OpenAI Employees Warn of Advanced AI Dangers
macrumors.com
macrumors.com
Do they WANT to be working on something capable of 'human extinction'? Maybe I'm cynical, but I strongly disagree that anything OpenAI builds on the same trajectory as GPTx is going to end the world.
The dangers aren't AI Terminators; they're "Her" and emotional manipulation and long term psychological problems for _everyone_ because most humans simply aren't complicated and are easily manipulated.
That's how capitalism works so well and now capitalism has AI fentanyl.
I model AI as a 10x speed up of the kind of changes we saw in the industrial revolution. Before then, we might be concerned about coffee or alcohol, today we are concerned about designer drugs. Before, cavalry and spies, today tanks and laser microphones and satellites. Before it was evil kings, now they have competition from evil megacorporations.
So sure, Terminator robots… made by the next Jim Jones. Westworld, brought to you by a half-assed startup branding itself as "Uber for Disneyworld" (or, given what happened with their self driving cars, actual Uber). Colossus the Forbin Project, brought to you by the radical outsiders who left Greenpeace. And so on.
But also every other thing, including all things that have already been shown by corporations perusing profits above people, dictatorships in general, and every simple mistake made by a well meaning corporation or democracy or software development team.
If we're lucky, it will be incompetent because the script writers have no idea how real life works.
If we are moderately lucky, such an AI will do this in a way that allows one random underdog to push a surprising oversized "stop" button with seconds to spare despite having previously demonstrated overwhelming force against a major nation's entire armed forces.
If we're extremely unlucky, it will be acting out a horror film with great competence.
If I had to bet on one of those three, I'd pick the first option — most scripts are not written by domain experts — but my main expectation outside this is "more of the same stuff we already see with corporations, but faster, and government regulation will continue to be 20 years behind the tech just like it is with the internet in general".
What does "end the world" mean here? Is the concern truly ending the physical world on earth, killing us all? Or is it ending the world as we know it, like ending our current civilizations and maybe even ending the model of humans being the leading species on earth.
With regards to OpenAI and GPTx, the question there is whether we believe we publicly know everything OpenAI software is capable of today and what capabilities they are actively trying to develop. We may think GPT isn't a serious risk, but that could be wildly wrong if OpenAI has much more powerful systems either under lock and key or behind the black box of GPT's public API.
Would you argue that its impossible for OpenAI or a similar company to accidentally develop something dangerous while they believe they are developing a safe, controlled but potentially much more powerful tool?
until then, they have a free pass to get away with such scare-mongering bs.
I don't have to be in a car crash or get shot in the head to know this is a bad thing, and nobody sane is going to bother causing the end of the world just to convince you it's possible.
What, precisely, does that even mean?
Critics (I initially said "you", but rereading this is ambiguous) clearly don't accept anything that currently exists as such a demonstration: not the models which are superhuman at strategy games; not the automation actually used by the real militaries despite dangerous flaws (whose bugs have resulted in NATO early warning systems being triggered by the moon and Soviet ones by the sun, or planes nose-diving because of numerical underflow); not the use of LLMs to automate propaganda; not Cambrige Analytica; not the lack of controls that resulted in the UN determining that Facebook bore some responsibility for the (ongoing) genocide in Myanmar; not the examples given in the safety report on GPT-4 prior to release showing how it was totally willing to explain how to make chemical weapons; not the report the other year where a drug safety system was turned into a chemical weapon discovery tool by deliberately flipping the sign of the reward function; and not the OpenAI report on maximal misalignment in their own models caused by flipping the sign of a reward function by accident.
What is the smallest "small scale" demonstration that people who currently laugh at the idea of the possibility of a problem, won't ignore?
I don't understand, are you suggesting flipping the reward function of reproductive fitness itself, in vivo, of DNA/RNA?
And how is "protect" supposed to demonstrate danger? That's like saying "ACAB protestors are dumb, I'll only believe the police are evil when they catch a gunman"?
The right approach to take with AI is to develop it as fast and as open as possible, the safety issues will be polished as it happens, the more eyes on it the better.
If it's becoming insufferable it's because nobody is taking it seriously.
2023: "Why isn't this AI research company that says the models are potentially dangerous in the wrong hands, not letting me download the models?"
OpenAI can switch it off… in theory. In theory, the Soviet Union could have "just switched off" Chrenobyl. In practice, the emergency off button made it explode, and other reactors on the same site had incidents both 5 years before and 5 years after the famous one.
None of the downloadable models can be "switched off" remotely, any more than we can force people to stop using buggy software.
And you would need to have that capability — was it called EvilGPT or ChaosGPT, the one explicitly tasked with trying to destroy the world? Because people absolutely will keep pushing that particular button until it actually works: some because, like you, they don't believe it ever will and wish to mock the idea; others because they are actively misanthropic.
I don't think it's something that's bound to happen, but I do think it's a real concern - and I'm not personally worried about an I, Robot or Matrix scenario where AI decides to take control of humans for whatever reason, but by the idea of it becoming easier for a small number of immoral people to control vast power without having to persuade armed people to do what they want.
For example if the AI / robotics arms race sees one major military power take a significant enough lead, it could see them become temporarily invincible (if it includes AI-based missile defence which allows them to secure the entire country's skies without needing millions of people to do so) and they might decide to attack others to try to keep it that way.
If, before any other country had caught up enough, the leading country were to have either a single dictator, or a political party in power, who wants to act like the empire builders of several centuries ago did... maybe what hasn't been possible for many years now might become possible again, and we could see a country like the US or China decide it would be better if the whole world were controlled by them, and any country that didn't volunteer to become a state under their power would be forced into occupation by swarms of AI-controlled fighter drones.
Without needing to accuse any countries of currently being run by people who would choose to colonise the world if they had the choice, it's clearly not unprecedented for people to make that choice even when knowing it would cause many deaths on both sides - this time it might be almost entirely limited to deaths on the other side, and potentially even not many of those if everyone learns quickly that there's no point resisting against the invincible robot army?
The crazy thing is, if you take these claims seriously, then every AI company in America should be subject to ITAR [1].
[1] https://en.wikipedia.org/wiki/International_Traffic_in_Arms_...
* Imagine AI giving a terrorist network a recipe for the most toxic nerve gas ever discovered. (This has already happened to AI researches)
* Imagine AI being used for deepfake propaganda inciting a war between superpowers. (This is arguably already in progress)
* or it could be the usual sci-fi classic of an AI intelligence becomes superior to humans and just takes over (using the above methods and more).
AI has the ability to be completely undetectable and incredibly insidious. We could be destroyed by a force we don't even notice.
- increasingly censored, controlled, curated gate-kept providers - an inherent lack of trust in public facing information - outright bans on information access
So ushering llms could be an undo the rate of societal and economic progress, create conditions for widespread political and social unrest, and create vacuums for authoritarian leaders and information brokers to consolidate control and power
You can translate the word "oversight" as "help us hammer down our potential competition with convenient regulations".
No the half baked shit-show of GPT is not anywhere near a threat to human society and much less so being close to AGI. It barely works well at its most basic "serious" tasks and so far, after a bit of back and forth about its usefulness, has been shown to be mostly good for helping at moderate organizational tasks, and SEO content sludge of a kind that used to be handled by third world content farm writers.
Even for serious content writing that needs to be more or less correct, many people refuse to use it. For programming, likewise. And this is going to be an "advanced AI danger"? Nope.
On a partial side note, "misinformation" and manipulation been with us since long before AI or the internet. If anything both were previously worse and more difficult to fight than today when free access to so much information also at least lets good sources be just as accessible as the garbage.
And much of the worst misinformation throughout history has come from the very governments and large businesses that are now harping again about its dangers, with a sidelong look towards how they can again better control what people say, hear, watch or read.
They play to the choir about the risk of human extinction. (Sure, whatever, WWDC is coming up and now you can pitch your employee activism as help with fundraising.) But altogether, it's a glorified town hall. A letter of employee grievances aired in public because apparently that's how HR decisions are made now.
Employees should have protected ways of disclosing unsafe practices, particularly when the potential impacts are so large.
That said, the technology is very likely to make the world a shittier place, and it's not due the technology, it's due to the people and grifters around it.
Actual, I want the entirety of the “AI revolution” to fail.
These technologies promising productivity improvements come, yet I’m still here waiting to get my slice of the pie. Who’s eating my slice??
I’m now more productive than I was last year, considerably so. No pay rise! I’m still working the same hours. So are all my colleagues. Stress is at an all time high. Expectations have increased.
WHO IS EATING MY PIE!
you might as well tell the luddites weaving textile by hand that the steam engine will just make them more profitable, seeing as how they own their own business and all
On the image side of things I’ve been using stable diffusion to do style transfer animations and turning sounds into images.
For me it’s been both productive and fun.
The luddites were wrong and got butthurt, if everyone had their mentality we would still be riding horses and using abacuses.
The order of the threats is amusing. “Our image creator doesn’t represent the nuanced spectrum of genders! And it sometime says things favorable about Trump. And, oh, I almost forgot… it may cause our extinction.”
This can also refer to wealth inequality. Especially between people who work for a living vs people who own assets for a living.
And it's worth putting that first, because it's arguably most likely to actually happen of all these threats.
A bunch of straw-man arguments ("See they argued in support of the risks of something I don't care about or think is important, therefore there are no risks at all").
False motivations ("It's just OpenAI trying to capture the market"), while almost 70% of the people signing the letter don't work for OpenAI.
Absolutely absurd thoughts: "Do they want to be working on something capable of 'human extinction'?"
And from people who supposedly work in tech, but somehow are suddenly incapable of grasping the difference in scale of impact that automating something for extremely low cost has on society. "We always had these risks in society, so this is nothing new". Completely ignoring the difference in for example having to travel to to basement archives of every small town in the country to read through papers to research something vs having all of that information at the tips of your fingers.
I don't think I've ever read any set of responses on HN that I've been more disappointed in. And not necessarily because I disagree with them, but because of how little critical thought when into them.
I think there will be a time where we'll look back at comments like these in response to this concern as being absolutely painfully naive.
I’m not sure the cause, but HN is clearly getting worse on the frontier of broad specialties (AI, frontend dev, econ) whenever one has extensive domain knowledge in them.
The amount of scoffing at OpenAI since inception almost a decade ago, is wild, as they continuously flew by the expectations here… the old threads did not age well and they’re not on track to now either
Sometimes overhyped tech is just over hyped tech.
1. further entrenchment of existing inequalities ... we were doing this LONG before AI came along. This isnt a new problem, it's not like AI is the reason some groups are under-represented in tech. (and it's not like tech itself is to blame alone).
2. manipulation and misinformation... Go back in history and the progressive movement was all in on eugenics in the early 1900's. Alex jones is a thing today. This was already a problem.
3. loss of control of autonomous AI systems ... there's the doomer I know and love! If your an AIG doomer go read I, Pencil. Look at the guy who tried to make a toaster from nothing (and bread too). Read up about how brittle the entire power system is. Things are falling apart because we're too cheap to put the people and resources into maintaining them. Unless AGI wants to die in the process it's going to be a long time before a world exists where it doesn't need people.
Seriously, climate change, nuclear threats, a bad screw up in bio research... way more likely to kill us all than for us to get to AGI.
AI today is about as useful as a bitcoin in a black out. It's interesting but it doesn't DO anything other than waste power and resources.
EDIT: it was so worrying that all the main safety folks quit openAI and no longer want to save us /s
Much of the history of the world's revolutions can be chalked up to that.
Bronze weapons, horsemanship, chariots, gunpowder, ship-building, castles, siege tech, engines, tanks, submarines, warplanes. Each of those catalyzed at least a new regional power and in many cases a new world power. Those transitions were generally quite deadly and unpleasant to live through.
AI today is useful as an enhancer to human productivity. It can sometimes be a real time-saver in the hands of a trained professional - at least in fields such as software development where the correctness of an answer can be validated quickly and safely, and the harm caused when the AI gets it wrong is unlikely to be significant. [0] But in that use case it is supplementing humans, not replacing them - without a trained experienced human in the loop, the AI can’t do anything
[0] It is possible the AI might get it wrong in a subtle way which no human notices and then blows up badly in production - but humans already do that anyway - and we already have tools to try to catch some of those problems (security scanners, memory leak detectors, etc) - and it is possible some of those tools might perform even better with AI assistance
The massive consumption of resources on AI research is a real material threat that AI safety people DO NOT TALK ABOUT. What if the real paper clip problem is the power and resources we waste trying to get to AI?
And because if it turns into a global or multi national LHC like effort all those PHD's dont get valley rich.
You're not wrong, but they aren't right because they aren't putting the issues that would impact them first.