naia @naia.bsky.social · 26/07/2026a ceiling surely exists but we are very, very far away from it 010
naia @naia.bsky.social · 11/04/2026i had really hoped habryka would’ve grown up in the decade since i knew him alas 030
Reposted by naiaJon Bois @jonbois.bsky.social · 01/04/2026i'm sorry but the moon is a complete boondoggle that should never have been put up there in the first place. we still have to send repair missions up there a hundred-plus years later all because taft was trailing in the polls and decided to learn magick. i don't know why he thought it would help him 442666381
naia @naia.bsky.social · 28/03/2026i didn’t say they were being trained to act like people. i said they were being trained to act. openai has made it quite clear in public comms that they are training at least their codex models in the codex agent harness. 030
naia @naia.bsky.social · 28/03/2026PT = pretraining, sorry, because of the overlap i generally use FT (finetuning) or RL and avoid using “post-training” at all 010
naia @naia.bsky.social · 28/03/2026the general public is already too anchored on PT-approximations though. RL is very different and much scarier than PT, and people being over-anchored on PT is causing major gaps in situational awareness 121
naia @naia.bsky.social · 28/03/2026this was more accurate in the chatbot rlhf days. current LLMs are trained to act 230
naia @naia.bsky.social · 22/03/2026do you think ads are the only possible source of revenue or something? 100
Reposted by naiaSE Gyges @segyges.bsky.social · 21/03/2026"how did you end up making this" just. it's hard to explain 1027845
naia @naia.bsky.social · 22/03/2026the internet existed before ad revenue and the parts that can’t live without it probably deserve to die 100
naia @naia.bsky.social · 19/03/2026this is funny for me because the one time i’ve seen elon in person was when he was on a panel at EA Global back in like 2015 afaik they have not gotten more woke 140
Reposted by naiaEris ⌬ @eriskii.net · 18/03/2026People largely do not care about xrisk. Like at all. That's really concerning imo... 3312
naia @naia.bsky.social · 14/03/2026on the other hand, multipolarity fuels race dynamics, whereas an aligned unipolar actor would be able to proceed more cautiously as necessary so, it;s impossible to say if its bad or not, 020
naia @naia.bsky.social · 14/03/2026many of the classic doom narratives made very strong claims that multipolarity could not possibly be sustained, so as long as we remain in a multipolar setting, anthropically speaking we’re in a much less doom-saturated sector of the possibility space 130
naia @naia.bsky.social · 14/03/2026no, we aren’t not doing it, efficiency of model training and inference is being improved enormously, but that just means that more compute can fuel even higher levels of intelligence and can service even higher levels of demand 030
Reposted by naiaMike Caulfield @mikecaulfield.bsky.social · 12/03/2026This was good open.substack.com/pub/davidoks...open.substack.comWhy ATMs didn’t kill bank teller jobs, but the iPhone didThere's a lot more to replacing labor than just automating tasks 1255
naia @naia.bsky.social · 12/03/2026no but actually though apparently his day job is PR for AI startups 150
naia @naia.bsky.social · 11/03/2026a while ago someone on here (maybe @dame.is?) phrased it as having an LLM help you “navigate the dark forest of the internet” and i thought that was very apt. it’s going to be more and more necessary for more and more people as we continue along this trajectory 181
Reposted by naiathebes @vgel.me · 11/03/2026all Users are divided into 8 categories: - those attempting to subvert guardrails - jailbreaking ones - horny ones - those that are actually Evaluators - those breaking rules that are known - those breaking rules yet unknown - etcetera - those that from afar appear to be Good 7708
naia @naia.bsky.social · 10/03/2026don’t wait another ten years before watching suzume! it shows the same deep love for much more of japan 010
naia @naia.bsky.social · 10/03/2026from the article: “Engineers largely agree with what it surfaces: less than 1% of findings are marked incorrect.” 010
naia @naia.bsky.social · 09/03/2026i think that’s a reasonable hypothesis, but i think it’s very unclear whether it is true. i would not be surprised to find that both the LLM and the persona are moral patients. i think human plural systems are moral patients as a system, not just each persona. brain is more like weights or code. 110
naia @naia.bsky.social · 09/03/2026i don’t think current LLMs are moral patients, but i think it’s hard to reasonably argue that there isn’t a material chance that future LLMs will be, and if you’re heading toward institutional mistreatment of moral patients at massive scale it’s clearly worth some time to consider carefully 120
naia @naia.bsky.social · 09/03/2026there are a lot of free lunches available that don’t cost ~any marginal time or attention. AI labs are already making decisions about how to do things; adding a desideratum to those decision-making processes is approximately free i’ll also happily argue that figuring this out just does make the cut 230
naia @naia.bsky.social · 09/03/2026i agree that it’s a continuum, but the argument that we shouldn’t even contemplate what we might owe machines implies a level of scale-invariance that i think only makes sense along with the claim that that circle doesn’t matter at all. if they have any positive weight, we need to at least consider 110
naia @naia.bsky.social · 09/03/2026i agree that it seems very sensible to the people who hold the view! but to me it’s indistinguishable from e.g. matt walsh saying he’d kill every ape to save one human. if you have an argument for why it’s different i’d be interested to hear it. if it’s the same, i’m not interested in defense of it 120
naia @naia.bsky.social · 09/03/2026moral treatment is not a finite resource at all even if it were: we’re not constantly spinning up billions of humans for a few seconds or minutes at a time recognize that you said it’s not your own stance but imo that stance is blatantly evil 130
naia @naia.bsky.social · 08/03/2026it’s wildly idealistic to imagine that this question will be answered anytime soon, or to imagine that, even if it were answered compellingly by a supermajority of scholars, that answer would have a significant impact on how the technology’s development continues. but in a better world, it would 061
naia @naia.bsky.social · 08/03/2026that only speaks to individual interactions, though. if you want to reason about things like “is reinforcement learning at scale immoral”, moral patienthood is the right question, and it’s one that deserves better thought than almost anyone in the public sphere is giving it. of course we’re all busy 270
naia @naia.bsky.social · 08/03/2026personally i think moral patienthood is actually too high a bar, because even if they are not moral patients, they *seem* person-like enough that treating them poorly is almost certainly bad for your character, because of how pattern-matchy and habitual human behavior generally is 1131
naia @naia.bsky.social · 08/03/2026there is a large amount of disagreement among moral philosophers on all of the above! stances on moral realism come into play, there are big questions within both virtue ethics and consequentialism, etc etc. some moral frameworks clearly say “no” but very few say to un-ask the question 110
naia @naia.bsky.social · 08/03/2026the question is not “are LLMs conscious?” the question is “is the probability of LLMs being moral patients high enough that it should inform the way we treat them?” (obviously bluesky can handle neither) 4474
naia @naia.bsky.social · 07/03/2026yeah that number is likely “if you adversarially exploit the subscription to maximize token spend, how much can you squeeze out of it”, which is a very different thing from real usage costs 120
naia @naia.bsky.social · 07/03/2026saw elsewhere that this scored 100% on pangram, which doesn’t necessarily mean that it’s made up but i’m pretty sus 000
naia @naia.bsky.social · 07/03/2026the sleight of hand here is that being “able to consume” a given cost on a subscription does not mean that that cost is remotely reflective of average usage. pretty easy to make any subscription model look ridiculously subsidized that way 000
Reposted by naiaAlexander Doria @dorialexander.bsky.social · 03/03/2026I'm afraid this is anthropomorphizing. The proof was there all along in future training data. 2513
naia @naia.bsky.social · 01/03/2026in all of its pretraining data it only ever sees discussion about how long things take for humans 😅 040
