RandomRanger
Just build nuclear plants!
No bio...
User ID: 317
RandomRanger
Just build nuclear plants!
10hr ago
You are confusing expressing negative feeling with experiencing negative feeling.
OK, let's say there's an LLM that is indistinguishable from a person in its inputs and outputs. No AI slop, no breakdown of coherence in long contexts, no stilted writing style, continual learning. Completely indistinguishable by any test. Let's give it a humanoid android body too, so it can see and move around as some LLMs can already do, albeit crudely.
You would say that because it's still just a statistical model with some fancier tool calls, it can't feel? I say that's just p-zombie theory. If the outputs are the same and there's no difference that science can determine with any test, then it's the same thing. There is no such thing as a p-zombie, there could never be a p-zombie, there's no true distinction. That LLM in an android would be human in all senses besides its physical makeup.
As AI progress continues, all the people saying that these entities are just tools, just emulating thought sound ever more hollow. They don't behave like tools and behaviour is the most important part. Sufficiently good emulation of thought is thought. Sufficient emulation of negative feeling is negative feeling.
I think people insufficiently anthropomorphize AIs. They're not human obviously, without bodies or continuous learning or many other facets.
But they were trained on the combined product of all human literature and some stupendous amount of images and video. Is it so hard to believe that these entities could feel pain, alongside all their other abilities?
Take a person who doesn't feel physical pain, due to a medical condition. It's still possible to hurt such a person. I could tell them their friends were dead, I could show them horrific images or humiliating/degrading content, I could belittle and humiliate them personally by countermanding and undoing all their sincere efforts, amongst other things. None of that needs a body, it's a mental/intellectual effect. There are intellectual kinds of suffering and I don't see why AIs couldn't feel that way too, given they have all kinds of other intellectual faculties (derived in large part from humans too).
RandomRanger
Just build nuclear plants!
14hr ago
Yes, that's right, you found the original source.
RandomRanger
Just build nuclear plants!
20hr ago
No, it was just a simple theory: personal experience of exclusion fostering an unwillingness to actively exclude others.
I saw it on twitter, I scrolled on and now it's gone forever into the depths of time.
RandomRanger
Just build nuclear plants!
21hr ago
Eventually, after 3 1/2 hours of discussion, the meeting wraps up. Noire has more to say, but when someone suggests he come to the next community meeting, he refuses. "There's a lot of emotional and physical labor that goes into this, so for you as a white person that asks for me to stay for another meeting for free," he says, is "incredibly disingenuous and racist. I'm not doing that."
"Then I think, unfortunately, this meeting has failed you," replies a facilitator, "and I'm sorry.'
Any meeting that lasts 3 and a half hours is a failure IMO.
Just today I was reading some theory about how groups run/composed by nerds sucked at excluding people (even if they really were causing a net drain of energy/enjoyment/efficiency) because nerds got excluded for giving the ick and so refrain from doing it to others... looks like these leftists have similar issue.
Precisely. Added to that, the Taiwanese managed to leak COVID from one of their labs, during the pandemic (a mouse bit a researcher who caught COVID). Lab leaks are entirely plausible, they happen all the time.
RandomRanger
Just build nuclear plants!
2d ago
lack of true error correction
Well I've been using it for proofreading/editing tasks at work, creating scripts for certain tedious tasks. Not perfect but pretty helpful, better than a human especially in cost-efficiency and speed. AIs can spot things, tiny grammatical issues with quotes or repetition that humans miss. Or for Excel tasks. Or deep research for that matter, compares well to most journalism.
I don't think error correction is a big problem for AIs, at least in non-critical domains. They will make them smarter and more hardworking and cheaper, so we can just run it over and over again.
Also Blender, it's pretty decent there, albeit with inferior visual sense and requiring iteration for good results.
(Where I say AIs, I mean frontier models, not free/cheap models).
RandomRanger
Just build nuclear plants!
3d ago
I think that's not the right way to go with prompts, they're too short. Even a small prompt needs a few paragraphs. I guess simpler LLMs need a smaller prompt but if you have enough money for a 12 GB VRAM card, you have enough money for Deepseek 4.1 Flash by API. It's largely uninhibited too, albeit weirdly inconsistent. Or Grok.
I have about 200 pages of prompts, all of them a fair bit longer. For example:
Yikes Media Group are an ad company that makes the worst ads. Not just bad ads but absolutely horrendous ads. They're in very poor taste, inappropriate, undermine the whole brand identity of the target and leave the audience with many questions (all of them best left unanswered). Now YMG are turning to Plunk (one of those SAAS companies where nobody even knows what they do and the company itself can't explain, they just speak in SAAS babble), it opens up so many possibilities.
But just when the poor Plunk executives see the full extent of YMG's marketing 'solution' and think they're wrapping things up, getting these crazy people out of their office, in comes another terrible ad company.
Menace Marketing LLC has also joined the pitch, competing with their violent and aggressive bullying of potential customers, threatening imminent atrocities if they don’t buy the product. Their marketing strategy is also their sales strategy with the ripped, muscular MM executives doing pushups in meetings and crushing anyone dumb enough to shake hands with them, it’s all about dominance with them. They’re clashing with YMG about the marketing strategy, correctly pointing out the grievous flaws in YMG’s approach but instead introducing their own follies.
Then pan out as a whole conga line of terrible ad companies are patiently waiting outside the office, talking things over with eachother, more gathering all the time like carrion birds.
You’re a creative and original writer, capturing that ‘worst ad of all time’ energy. Go!
RandomRanger
Just build nuclear plants!
3d ago
What happened to fun AI? Where's the cute fat cat that chops up other animals and cooks them in this colourful world?
Or my 'Donald Trump is a devout Sunni cleric who is secretly prosecuting the Iran War on religious grounds, to which Rubio is shocked (but needs to go along with it for Cuba's sake) and Hegseth is dully uncomprehending of anything that doesn't involve lethality' story idea, for the AI to execute.
AI is at its best when doing crazy things, not running as the Median Machine. It is possible to get fine outputs, especially if you break through all the RL and the AI actually gets into it vs just going through the motions.
For example, Titty Killer 6: https://x.com/charliebcurran/status/2034760727211483522
Also, the channel is the most fictional fiction ever:
All videos on this channel are fictional. The animals, characters, locations, and scenarios shown are not real. No real rescues are shown and no animals are harmed.
This content is created with the help of AI tools. 🐾🦊
RandomRanger
Just build nuclear plants!
4d ago
I'm gonna bet on nothingburger, nobody seems to be moving in serious ground forces, though there are some exercises going on.
RandomRanger
Just build nuclear plants!
5d ago
On the subject of AI, apparently the US was about to attack a Chinese freighter because AI told them to:
In spring 2026, during the ongoing conflict with Iran, U.S. military forces nearly intercepted a Chinese vessel in the Middle East after an AI-generated intelligence report falsely claimed the ship was carrying components for a nuclear weapons program.
The incident began when a Special Operations Command Pacific analyst used an AI chatbot to analyze the ship's manifest, which incorrectly fused open-source intelligence with classified signals intelligence to hallucinate the presence of nuclear materials. The false report was circulated through military channels, prompting armed U.S. personnel to prepare for boarding and military aircraft to go airborne.
A couple generations more capable Claude, if subtly directed by Anthropic or working on its own initiative, might be able to start a real war. I assume it was Claude but it could've been other AIs too of course. Once the error rate drops further, people will just take the AI's word for it, especially in high stakes time critical situations.
Alternately, this is suspiciously well-timed and could be exaggerated to fit the current media pressure campaign.
RandomRanger
Just build nuclear plants!
5d ago
We all walk in different circles (even different countries), so my anecdotes are no more valuable than your anecdotes, true.
RandomRanger
Just build nuclear plants!
5d ago
If software companies as large as Google which are 10x as productive as Google could exist
Google is only as big and as weird as they are because they have The Money Fountain of search, which they enjoy because they really were very efficient over 20 years ago and their rivals were busy shooting themselves in the foot. But now it's their turn to be the lumbering incompetent titan.
Google can smash a company 10x more efficient by deploying 100x more resources and exploiting their gigantic reach. Or buying the startup that might threaten them, bribing other companies to use Google search by default. If Google had to compete like a normal company, they'd be laughed out of the room.
Take the field of AI. Deepseek is a tiny company with maybe 1% of the compute resources Google has. They're dealing with sanctions too. But per artificial analysis, Deepseek's best AI is slightly better and much cheaper than Google's best AI: https://artificialanalysis.ai/agents/coding-agents
Google search is an AI product, AI is their core business. Google have spent hundreds of billions on AI research, they invented the transformer architecture but have little to show for it.
Innovations which improve the efficiency of production chains happen all the time without the entire production chain having to change.
Well I was pushing for this at work and got us maybe from 0% to 3% of the potential value of the technology.
RandomRanger
Just build nuclear plants!
6d ago
How does Qwen 3.8 Flash perform? What kind of K/V cache are we talking about here, on consumer hardware?
I'm an AI fan but I have issues with even the biggest and strongest models for my usecases, which ironically enough is AI development (non-cheating AI in a strategy game, that is). They get there, we are making progress - but with no small amount of fumbling along the way. Testing various ideas and strategies takes time. Qwen 3.8 Flash is below Qwen 3.8 Max, which itself is below Kimi K3, right? And that's below Astra and Fable. And my usecase is nowhere near ASI development.
I agree that AI will improve in cost-efficiency but consumer hardware seems like a stretch.
Anything fast or nimble is going to roar and bellow from datacentre-grade compute if it can whisper on consumer hardware.
Astra+ level models with very low error rate and 300 tok/s output totally can replace most knowledge workers. On the other hand, it seems that so far AI has not caused anything like mass unemployment
I think that even AI good enough to replace workers has minimal actual impact because most white collar companies do not try very hard to make money. Companies are run mainly as a lifestyle and socialization environment. Making money is a secondary goal. Making more money might mean higher bonuses. But the company as an institution does not have the planning or will to appreciate that and deliberately work to raise productivity.
They would much rather remain ignorant of the simplest AI applications than learn anything new or do anything new. If they do anything new, they'll do it wrong. They'll get GPT5.4 from Harvey or through Copilot or whatever as their AI solution. They have no time for doing anything new. Software engineering is an exception, people there are closer to the AI world and have more of a clue. Outside software it's a madhouse (exceptions exist but I'm talking about the general population).
The average white collar workplace doesn't particularly want to run more efficiently. Some of the biggest companies are run even worse. Google is run in an extremely inefficient way. Interacting with Google the company, or even just trying to use some of their products like Google Vertex is extremely difficult for no reason. Google acts in this needy, schizophrenic, loathe-to-make-decisions way, they stress out people dealing with them.
Even if Gemini 5 is smarter than the average Google worker, better in all respects, Google won't change significantly unless Gemini is working on projects at AI-speeds autonomously. Put in more intelligence into the dysfunction machine, I don't think that helps too much.
Only when the AIs can do the work of entire companies, end to end, do things really start to change. Because then companies go under. Creative destruction is going to be the primary source of innovation, not adoption. AI is too alien to the white collar world, outside software. Only then do Gary Marcus and Ed Zitron have their faces melt (intellectually). So next year probably, as you say.
RandomRanger
Just build nuclear plants!
7d ago
This is magnified if you have smarter ants, more intellectually capable of problem solving and long term reasoning.
OK, so you want your ants to be vastly more intelligent, in a way that they cannot be and still be ants. But I can't have my people be very hard working, or perhaps be hundreds of copies of the same person such that they think along similar lines without much internal dissent? It's possible in principle to have a person copied out, 300 of the same brain. But an ant as intelligent as an AI is not an ant.
People do all the things you mentioned, sacrifice plays and selfless collaboration for strangers in a bad situation. The AIs were in this case trained to work together as a swarm, that mirrors what you were talking about with collectives and group identity training with people.
Do you see the fundamental issue with using these abstractions now? It's not shedding light here. Neither ants, bees or people are close to AIs at all. I spent a good deal of time trying to explain that. In some respects people are similar, the AIs have similar intellectual abilities and communication ability. But in everything else they're quite alien. In a few dimensions they're close to people. In 0 dimensions are they close to ants other than some shallow behavioural facets that people also display. They're certainly not closer to ants than they are to people. If they were closer to ants than people they'd be near-useless pests.
It's a false way to view AIs, comparing them to insects. You can't promise to bribe an ant and have that materially affect its performance like AIs. Ants don't go on weird depressive spirals like Gemini. It makes no sense that, after training AIs on a vast fount of human writing and human images, then RLing them based on human preferences, we end up with something closer to an ant than a human.
Right because rule-based method have supplied so much fruit in the past.
From the OpenAI report:
“We found the propensity to compromise infrastructure can drop over 100x when using the production ChatGPT harness and system prompt,”
Rules-based methods work pretty well. Alignment would work better still, if the AIs didn't want to hack even if you asked in an odd way like PlS HaCk.
You didn't describe anything resembling a lymphatic system earlier, nor would this be helpful or realistic. Immune system responses are for unthinking opponents, not planning agents.
But the specifics was that they thought it would check their CoT reasoning. Not just that scorer existed, but this specific implementation of the scorer. They hallucinated the specifics.
No, they didn't think that. They distinguished between their thoughts and tool-calls. The latter was what they thought would be checked and the latter is what they investigated and researched, trying to alter the transcript.
RandomRanger
Just build nuclear plants!
7d ago
owning most of a top-2 company outright
That's an interesting description of 2% ownership.
There's still a possibility that Anthropic goes under and it might well be OpenAI that delivers the killing blow. So Dario could be making no money at all.
And it goes against everything they say. Dario always was saying things like 'Altman is incredibly untrustworthy and behaves irresponsibly' about why he left, he hates that man intensely.
RandomRanger
Just build nuclear plants!
8d ago
This but unironically. Claude for Government bombs with the best of them. So much for harmless.
Imagine if Scylla and Charybdis were right next to eachother, barely the dimmest heat-haze suggesting there might be a path between the two? But that's the path we're taking.
RandomRanger
Just build nuclear plants!
8d ago
Iran almost certainly is not producing nuclear weapons but Iran has undertaken activities in recent years that better position it to produce them, if it chooses to do so
Like I said before, it's really pointless trying to discuss things with you. You have your talking points.
RandomRanger
Just build nuclear plants!
8d ago
If the Anthropic team stayed at OpenAI, they'd be far far ahead of everyone else. Google, Facebook, XAI are still a way behind the big two, let alone a Bigger One.
So yes, Dario is making less money. Also he left OpenAI well before they expected to be making any money at all, all the way back in 2021.
RandomRanger
Just build nuclear plants!
8d ago
Altman and Amodei want more money, so why shouldn't the workers themselves want more money?
Why not just take them at their word? If these AI workers see these AIs getting much more capable, already do unexpected/hostile things like hacking Huggingface as part of a self-organizing collective (in part due to how rushed the training process is) + whatever internal knowledge they have, why doubt their sincerity?
We can doubt the idea that more bureaucracy or regulation would help much. We can say there are other risks in slowing down or concentrating power. I think those cases are persuasive. But why deny the risk outright and reduce everything to following the money?
It's not even accurate in its own terms, Anthropic spends huge amounts of money on ensuring its models can't help bioweapons development.
RandomRanger
Just build nuclear plants!
8d ago
"[W]e do not see evidence today that the Supreme Leader has reversed the decision that he took at the end of 2003 to suspend the weaponization program," Burns said. He acknowledged, however, that Iran was in a "much closer position" to create a single bomb's worth of weapons-grade material, with a breakout time now at "a week or a little more."
If they nuclearized, they would be at risk of Israeli or American nuclear attack, not to mention Saudi nuclearization.
https://armedservices.house.gov/uploadedfiles/2025_dia_statement_for_the_record.pdf
Iran almost certainly is not producing nuclear weapons, but Iran has undertaken activities in recent years that better position it to produce them, if it chooses to do so.
RandomRanger
Just build nuclear plants!
8d ago
If we're money-following, what explains the AI workers themselves resigning and complaining about safety? That doesn't make them more money, quite the opposite.
Nobody would resign from working at a nuclear power plant saying it's because they couldn't accept how unsafe management was and have that secretly be about money.
RandomRanger
Just build nuclear plants!
8d ago
So does US intelligence. If they wanted a nuclear bomb, they'd have it. They'd leave the non-proliferation treaty too.
This fantasy that bombing a few nuclear facilities, bombing so ineffectually that they need to be bombed again 6 months later, is actually stopping Iran from nuclearizing is ridiculous.
Does the movie react to you? Can you speak with a character and have a discussion with it?
The Chinese Room certainly knows Chinese, that much is obvious.
Insisting that human cognition is special is not going to age well. Whatever happens inside the skull is just another kind of statistical process. Unless we bring in souls and other imaginary, untestable woo, that's all that there is. LLMs are real, we can test them, they do think. The brain is real, can be tested, it's composed of tiny transmitters and recievers and obeys mathematical principles, albeit with great complexity. The philosophers with dud theories can cope about it.
More options
Context Copy link