https://x.com/PirateWires/status/1950212890184864106

"AI psychosis" from talking to LLMs is not a real thing. people who succumb to this are just stupid/mentally ill in the first place. this is self-evident but whatever, good to see The News reporting this correctly

https://x.com/tokenbender/status/1950215097147936800

the "AI psychosis" forcememe is definitely the new "angle" for deranged AI safetyists, though. i'm seeing a lot of kvetching about it

https://x.com/uberboyo/status/1948646282819514674

do LLMs "prove" that plato was cr about Forms? i pretty much agree with this guy's response on the matter: https://x.com/curtdoolittle/status/1948802577854070842

https://x.com/rohanpaul_ai/status/1948572304809611701

these guys think they have figured out the mechanism via which LLMs can "learn" in individual contexts during inference, without updating their actual weights via new gradient descent. i think people have already intuitively understood something like this to be happening, but this presents a formal hypothesis for how it works

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140651)

like i don't even know what people mean anymore when they say "china is mostly hot air" (not you, other people). obviously this is complete nonsense and china is objectively a very consequential and meaningful actor in the world. but you can point out that china also has some specific, serious problems that will affect things going forward. but nobody ever does any actual analysis, they just say "it's over, china already won" or "it's over, china is a paper tiger." it's like when people say that america is "done here" (this is similarly obviously complete nonsense - america is still the most powerful and influential society and culture in the world by far, and will continue to be for some time)

anyways i'm rambling but to answer your question: i think that china will continue to be more and more technologically, economically, and even culturally relevant going forward. i think it's clear that they will eventually catch up in AI production and inference compute capacity and then they will go closed-source. what i don't know is how much that will actually matter in practice. i don't think americans or american firms will be willing to use chinese AI over american AI services. even though it will be cheaper. the USA fedgov may even legally restrict americans from using it (i think this would be cr). i think that best case scenario from our POV is that the competition from chinese firms drives down the price of AI services in the domestic US

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140677)

in general this is a huge problem in the current year for everything. people think that because they see pictures and videos and "hot takes" from people online about a place or thing, they're "informed" about it and have an educated opinion on it. hundreds of millions of midwits sneering at each other about shit they have literally zero firsthand experience with. i'm sure LLMs will make it even worse. it's all so tiresome

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140682)

What is on the event horizon that will be a "big deal" to doubters who say "its just text auto-complete and nbd"

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140679)

this should be huge. no one cares.

so as youth, esp male youth, get eviscerated by AI, there wont be any noise about it.

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140891)

society will slowly adopt the technology like any other technology. it will take time but it will also move fairly quickly. humans will keep losing jobs in every industry, some worse than others. people will get poorer. there will be more calls for redistribution

i want to be clear that AI is definitely going to have an overall negative effect on people's lives. people WILL get poorer and the benefits WILL go to the capital class and to those in the AI industry and its inputs (energy, hardware, etc) (this is why i've professionally pivoted to AI). there is currently a fiction, being spun in no small part by those in the AI industry and their shills, that AI is a "rising tide that will lift all boats." this is false. AI is going to make society worse for normal people, unquestionably

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140715)

as wealth disparity increases, industries that cater to rich people will expand too. anything that requires personal human relationships and human judgment and physical presence or touch. custom carpentry, for example. custom welding. gunsmithing. hunting and fishing guides. high-end escorts and gigolos. not even flaming about these btw

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141122)

3 is where this clearly comes apart and starts to strongly resemble the Stephen Jay Gould style delusions about general intelligence measures. these people want to have a vibes based argument.

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141837)

https://x.com/jd_pressman/status/1950381762040983808

i think he's being a little bit hyperbolic but i don't think he's being unfair or uncharitable. the fact is that the tech that we're seeing on the ground right now is nowhere near what people have been expecting will emerge. that isn't to say that it won't continue to slowly progress, but things have slowed down a lot, and it's not actually clear what the next transformative technical step could be. there are candidates for more incremental LLM improvements, but nothing potentially transformative except non-LLM techs which are currently still in early stages of development

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141897)

https://www.theverge.com/news/713879/it-looks-like-microsoft-is-getting-ready-for-gpt-5?utm_source=listedaidaily.beehiiv.com&utm_medium=newsletter&utm_campaign=breaking-down-trump-s-big-gift-to-the-ai-industry&_bhlid=2cd519cef77e3139be0e6b4f12a1deb04dbb6fdb

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49140755)

https://www.youtube.com/watch?v=33TkQ4ZCTww

these people say they invented a new technique that has LLMs use natural language "chain of thought" to prompt engineer new, better versions of a prompt from a given prompt input, resulting in superior output results. they've tested it against reinforcement learning training techniques that actually change weights and claim that it performs better. which is pretty crazy if true. i mean all this does is you feed it a prompt and then the AI optimizes the prompt and you're getting a bigger performance boost than RLHF. you'd be saving soooo much money

if this is legit i'm sure we'll see it widely implemented soon. i'm always so blown away by these clever little ideas that researchers come up with

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141237)

AI is just really cheap labor that's all

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141690)

this isn't a "gotcha" non-question im legitmately asking what you think

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141694)

AI is definitely going to make life worse for most people. people who claim otherwise are dishonest or coping or stupid or some combination of those. i am excited about AI because in spite of this, it will give high human capital people an enormous force multiplier and much more sociopolitical leverage and agency. this is a very good thing, and the good will outweigh the bad imo

also, any large structural change to our society is good, because we're otherwise staring down a road of certain gradual failure and destruction. there's no other possible mechanism by which to fix our society. whereas it's very possible that someone(s) could leverage AI to seize absolute or near-absolute political power and actually be able to make real, substantial changes that could save western civilization

as an aside, you misunderstand my beliefs as "libertarian" based on that one random thread. i'm concerned about human biology rather than the human economy, which i don't think is very important at all. humans becoming post-malthusian has resulted in us literally de-evolving to an inferior, more primitive form of life. i'm not talking about mass immigration or changes in racial demographics. within people of european descent, we are already much inferior biological specimens to our ancestors. not only this, but westerners have ceased to even care about their own existence and future without malthusian pressures and intra-species martial conflicts to give their human condition meaning. it's clear that bread and circuses can and will never be enough

this is a big problem - probably the biggest problem that the species faces. there's no solution that doesn't involve a lot of people suffering and a lot of people dying. this is why i'm so callous and dismissive of stuff like "give poor retarded people welfare" or "the minimum wage" (btw raising the minimum wage just means you have to give people who are priced out of the labor market welfare, or else they die - you didn't seem to acknowledge this fact in the other thread). when you start from the above recognition of the current state of the human species, this stuff is just obvious and uninteresting and not worth talking about because people are only ever going to respond with "but that's mean, you're mean" so discussion of it is unconstructive and pointless

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141753)

i think we've probably talked past eachother a lot because my thinking is focused primarily on today and in the immediate future and you seem more concerned about the very long term trajectory of humanity. i think thinking your way is misguided because i dont think we can anticipate the scope youre conceiving nor do i think humans that exist right now care much i. i think focusing on todays issues and stewarding things to give the next generations the best tools is better than thinking further down the line.

you interpret it is giving them welfare i interpret it as ensuring a fair system that works for the purpose its intended: humans to flourish. your thinking here is in line with your grander thoughts about humans trajectory but i think that misses the trees for the forest. you arent considering that the purpose of the economy is humans thriving, not that humans are merely grist for the economic mill to produce... what? your arguments here are identical to every other labor law debate. in the absence of government most people are slaves to a despot. so the question isnt really "minimum wage?" but "balance between free markets and social welfare?"

you packed a lot in there and dont think i didnt consider something even if i didn't give it much engagement. i, uh, have a problem with succinctness.

so you would have ai "do it's thing" with little regard for the immediate impact in the hopes that what comes out at the end is better for humanitys long term future?

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141797)

in the face of these otherwise overwhelming threats, AI seems to me like an unexpected gift that has materialized when we need it most. like the sim creators realized that we're actually in danger of losing the game for real and so they threw us a deus ex machina. we're still facing a massively uphill battle, but the odds of us getting through it are now like 10% instead of 1% imo

it's not that i don't care about people in the present day, it's just that a bunch of people getting somewhat materially poorer for some limited period of time just doesn't matter much, imo, compared to the prospect of the human story ending. and if we end up losing regardless, then it didn't matter anyway

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141890)

putting aside the potential of AI and the value of your goal, how do you expect to accomplish this? this is what i meant by utopian/unrealistic goals previously. i see this type of thinking a lot in Engineers where their ideology and goals are structured as if life is a game of Civilization.

the crux of the issue is this: it is not possible to orient people towards the goal you see as necessary nor is it possible to ensure continuity across several generations being that we're mortal and future generations will do what they do.

i'm interested thoughts on that aren't based on the value of the goal (i.e. stressing the importance of your goal for humanity's sake doesn't move the needle on getting others on board today unless you can explain how). how is your conception of what we "should" work towards not sci-fi naval gazing?

(http://www.autoadmit.com/thread.php?thread_id=5756020&forum_id=2).#49141987)