Everybody's Lost Their Minds
165 points by j11g
165 points by j11g
This article is really salient, and captures really well how I feel about the whole thing. It really can feel like everyone is losing their minds at times.
Ditto. And this topic comes up pretty much all the time in casual conversation, especially if people know or learn about what I do for a living. My standard response is "I don't have a polite way to talk about this with you."
I'll have to start using that one. Drives me insane when I tell people I'm a computer scientist and the first thing they ask about is AI. I know it's not their fault, but it grates away at my sanity regardless.
Previously if I didn't want to talk about my work and/or didn't want to be asked to fix somebody's broken computer, I would say I was a mathematician. Sadly, AI has ruined that response now too.
I’m (pleasantly) surprised by the evolution of anti-AI feeling on lobste.rs. There seems to be more and more a general consensus about the whole AI bubble on this place, populated by people usually very experienced with computers.
On the other hand, Hacker News seems to have taken the exact opposite direction to the point it became nearly unwatchable if you simply don’t want to hear about vibecoding practices and release of new "frontier" models.
I’m wondering if this schism is representative of something or it is just a feeling I have because I’ve hidden the "vibecode" and "ai" tags on lobste.rs
One being a community where anything unrelated to tech is off topic, and the other being the bulletin board of a major venture capital firm where their own startups are given privileged real estate, definitely leads to pretty different selective pressures. At this point I don't even check HN - if it's interesting it will make its way here and the discussion will be better.
You know how Alan Kay is always saying that computing, as a field, is a pop culture? HackerNews is US Weekly for the tech crowd. Whatever is the latest fashion, it will be hashed out there endlessly. It is also, to my mind, much more concerned with moneymaking, companies, and profits than lobsters.
On the other hand, Hacker News seems to have taken the exact opposite direction
We might be seeing an example of schismogenesis, a term I first learned from The Dawn of Everything. As an individual, I feel myself increasingly apart from the kind of discourse that now seems to dominate HN, and stopped visiting it some time this year or the last when it got unbearable. But the more averse I feel to the people who are into what's going on with LLMs, the more I might be reinforcing the kind of views that make me feel different from them.
I considered myself strongly pro ai coding, until I started reading the comments there.
LLM coding is a useful tool with many sharp edges, and it will remain that way for the foreseeable future
The slop avalanche that it enables is a very sharp edge
This. I'm all for automation in many cases, but I don't think we need to be automating the production of bullshit.
HN’s core is about VC, entrepreneurship, and finding your exit. That is the business of the underlying company. Being completely taken over by accelerating that loop is consistent with the core ethos of the site.
it became nearly unwatchable if you simply don’t want to hear about vibecoding practices and release of new "frontier" models.
I think they'd (or rather we'd) benefit greatly from a tagging system like Lobsters has. I highly prefer the speed of this site, but HN's rapidfire pace and laissez-faire posting rules works great as a "maybe you'll stumble on something interesting" list, which I'd love to browse... were it not a good 80% AI.
I always thought “you’ll get left behind if you don’t get into AI” was ridiculous, but now the communities I associate with have gone all-in on LLMs and I don’t want to follow them there, so it feels kind of true.
This week I’ve been at OSFC which I had thought of as having a certain hacker culture: rebelling against the establishment of big tech, and doing engineering for the love of the game. I was surprised by the talks espousing the virtues of LLMs.
I mean no disrespect to these speakers whose projects and talks were very interesting, I am only surprised by my misaligned expectations.
I no longer feel like I align with communities that I used to, but it also doesn’t feel like I’ve changed, more like “everyone” else has. Maybe this was always the culture and I just didn’t notice.
Maybe this was always the culture and I just didn’t notice.
"Open source" has always been an uneasy alliance between people who want to build liberatory software and people who are looking for a more effective means of enriching themselves at the expense of others. The intent behind coining the term "open source" was to de-fang the free software movement of its liberatory origins. So no, this is not really anything new, except for the fact that they've become a lot more obvious about it. Bit of a silver lining, honestly.
I no longer feel like I align with communities that I used to, but it also doesn’t feel like I’ve changed, more like “everyone” else has.
I don't feel like I even align with my coworkers who have gone all-in anymore, people I've met in person and have had deep conversations with and I know they were curious, thoughtful individuals. The tools I've used that have gone all-in on AI have all seemingly stagnated, tech culture has homogenized towards one thing, the almighty AI. I don't have anything I'm looking at on the horizon that people are building anymore, and it's not like I'm not looking for it. It just seems to have stopped existing (maybe more appropriately, it has stopped getting funding).
I've been going through something similar and it has been painful. In particular people I looked up to and considered role models no longer are, and it was incredibly jarring suddenly being cut off like this. I have to find new role models now, or find my own way against the current as long as I can.
I always thought “you’ll get left behind if you don’t get into AI” was ridiculous,
My response to that is simply: "Being left behind sounds nice, how do I do this sustainably?"
I was looking for up to date information on Data Center energy usage. The most recent thing, prior to the reference in this article, was this article from a year and a half ago:
https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/
The more recent estimate from what I can tell is a reputable source seals the deal to me. This wave of LLM technology has to go.
For anyone confused by this story’s tags: I used the “suggest” feature to try to change the tags from [ai, culture] to [ai, culture, rant], but that resulted in the tags changing to just [culture]. That’s probably because someone else had suggested [vibecoding, culture]. In that case, Lobsters uses the intersection of suggested tags. After applying tag suggestions, Lobsters considers the story to have been moderated, which hides the “suggest” link.
Maybe it’s this intersection behavior, but I find it exceedingly grating that I mute ai and vibecoding (at the cost of hiding unrelated very technical posts that fail some purity check), but somehow keep seeing articles entirely centered on “vibecoding,” as long as they are critical of it.
Whenever I find myself unavoidably in a conversation with someone who is talking about AI nonstop, I like to continually ask them: "to what end?" I discovered the potential of this strategy when I was at a party chatting with some DeepMind engineer. He was going on and on about how he would make agents to run all aspects of a company (what the company would do was never specified). I asked "to what end?" and he was, to put it mildly, annoyed by the question. Something about "the economy has to still run, people are going to need to figure out what to do with themselves." This was said by an ostensibly educated person.
I've had a similar experience talking to a journalist with math phd. I asked "even if AI greatly increases productivity, who's it doing to benefit? There's only so much service economy a finite planet can sustain" and then they completely stunned me by responding with a "nuh-uh, I think Earth is effectively infinite actually"
I wish I had the courage to broach the subject of ethics at work. I have been shocked at how many people are ready to toss all ethical considerations to the wayside for the perception of productivity. This isn't even like clothing being made by sweatshop labor; you absolutely have a choice to not use these models and directly fund their makers who openly despise workers.
I am a fan of the righteous indignation, but I am afraid that this kind of tone does not reach past the audience of haters. We have to preach past the choir to people on the fence and the other side. Whether a call to action is effective in the end, I can't say, but it's worth a shot. To that end, I've been writing my own paper. It is slow going because I have been researching all of my claims and trying to write with a kind tone, but it is incredibly difficult to not let my anger and indignation rise to the surface. The author has my sympathy.
If you actually thought your product will kill all humans, then you could, you know, like, just stop building the torment nexus. ... Alas, that wouldn't cockblock your competition...
I don't especially enjoy that I keep defending large corporations I don't agree with. But if you're going to believe their reasoning, for sake of argument, then there are plenty of reasons why they (the supposed ethical players) should be the ones to build the torment nexus. This is why they're so obsessed with alignment: their goal is to make the nexus sans torment.
And yes, I think they do see it as a race and a competition, because a less scrupled organization/country consolidating enough resources to "win" creates an existential threat.
Edit:
I also wanted to add that the "ethics aside" argument also drives me nuts. And yet I feel I can't wholly avoid using models myself. It's disappointing to me that I somehow object but not avoid.
I also wanted to add that the "ethics aside" argument also drives me nuts.
You are doing it right now in your comment, though, by excusing these companies for building the torment nexus. Like this work MUST be done. It's just not true. Life will continue on after this fallout, just like life continued on before this technology.
I don't especially enjoy that I keep defending large corporations I don't agree with.
Then.. don't? They don't need a knight in shining armor on random forums.
Then.. don't? They don't need a knight in shining armor on random forums.
And I'm not trying to be --- I added that note on to indicate that I was not commenting because I'm ideologically aligned with Big AI. I do believe in calling out feeble arguments, though, because I think they discredit a position that I largely agree with. (Edit: and they also are a distraction from the broader ethics argument, which I think is presently more important --- yes I am aware of the irony in me engaging with said distraction, but I had hoped to help us move past the whole AI doomsday discourse)
You are doing it right now in your comment, though, by excusing these companies for building the torment nexus. Like this work MUST be done.
I tried to be explicit in my comment, but what I tried to say is "IF you believe what they believe, THEN their actions are not so immediately unreasonable." I did not by any means survey the whole of rationalist literature on AI doomsday before I came to this conclusion, so if you understand it already or want to understand it and then debunk it, I would be all ears.
I don't think the companies should be excused for building the torment nexus. I do think a subtext people are missing here is that a large number of the people behind the LLM revolution we've seen are not wanton capitalists, but people who specifically have thought carefully about what sort of torment the nexus will deliver and are yet still building it. They're some flavor of rationalist and they have extensively written about their opinions and forecasts.
I want to reiterate that my whole comment is predicated on the If you actually thought your product will kill all humans part. I don't personally ascribe myself to this ideology. But if you did, then I'm saying that you can't immediately say "well of course I wouldn't build the torment nexus." It doesn't logically follow that this therefore prevents its existence. And --- again --- people have written at length on this (subject to the same precondition).
Your point made perfect sense. It's entirely reasonable to point out why someone else's reasoning is valid from their perspective, while disagreeing with the axioms or assumptions that led them there. Indeed, understanding the reasoning of someone with whom you disagree is one of the best ways to formulate counter-arguments or counter-strategies. If you write off your opponents as stupid or illogical, you'll be less-well-equipped to act against them.
I do believe in calling out feeble arguments, though, because I think they discredit a position that I largely agree with.
God, I feel this in my bones. Very little irritates me more than someone making a weak argument in favour of something I agree with.
Your point made perfect sense. It's entirely reasonable to point out why someone else's reasoning is valid from their perspective, while disagreeing with the axioms or assumptions that led them there.
Which is indeed something that rationalist-types especially enjoy doing. The irony, again, is not lost on me.
I just think it's easy to dogpile someone because your first reaction to what they say is "this cannot be." Hanlon's razor, or some adaptation thereof, applies.
understanding the reasoning of someone with whom you disagree is one of the best ways to formulate counter-arguments or counter-strategies
Or… for completeness sake here, potentially changing your perspective!
Rationalist thinking driving all those AI labs is essentially the same as evangelical christians'. It's a system of belief where helping build it is good actually, because the nexus is inevitable anyway, and when finished it shall spare those who contributed to building.
The "existential" threat is already here, it is already underway. A whole industry has arrogated for itself the right to forever limit the acquisition, practice, and proliferation of deep literacy -- which is already the single most consequential mind-augmenting technology ever devised. That is has happened at all, let alone so easily, is a moral offense to the entire human race.
This is why they're so obsessed with alignment:
They're so obsessed with alignment that they're selling versions of their AI to the military. I call bullshit.
their goal is to make the nexus sans torment.
This is impossible; the best outcome of an aligned ASI is that we end up no better than pets. It gets worse from there. With the small possibility that it decides we're either too boring to interact with or better off without it, and it fucks off forever. But we're very much not building that.
They're so obsessed with alignment that they're selling versions of their AI to the military. I call bullshit.
To be clearer than I was before: the discussion at hand is on the motivations of those involved.
In that case, if you want to call bullshit, I request that this comes with evidence. Have you read the rationalist ideology they rally behind? Can you argue --- in terms of their motivation --- why they should not sell to military? I don't doubt that you may have, but as it stands it reads as if you are jumping to the easiest conclusion.
I don't personally like engaging with their myriad manifestos and deductions, but I am at least privy to enough to believe in the motivations of many of these AI people.
Oh, and I don't think that this is wholly incompatible with some number (e.g. the business side) being ruthless capitalists who just want to extract as much capital as they can. Perhaps, if you want to go to conspiracy theories, they've made up or exploited this whole AI Doomsday bit to motivate scores of smart people. But that doesn't change the motivations of the many who do believe in it.
To be clearer than I was before: the discussion at hand is on the motivations of those involved.
You're right. I shouldn't assume that an alignment model is one which refuses to kill people.
the best outcome of an aligned ASI is that we end up no better than pets
I technically agree with this, but the phrase “no better than” implies a bad outcome, which I disagree with. If an aligned artificial superintelligence were created, and it really had people’s best interests at heart, I think that being its “pet” would be a much happier life than living in today’s society. One good outcome I can imagine (though I’m sure some sci-fi authors have imagined better ones) is a society where people don’t have to fear homelessness or starvation. In that outcome, the ASI that runs society could use its surveillance to identify people who would be compatible in friendship or romance and then arrange for those people to meet. It could recommend media to people based on what it knows about them, and coordinate events such as book club meetings or sports games. It could choose to hold back from publishing its own solutions for math problems so that humans who enjoy mathematics could still have fun doing it.
If you’re imagining that the best outcome of aligned ASI results in being trapped in some kind of human reservation and feeling restless because you perceive no agency over your life, then I think that hypothetical ASI is either less aligned or less intelligent than in the true best outcome.
I do agree that, sadly, the AI industry is not on track to build an humanity-aligned ASI.
One good outcome I can imagine […] is a society where people don’t have to fear homelessness or starvation.
You don't need ASI for that, just progressive social policies.
I think your description of a benevolent ASI is pretty much what human cultures already are (and have been for thousands of years). We get conditioned into whatever culture we are born in. Successful cultures try to accept and cater to human needs, but they still condition people into their own logic and demands. And so called individualistic cultures are especially successful at making themselves invisible and making humans believe they are individually free.
I don't see how an ASI would be any different from that (except maybe having more conscious awareness of this dynamics).
So why not heroin? It still makes you feel good at the cost of everything else, but not everyone else. The downside is mainly imposed on you and your local community, and not on all of humanity.
If all you want to do is watch Netflix for a while and die, we have solutions today.
You know what the worst part is? It's knowing that when the bubble pops, everyone will be even more fucked, and still wishing it would just fucking POP.
AI is the tool that dulls its users; it incrementally replaces understanding with a new dependency and addiction as you actively de-skill yourself.
Aha, so the way AI gets smarter than humans is that humans get dumber, I see the plan now.
I smiled when I read this one. It's a clever observation (not sure I agree with it necessarily)