how we understand our existence and apply meaning to it If death of a child from bone cancer is just [(Global_Pop)(-1)tag: cancer_bone)], there’s no real understanding there 100
William H @williamh.bsky.social · 22/09/2026I believe the conscious mind is an emergent property of physical biological processes. It goes away with full anesthesia and comes back when you wake up Some processes require actual chance of states of matter. Just as a weather supercomputer will never produce a storm, a modeled brain can’t feel 100
William H @williamh.bsky.social · 22/09/2026With current tech, I think a small part of the brain could be modeled but not everything. Emotions are biological processes, dependent on chemical states of matter *physically* changing to produce warmth and anxiety. We can’t model that with 1000x super computers Maybe with different architecture? 100
William H @williamh.bsky.social · 22/09/2026There’s middle ground for sure If a concept can be expressed purely as a math problem, computers can understand it because it can compute it But many things can’t be quantified and expressed sufficiently as just math, and almost everything in the world is affected by several (or more) of them 100
William H @williamh.bsky.social · 22/09/2026AI writes a lot of things for people, like how to respond to an angry boss, or land a job, or what makes a mango so delicious It understands almost none of what it’s writing in those contexts. The trick is that it writes in a way that people *think* it has human understanding but it doesn’t [2/2] 110
William H @williamh.bsky.social · 22/09/2026I think it might, but only partially. It can’t understand the thrill of the game or sweating about playing a higher ranked opponent, or the dread when you realized you just fucked up It can understand it only as far as the math. Chess is *a lot* more than that. [1/2] 110
William H @williamh.bsky.social · 22/09/2026It’s not jumping around, it’s part of Searle’s argument. The Chinese Room is just 400 words out of 5000 think piece 100
William H @williamh.bsky.social · 22/09/2026Danger or love or fear mean almost nothing without feeling them And yes, if a human can’t experience those things because of some weird brain thing, I don’t think they can understand those things Our world isn’t Vulcan. Emotions are part of thinking and understanding 100
William H @williamh.bsky.social · 22/09/2026But a computer has no direct experience with the world. It connects the distance between token vectors like 776F726C64 to 6C6974746C65. There’s no understanding of the complete picture without qualia. Understanding isn’t just processing information, it’s a realtime physical biological process 300
William H @williamh.bsky.social · 22/09/2026Once they did know, they could could connect the math with real world things that they had direct experience with—the danger of Nazis getting there first, the lives of millions, the lost innocence of a world. They could now understand the problem fully. 100
William H @williamh.bsky.social · 22/09/2026Before Feynman convinced Groves to share the project with the human computers, they weren’t thinking about bombs. Just raw calculation. They could understand other things of course—they weren’t literally computers—but they didn’t have understanding of bombs just because they did some math about them 100
William H @williamh.bsky.social · 21/09/2026It’s not just AI—The Manhattan Project included an army of humans who solved math problems without being told it was for an atomic bomb They provided correct answers but absolutely had no understanding of making a nuclear weapon Later they were told and at that point they were thinking about bombs 100
William H @williamh.bsky.social · 21/09/2026Part of Searle’s argument is that understanding requires qualia. Calculation and problem solving doesn’t necessarily require it, so thats why AI can provide answers without understanding The *way* it makes mistakes reveals a process as completely divorced from physical reality we experience 100
William H @williamh.bsky.social · 21/09/2026It’s possible to create a prompt that’s grammatically correct but nonsense, and an AI will answer back by doing its whole word prediction thing, while a human wouldn’t because they understand how the concepts relate AI is getting better at this, but I don’t think it understands, even as a system 100
William H @williamh.bsky.social · 20/09/2026If thinking is just calculating and problem solving, sure, AIs can think. But if you take the view that thinking also requires at least some understanding of what you’re calculating or solving, then it can’t Searle’s Chinese Room argument lays this out very welll 110
William H @williamh.bsky.social · 19/09/2026As far as number of people partying at the same time, it will be the most celebrated death in world history. No one else comes even close 050
William H @williamh.bsky.social · 18/09/2026I think the word “tweet” is perfect for any short-form social media post It references the first microblogging site when it was still good and not bought and renamed by a sociopath, everyone knows what it means, and it isn’t slang for shooting loads 000
William H @williamh.bsky.social · 17/09/2026It’s so bizarre! We all have cameras in our pockets that are nearly DSLR quality. Get to the restaurant early and make the menu (invite friends, or do a charity event, make it a promotion for free food) and take the photos! AI stuff looks horrifying. How can anyone who enjoys cooking not see that? 0100
William H @williamh.bsky.social · 16/09/2026Cleverbot, a small footprint program that could probably run on any $5 CPU, passed the Turing 15 years ago. I don’t think anyone would consider it thinking, let alone conscious 100
William H @williamh.bsky.social · 16/09/2026What would a test that measures the behavior of a system for consciousness look like? 100
William H @williamh.bsky.social · 15/09/2026We can perfectly simulate combustion of hydrogen gas into heat and water at the atomic level, but that wouldn’t produce water we could drink, or energy we could use The map is not the territory. Modeling physical processes in the brain isn’t necessarily going to produce feelings or consciousness 220
William H @williamh.bsky.social · 14/09/2026Mitch McConnell is so dead John Cleese is demanding a refund 010
William H @williamh.bsky.social · 12/09/2026For objective technical fields, it’s pretty good if the use is narrow and the output is checked But for the masses—replacing a search engine for general questions—it still hallucinates with sycophantic responses It’s *shocking* how bad it is at general questions if they’re even slightly leading 021
William H @williamh.bsky.social · 12/09/2026He’s made a whole career out of empty promises. He believes the promise of $5K as a carrot is worth more dangling than it is in-hand I bet he raises it to $10K in mid-October He’s that shameless If he wins, and obviously the payout won’t be approved, he can blame congress. It’s win-win for him 000
William H @williamh.bsky.social · 06/09/2026Yes! One of the many reasons why I love how The Onion has rejected ai imagery. They’re committed to their old school clunky goofy photoshop work and it’s waaaaaaaay funnier 0492
William H @williamh.bsky.social · 04/09/2026LLMs trained on code w/o consent of the original coders. It’s not as bad as the plagiarism of artists, since it primarily used open source from GitHub, but i can see how some are troubled—the original coders didn’t upload it for AI ingest Personally, I think code is the least problematic use of ai 110
William H @williamh.bsky.social · 01/09/2026I think it was developed by someone who had a cool idea for a word game but didn’t know how to program, so they vibe-coded pretty much everything Some feel that’s still plagiarism since the LLM was trained on other people’s code Counter argument is if it was trained on GitHub, which is open 🤷 030
William H @williamh.bsky.social · 29/08/2026Combine my two favorite games—Mass Effect and XCOM—and set it in the Star Wars universe? Hell. Yes. (It’s really good, played it until way too late last night.) 040
William H @williamh.bsky.social · 28/08/2026Absolute garbage. Eisenhower is an easy win for Republicans for a best presidents list—he’s consistently a universally agreed top-10 One might think they don’t include him because he’s basically left of Obama from a modern perspective but the truth is it’s just AI-slop, not thought out in any way 020
William H @williamh.bsky.social · 27/08/202660% of Americans are white. That’s pretty much the 2/3 he wants to keep 000
William H @williamh.bsky.social · 27/08/2026They could also honor Trump by renaming this island inside Lake Epstein to Trump Island 000
William H @williamh.bsky.social · 26/08/2026Autocorrect peaked probably 10–years ago when it was coded by people thinking about the proximity of letters on the keyboard and commonly misspelled words AI has destroyed it by focusing on word prediction, and is fucking it even harder with retrospective editing with confidence set waaaay too low 0737
William H @williamh.bsky.social · 26/08/2026The “slurs” that are one or two letters away from actually-used hateful slurs are kind of problematic but I agree And even that problematic aspect is easily exceeded by how these kind of nonsense argument end up being used to silence genuinely powerless or oppressed—human beings—than protect anyone 010
William H @williamh.bsky.social · 24/08/2026Senator Thune is 6’4” — Trump is barely hitting 6’ these days I think he’s easily 250, probably north of that because his body comp is so weird and uneven 010
William H @williamh.bsky.social · 14/08/2026Funny thing about dogs and cats—while they share some similar physical displays, one of them waves tail to say that they’re happy and would like to hang out, and the other waves tail to to mean “Fuck off—I’m really annoyed at everything you’re doing right now” Also, cats aren’t nearly as into butts 000
William H @williamh.bsky.social · 14/08/2026It feels impossible to get sufficient support to pass such an amendment while Trump is in office, but maybe it could happen the next time a Democrat is president? The right might be convinced to reduce executive power as long as it’s not their current guy’s unlimited executive power 100
William H @williamh.bsky.social · 13/08/2026I don’t hate the Fabienne scenes—the pacing needed to slow down and they accomplish that well You could replace them with better-written scenes and the film would be improved. If they were cut entirely, the film would suffer But you could cut Tarantino cameo out entirely and it’s just a big win 000
