tweetermen.bsky.social @tweetermen.bsky.social · 20hThe editor that approved this makes me think AI can take over some jobs already 000
tweetermen.bsky.social @tweetermen.bsky.social · 11/10/2026Including “passed phase 3 clinical trials” really undercut the whole thing. 130
tweetermen.bsky.social @tweetermen.bsky.social · 11/10/2026I mean, is anyone an ML expert on his level 😂 thank you for sharing the links, I will give them a read this evening! I find myself frustrated that I don’t believe the all powerful super intelligence hype but concede I can’t really articulate why 110
tweetermen.bsky.social @tweetermen.bsky.social · 11/10/2026I always look forward to his interviews but I wish that he would explain why he doubts super intelligence; I’m surprised how quickly he dismisses Hinton’s reasoning as “not credible” 150
tweetermen.bsky.social @tweetermen.bsky.social · 10/10/2026This embarrassing but I did read an article that didn’t quote the tip and didn’t read your screenshot 000
tweetermen.bsky.social @tweetermen.bsky.social · 10/10/2026Are the contents of the tip public? I’d like to report the same tip to see what happens 100
tweetermen.bsky.social @tweetermen.bsky.social · 08/10/2026I would think AI will help with FTQC just fine? 000
tweetermen.bsky.social @tweetermen.bsky.social · 08/10/2026Would you concede that in all likelihood 'some of their results turn out to be somewhat correct, not plagiarized, and not incorrectly autoformalized' is "most of their results", or do you actually expect a majority to not clear that bar? 000
tweetermen.bsky.social @tweetermen.bsky.social · 28/09/2026Thanks for replying! Big fan of your blog (but haven’t looked at the guide closely). Since Mythos/Fable, I’ve been shocked by agent capabilities you and others have demonstrated. Are you still finding tasks where your expertise and guidance are critical, and if so, any commonalities to them? 010
tweetermen.bsky.social @tweetermen.bsky.social · 27/09/2026I think the baseline training of the dog and the owner in the moment makes a difference. If my usually well behaved dog ignores me saying “sit! stay!” to bite a mailman, then I feel rogue is pretty accurate bc it defied its training and instructions like an OpenAI agent 150
tweetermen.bsky.social @tweetermen.bsky.social · 26/09/2026@simonwillison.net I saw your x post about agents making development harder. As someone who did a lot of hand coding but manages now, do you have suggested resources for keeping up? Besides hands on experience of course 130
tweetermen.bsky.social @tweetermen.bsky.social · 24/09/2026damn the movie really is the perfect metaphor for agents 000
tweetermen.bsky.social @tweetermen.bsky.social · 24/09/2026Your data scientists were so preoccupied with whether they could, they didn't stop to think if they should. 111
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026Okay that feels like a precondition to me yes 100
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026damn I need an edit button but you get it 000
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026I mostly agree. I call consciousness metaphysical right now not bc it requires something new and mystical, but because we have no meaningful connection between physics and consciousness right now. And fair point I'm totally convinced they are not consciousness but I can't prove that at all 110
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026I'm not even sure which way that goes, do you experience the passage of time because of consciousness, or is the passage of time a precondition to consciousness. I'm not sure how you can 'experience' anything without consciousness. Of course, I mean this all metaphysically 200
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026Maybe, maybe not. I don't think it answers the question of 'why does that configuration give rise to qualia instead of just not giving rise to qualia' 100
tweetermen.bsky.social @tweetermen.bsky.social · 23/09/2026i think the first guy that said metaphysical really meant “it’s hard to explain ‘understand’ without referencing conscious experience.“ I think the brain vs computer thing is up in the air till we can explain why living things have experienced and other things don’t. 200
tweetermen.bsky.social @tweetermen.bsky.social · 17/09/2026Someone tried to help me understand the longterm pov with: - World A: 10 bln people with avg happiness score of 8 - World B: 10 bln ppl w/ avg score of 5 - World C: 5 bln ppl w/ avg score 8 And then said 'doesn't your intuition tell you A is better than C?' No, no it doesn't, like at all... 011
tweetermen.bsky.social @tweetermen.bsky.social · 17/09/2026Ah okay. So in the case of AI, it isn’t so much as “I don’t think AI can plausibly have X consequence” as much as it doesn’t help solve X? 110
tweetermen.bsky.social @tweetermen.bsky.social · 17/09/2026Can I ask why no dooming? I see doomers lay out lots of arguments, I see lots if people dismiss then, but I’m really not seeing why they dismiss them. Like what am I not seeing that the non-doomers are seeing?? 110
tweetermen.bsky.social @tweetermen.bsky.social · 17/09/2026Here it is scottaaronson.blog?p=3903scottaaronson.blogBeyond fictionI now know firsthand what it’s like to be arrested by armed police officers, handcuffed, and sharply interrogated, while one’s wife and children look on helplessly. This is not a prank post.… 010
tweetermen.bsky.social @tweetermen.bsky.social · 17/09/2026He stole the tip jar and the police weren’t nice to him so as usual it becomes about Israel or WWII. I am curious what he says that’s wrong about quantum computing though that you and @beenwrekt.bsky.social Ben refer to. I thought he was reliable there 110
tweetermen.bsky.social @tweetermen.bsky.social · 16/09/2026Obviously not obligated at all. He knows way more than me and I wonder why he isn't worried. I haven't understood it from his blog posts. NS worries me because of how fast AI went from bad at math to brilliant. Plenty of the AI discourse before it worried me as well, like bio risks. 000
tweetermen.bsky.social @tweetermen.bsky.social · 16/09/2026This would be awesome. I believe Nathan Witkin is looking for someone to help with this exact topic 000
tweetermen.bsky.social @tweetermen.bsky.social · 16/09/2026I find your stances enigmatic. You don’t think it’s normal tech, you don’t think it’s snake oil, you don’t think it’s silicon god. I would like to not be worried, but I guess just can’t see why someone wouldn’t be? NS has me rattled. PS I say this because I find your writing insightful 110
tweetermen.bsky.social @tweetermen.bsky.social · 16/09/2026So I read it. Enough hedging and both side interpretations that it’s hard to act like “he called it.” I don’t deny rationalists have been better at predicting AI capabilities compared to most people, but people didn’t think about AI capabilities. And predictions of RSI doom predate them, by a lot 000
tweetermen.bsky.social @tweetermen.bsky.social · 14/09/2026blog.dileeplearning.com/p/amelia-bed... seems in your laneblog.dileeplearning.comAmelia Bedelia and AGI Safety. Part 1..on the questionable beliefs behind AGI existential risk concerns 090
tweetermen.bsky.social @tweetermen.bsky.social · 13/09/2026“Amelia Bedelia with superhuman strategic planning abilities and encyclopedic knowledge of everything in the universe but zero sense of nuance” It’s always interesting to me when super human persuasion is part ofthe doomsday. How could ASI have that understanding of psych but not any nuance 020
tweetermen.bsky.social @tweetermen.bsky.social · 12/09/2026(The physical limits of intelligence bit) 000
tweetermen.bsky.social @tweetermen.bsky.social · 12/09/2026Imph a key point that isn’t always stated is “I see no reason why it can’t keep improving itself. Since I see no reason, it will keep improving to ‘the physical limits of intelligence.” So if something is allowed by physics, then it’s allowed for ASI. That’s a dwarkesh quote 110
tweetermen.bsky.social @tweetermen.bsky.social · 11/09/2026then who can I ask if it falls into that theory building category (same for the BDS solution that’s rumored) 101
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026But then again they have predicted things better to me 000
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026We’re going to go extinct when there’s finite blow ups happening all over the place 120
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026Okay I think I follow, this very in line with Gowers problems solvers vs theorem builders construct right? 110
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026Okay I follow, thank you. And yes by narrow I meant the shape of solutions. Not that a MP problem was attention bottlenecked or low hanging fruit 000
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026Ah okay and so far big math results have been narrow example sort of successes? This math is well beyond my knowledge, which tbh makes being shook but the reste recount results annoying 100
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026Not to be a dick but is there a reason you don’t think other problems are exposed to the same approach? Also did you see Gowers talking about his nose test 100
tweetermen.bsky.social @tweetermen.bsky.social · 09/09/2026Same but after today I’m like 0 for 15 on calling things 020
