Palantir CEO Rejects Calls to Pause AI Development
newsnotfound.com
newsnotfound.com
I agree there's no reason to "pause" anything but it's just a broken clock being right twice per day.
This seems like the kind of team sports mentality that has destroyed politics.
The controversial things everyone talks about are their other products like Gotham which is for the police/military.
In germany it is Unconstitutional to use.
Here is a german article about it: https://www.hessenschau.de/politik/warum-die-polizei-softwar...
And Germany has bought and used those same tools, it's not a US specific thing.
At some point it comes down to your own personal ethics on this, but to me you can't say "that Palantir"... it's all Palantir.
Of course that should not be controversial for Airbus, which themselves operate military industry and directly cooperate with firms like Raytheon and are competing with Lockheed and such. Working with Palantir is just another drop.
Microsoft, Google, and Amazon all have substantial sales to the DoD. Part of the military industrial complex?
What about Uber, Hilton, AB InBev, and Coca-Cola? What about Harvard and Yale? What about White House staffers?
I would say absolutely yes. All these companies make US military operations possible.
>What about Uber, Hilton, AB InBev, and Coca-Cola?
Not really. If they are used at all, they are used incidentally so. Certainly the military does not need them.
>What about Harvard and Yale?
Difficult question. I tend towards yes.
>What about White House staffers?
Depends I think.
There is valid criticism for some of these military-industrial companies lobbying for various horrible historical acts to improve their business, and that obviously is immoral - but that's not immoral because they make weapons but because of these specific acts which put them in the exact same position as United Fruit Company, which caused violent conflicts to benefit their banana business.
I share your ideal but realistically this is a useless stance if there is anyone in the world that doesn’t share your ideal and doesn’t mind slaughtering others.
As for the other parts of Palantir that do business with other parts of the security state like Boeing, well, that's why they incorporate. To avoid legal and social accountability for their criminal actions by abstracting it away even within their organization.
https://www.businessinsider.com/palantir-wikileaks-apology-2...
https://www.businessinsider.com/palantir-wikileaks-2011-2#-1...
Perhaps it's plainly true in some trivial sense (all large orgs have committed some crimes), but not in any legally-relevant sense. Otherwise people have to explain the obvious incongruity between this "plain truth" and the obvious fact that Palantir is not facing the sort of legal problems that would be expected of a "criminal organization" (and no, saying "the US judiciary is just very very corrupt" is not a "plain truth" either.)
By all means criticise away, by constructing an argument and preferably providing sources.
Saying "X is bad" is cheap and does not further discussion or teach anyone. In fact, that kind of behaviour being normalised is the most susceptible to brigading and bot farms.
no person on earth -- who is informed by anything aside from safeguarding or expanding their own capital -- would make that argument.
Comments such as: > Yeah, no shit. That's what I'm saying, Peter
Are in contradiction with HN guidelines and what makes this site valuable, in my opinion, is informed discussion. There are multiple forums where sarcasm and low effort comments are the norm; HN should be more than that.
This isn’t like a big conspiracy theory people have won Pulitzer Prizes over it and all that.
It’s a completely reasonably inference that this person advocating for fewer restrictions on a powerful new technology is hoping to use it to conduct more sophisticated mass surveillance on Americans.
As I said, nothing could be more substantive to this thread.
Even were that the case, it's unclear to me that merely providing material aid to a criminal government organization would make a company a "criminal enterprise".
Perhaps uncharitably, this seems akin to arguments that would condemn a large share of contractors involved in military / police / intelligence work as "criminal".
Yes of course the inability of the legal system to deal with it is a deep problem.
My argument is that people who break the law are criminals. It’s not a subtle point I’m making here.
The only point of confusion seems to be your surprise to discover that people with power commit crimes openly and with impunity.
In any case, even when illegal government programs exist it's again unclear to me what sort of due diligence a contractor would have to make to avoid being labeled as a "criminal organization". Similarly, a lot of people consider soldiers who fought in Iraq to be murderers, and I see the point but it would come off as confusing in a normal conversation about a particular person to just casually label them as a murderer because of this.
I am pleased that all in this discussion have now demonstrated such a rigorous grasp of the obvious.
To your more specific point, my first recommendation for government contractors who wish to avoid being labeled as a “criminal organization” would be to avoid repeatedly and blatantly committing crimes.
Has there ever been an era in American history where there aren’t private entities with high government officials as allies who engage in organized criminal activity to further our international trade interests while avoiding consequences?
Not sure if you’re a history buff, but…
"Treason doth never prosper, what's the reason? For if it prosper, none dare call it treason."
It may be illegal, but this is not legally relevant at the moment.
I do find that Karp sounds like a huckster, but criminal … sad that this got so many upvotes.
When it comes time to vote though people tend to go with the person telling them they'll have more.
If you want to suck co2 out of the air, no one will have issue with it, you will just run out of money without making a dent.
Also these are similar arguments to what has been used to criticize nuclear energy for decades, which has obviously been a disaster on net.
About climate change..? It doesn't serve your argument to so blatently misrepresent reality.
1. This article has nothing to do with climate change, so why is such a comment the top reply?
2. "nobody says we should stop anything [to address climate change]" is a flagrant lie. You guys are being baited.
Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll look at the data.
2. if it was not obvious, my comment was sarcastic. Of course there are people who care about both, what surprises me is all this fanfare from CEOs about something so speculative that is completely science fiction right now, when the existential threat is here and now, and very few seem to point at some urgent solution like, IDK, stopping using fossil fuel in 5 years? We have about a decade to fix this [1] and the progress so far has been, at best, lacklustre [2].
[1] https://www.ipcc.ch/report/ar6/syr/resources/spm-headline-st... [2] https://climateactiontracker.org/
Unfortunately, with AI things are different. By the time humans notice a real problem, we may be months if not hours away from death.
The basic idea is that once AGI hits an unknown capability threshold it will likely recursively self-improve into something very dangerous and difficult to control, and will likely be able to come up with an effective plan to remove any obstacles to its intended desires, ie. humans.
People have varying degrees of confidence in this scenario, but even if you peg it at like a 10% likelihood you're basically conceding that it's the single most important policy issue in the current political climate.
What if the likelihood of AI "solving" climate change is 10% and the alternative is human extinction?
Can we agree that coming up with sci-fi scenarios and assigning arbitrary likelihood figures to them is pointless?
Good discourse only happens when people take efforts to honestly represent the state of affairs.
Happy to be given counter examples.
Imagine if a real natural disaster occurs within the next 5 years, like another Chernobyl.
It will be pretty tragic for our species.
edit: I'm realizing there may be different classes of people who want to regulate AI in some capacity. I'm basing myself off of this train of thought from the Center for Humane Technology: https://www.youtube.com/watch?v=xoVJKj8lcNQ
Palantir is mostly in the analytics, ingest, and business intelligence space. That said I think they are making an AI Platform of sorts, but it's not that impressive.
edit:
there were some interesting responses, but it seems HN has pruned the entire comment chain.
i'd like to encourage others to consider whether censorship is necessary in adult spaces, especially HN - i did not ask to be insulated from criticism. critical comments are a chance to have a healthy discourse.
for what it's worth, to answer the person who asked me what it feels like to be a murderer, since their comment didn't deserve to be censored - i don't feel like a murderer. this is my first time even hearing about anduril, but i did notice that there is a whole spectrum of jobs open on their page and it felt right to pass that on.
given that we're all adults here, capable of self moderating our own emotions, im not impressed that the comment chain was removed, and am hoping to interpret this in the most charitable way possible: that it may have been an automated measure responsible.
They offer $112,000 - $198,000 a year for this position, which requires some pretty wild math competency, electrical engineering, simulations experience, and security clearance:
https://jobs.lever.co/anduril/b78e5e78-f09f-4be8-995c-71ead3...
Or you can write React and TypeScript for $160,000 - $240,000 a year
https://jobs.lever.co/anduril/610cb495-bfd9-4a0c-8c55-a2756a...
AI risks include, but are not limited to, "if you get this wrong everyone dies" and also "if you get this right, your idea of 'good' is indelibly imprinted on the future light cone, and we don't fully agree with your idea of good".
It sounds reasonable, and if it was possible to do, might be effective.
But it's absolutely not possible after a certain point, and that point is past.
The people who don't realize that just look silly, or they have an agenda.
The reality is the cat is out of the bag in so many ways, particularly if you’re a bad actor LLAMA models are available to you. It’s already possible to just create infinite propaganda/spam now. What would 6 months possibly solve?
I presume he means among producers of AI products. But factoring in others (like researchers, lawmakers, concerned citizens, me, Sarah Connor) you could flip that around to 'only those with products don't want a pause'. Which is a bit like meth producers not wanting to surrender their labs.
But something tells me Alex Karp isn't concerned about ethics.
There has been press hype linking Palantir with AI however:
FT: AI has given Palantir its mystique back
The Register: Former NHS AI leader joins US spy-tech firm Palantir
It’s fairer to say it probably aspires to be seen as an AI stock.
They’re financially incentivised to lobby for unregulated AI to realise “upside”.
These kind of libertarians are all about "economic liberty", e.g. unfettering and deregulating corporations. They care little for individual liberties; a surveillance police state administered by corporations is right up their alley. In their view, the problem with the Stasi is the Stasi wasn't a for-profit corporation, and was consequently inefficient at what they did.
I assume many parties will be pursuing pushing the envelope on LLM training (and sometimes deployment) regardless of regulations/pledges/treaties.
The mass extinction combined with a 2+C world will kill enough of us fast enough that most of these imagined problems probably won’t be a threat.
Who will care about killer robots when feeding ourselves, getting clean water, and finding clean air will be a constant challenge?
I’m all down for regulation and forcing AI companies, as the others, to pay for the externalities the enforce on us.
It isn’t a rogue language model that it suddenly going to brute force it’s way to sentience and kill us all; it’s billionaire tech companies with profit motivations. That they can point the finger to a computer and say, “AI did it and you didn’t listen to us when we tried to warn you,” is some fabulously weird post-capitalism.
One of them is.
AI also enables every conversation to be transcribed and indexed in a way that would make the Stasi envious.
This is because evil isn't limited to fat-cats twirling moustaches — such villains are just the easy-to-spot examples in the modern world because the big scary Communists fell on their faces decades ago.
> it’s billionaire tech companies with profit motivations
Could be. Could also be the next Pol Pot.
Did the PRC's Great Leap Forward cause such death on purpose, or by accident? Were Stalin and Sir Robert Peel malicious or incompetent with the Holomodor and the Great Famine of Ireland respectively?
Once AI gets to that level we will also have AI helping to defend against it though.
Especially coming from the people who created attention machines that have changed the course of elections, had a major impact in spreading misinformation during a global pandemic, and are contributing to a mental health crisis. And yet we don't hear them call for more regulations for their own markets...
Quote me.
Karp believes that the West currently holds key commercial and military advantages in AI and should not relinquish them.
Delusional take, considering the East holds the AI advantage.
I think 'the east' has working multi-language LLMs, but the west's version doesn't quite perform as well with multiple languages
But since we're all speculating here, if I had to place my money on who will come out on top with AI, it would be the East. And by East, I specifically mean China.
While the West is scrambling to limit the pace of AI development, no such regulation exists in China. There have certainly been regulations to limit the power of Chinese companies in the past few years, but ultimately the shot caller is the CCP. An autocracy has more power and influence to direct its resources in a single direction much more efficiently than any democratic government.
Then there's the manufacturing side. The West does still control the most advanced chip fabrication technologies, but China is quickly ramping up development with results that rival those of TSMC/ASML[1]. If any country can be self-sufficient with chip fabrication, it's China. There's only so much harm US sanctions can do to a self-sufficient country.
[1]: https://thediplomat.com/2022/08/chinas-semiconductor-breakth...
> An autocracy has more power and influence to direct its resources in a single direction much more efficiently than any democratic government.
Fortunately, because it’s an autocracy, there is no way to know if it is the right direction, until it fails.
I worded that wrongly, but what I meant is that the current concerns with AI and discussions to pause AI development are entirely happening in Western countries. China will regulate its companies to comply with the goals of the CCP, but they don't have the same ethical and moral concerns as the West.
> Fortunately, because it’s an autocracy, there is no way to know if it is the right direction, until it fails.
They can also afford to fail fast. It's the "move fast and break things" approach to government. :)
There are yes. Environmental regulations that impact hardware. Personal information, privacy regulations that impact the training data. Business regulations that impact what businesses and business owners can do with labor, financial reporting, etc… We also have a fourth estate so if people like Yudkowksi get enough traction in the media the government will be force to act. That is worse than regulation by government, it’s regulation by the whim of the media owners.
These are not AI regulation per se, but they are de facto restrictions that aren’t present to the same degree in communist China.
Regarding the topic at hand: that’s plausible. I wouldn’t describe the west as “scrambling to limit AI development” though, more “establishing conventions and rules and accountability and and and”. I guess that might slow down research, but I don’t see it as that simplistic, like a blue bar ticking up in Civ; AI will need some major paradigm shifts before it gets really crazy, and I think a healthier and more sustainable field helps that rather than hindering it.
Plus there’s the fact that “the west” (kinda hate these terms tbh) has ~66% of all CS research - https://m-cacm.acm.org/magazines/2022/4/259407-trends-in-com...
> I wouldn’t describe the west as “scrambling to limit AI development” though, more “establishing conventions and rules and accountability and and and”.
Well, the discussions here are about pausing AI development. There's a growing public concern about its capabilities, and democratic governments must, in theory, abide to its citizens. An autocratic government can impose its will on the people, and even though China has passed AI regulation recently[1], it ultimately must serve the goals of the CCP. The Chinese government has much more power to direct its resources in ensuring it ends up on top in this race, while Western countries will be bogged down by bureaucracy, red tape, and more stringent regulation. Not to mention that the CCP doesn't share the same moral and ethical concerns. So any current lead the West has at the moment, will not remain so in a few years.
But this is entirely an opinion from a layperson, and I have no expertise in the matter. Cheers for the civil discussion!
[1]: https://www.reuters.com/technology/china-releases-draft-meas...
> TSMC is hosted by a strongly west-aligned country.
Which is right next to an increasingly hostile neighbour who keeps threatening to invade it. Unless the West is willing to go to War with China over TSMC, I don't see for how long it can maintain advantage.
Even Apple, Qualcomm depend on TSMC heavily. Lets face it. AI is primarily driven by GPUs sourced from Nvidia which depends on TSMC. China has been eyeing Taiwan for long now and we never know when it might decide to finally invade. There is no clear roadmap from the West on how it will defend Taiwan (except for saying US should blow up TSMC or TSMC should implement self-destruct mechanism in event of Chinese invasion [1]).
[1]: https://theprint.in/world/us-scholars-advice-for-taiwans-top...
Meanwhile version 2 has also been found in many places, like Mar-a-Lago and 10 Downing St, London.
(Although, given how Russia is doing, I wouldn't be surprised if a Spitfire won a few battles…)
Useful list here for some of what you're looking for.
With most high tech it's rather the other way around: the technology gets first used in the military and after some delay, the general public receives the tampered version to play with.
Isn’t that asking them to surrender instantly to any military who keeps the research going?
IMHO, there's nothing humane about indiscriminately lobbing large amounts of explosives at each other just because you don't have the technology to aim properly and strike more surgically, which are two things you might be able to do with AI. The Ukraine conflict is a stark reminder of how dehumanizing old school war can be. Scorched earth, lots of lives being lost, entire cities that get leveled to the ground.
Or perhaps AI can accurately determine the number of civilian casualties that can be had while staying below the threshold of public outrage.
Just spit balling here.
Does that mean we can’t talk about it or raise questions? After all, it is hard to prevent nations acquiring nuclear weapons too, but there is a tremendous amount of effort spent trying, isn’t it?