Five men control AI. Who should control them? | Hacker News Reader

Five men control AI. Who should control them?

economist.com

37 pointsandsoitis39 comments

Clearly the Economist and their panel of experts.

So basically the same 5 men, considering that the Economist is the mouthpiece of the capitalist global oligarchy

lol bit of a stretch there, seeing as there are dozens of companies training LLMs.

As training software and infrastructure matures plenty more entrants will enter the market. It’s not like this is a particularly challenging research field, just very expensive at the moment.

Which LLM should I be using for programming work that isn’t released by OpenAI, DeepSeek, or Claude?

Which one outperforms this small handful of options within the AI oligopoly?

This statement you’re making is like saying “there are dozens of Android phone OEMs” when in reality Apple is gobbling up 80% of the profits, Samsung/Google are consuming another 10%, and everyone outside of China has their app installs gated by Google Play or Apple App Store.

Demis Hassabis is on their list? He reports to Sundar (who reports to Sergey Brin?)

Also no mention of Huang? He's behind all of them in the supply chain.

No Chinese? Guess they're no good at ai.

That wouldn't fit the narrative.

Chinese companies are lagging behind and always have. They are fast-followers but don't decide the frontier.

The countries that they're in already do via the law. No one else should "control" someone.

The law is only relevant insofar as it's enforced. In America, that's a tossup.

Good point. People do not think of a scenario where one billionaire might decide to take their wealth and resources and hunker down on a dictator-controlled country where extradition does not apply, that person could easily experiment and create an AI that may not necessarily see us as relevant to their existence.

I probably won't be able to respond to this comment since some people on this forum have flagged my comments as inappropriate thus limiting the number of daily posts I can make :)

no one should have to control some one, until they become a threat.

when someone presents a threat, at large, they have limited entitlement to walk among society, or act without review.

> no one should have to control some one, until they become a threat

The Helots were a threat to Spartans. Black Haitians to the French. Jews to the Reich.

Threats feel like a reasonable reason to reduce another’s rights. But they turn out to be the most usual way of tricking oneself into becoming a monster.

Are you comparing the ai ceos to helots? I am confused

Congratulations! You just compared regulating the behavior of a handful of billionaires to the holocaust! You just equated the idea that there should be some democratic restrictions based on corporate activity with death camps that murdered millions!

You win the "most HN post of the month" award. Never change, HN. Never change.

>You just compared regulating the behavior of a handful of billionaires to the holocaust!

They literally did not.

"Jews to the Reich."

Yes they did.

> You just compared regulating the behavior of a handful of billionaires to the holocaust!

On the most surface level, sure. Regulating something and controlling someone are, to me, different motivations.

I am starting to believe a significant number of humans run a computation that goes something like this: "Can I control AI? Will I meet people that control AI personally? If no, why would I care if they're treated unfairly in the abstract? Most important thing for me is they don't affect my resources in any way. They're better off than most either way, if anything not willingly reducing their power shows greed and confirms they're threats."

I interpret it more generously. When a pet or a child misbehaves, we constrain their behavior. For most people, I’d guess that’s the majority of bad behavior they come across in daily life. (When adults misbehave, one usually distances or confronts. The latter isn’t an option for a difficult-to-reach public figure. And some of these figures make distancing difficult, too.)

i fixed that for you:

"The Spartans were a threat to Helots. the French to Black Haitians. the reich to the Jews."

justification, doesnt transform a victim into a threat.

The whole point is that the self-fulfilling prophecy and their own cruelty which created the victims is exactly what threatened them later. One reducto-as-absurdum hypothetical I give of this type of self-fulfilling prophecy from fallacious logic is that if group A decided that say, all redheads were all vicious bandits who would kill them on sight and therefore should be killed, guess who is now incentivized to kill Group A on sight?

I'll do a bit more "fixing"

"justification, doesnt transform a victim into a threat"

unless the victim is Palestinian, and the monsters are jewdaic zionist terrorists, for more than 100 years now

oops you broke it again stahlmann, a victim is a victim period.

it doesnt matter who is right or wrong at the start, there is the attacker, and the attacked.

victim, and attacker swap places as they go around the wheel.

now what breaks every thing is when a militant in combat is spun as a victim, defending from mother and child.

generalizing based on nationality or eye color or anything else that is the actual problem you seem to be concerned about.

let that be your last battlefield.

https://en.wikipedia.org/wiki/Let_That_Be_Your_Last_Battlefi...

to be fair, that's exactly what's at issue. controlling AI implies controlling society as intelligence scales.

This is singulitarian fallacies all over again like 'being able to make something smarter than a human means infintely smart, because it can just keep on making one smarter' while ignoring the multifaceted nature of intelligence, the time and other costs involved in creation and the costs. It just gets handwaved away as superintelligence somehow enabling goddamned sorcery to ignore physical constraints. Except reality does not work that way.

It reminds me of the 'Einstein's superintelligent cat' refutation to such fallacies. It went something like this: imagine Einstein has a superintelligenct cat. The room has only one door and it is locked. The cat is not capable of opening the lock due to lack of manual dexterity. The cat does not want to go into the carrier. Einstein is however an order of magnitude greater in mass. As much as the cat might want to escape Albert Einstein's grip he cannot. The superintelligent cat is going in the carrier.

The point being that, no, controlling or creating AI does not in fact equate to controlling society no matter how smart it gets. Even if we were so incredibly stupid to wire it up to be able to actually control an entire munitions factory it still can't take over society, and it only takes one bombing run or called in artillery strike to end the situation.

Yet in the real world we can trust private ownership of firearm factories, missile factories, and tank factories without a serious risk of a coup. Yet somehow AI is supposed to be what makes them a god-king? It strains credulity.

>Yet in the real world we can trust private ownership of firearm factories, missile factories, and tank factories without a serious risk of a coup

Because they are highly fucking regulated....

Start selling missiles to kids and watch yourself get put in a cage.

These arguments have been going on for more than a decade and have been silly the whole time.

> It reminds me of the 'Einstein's superintelligent cat' refutation to such fallacies.

One (of the many) problem(s) with this "refutation" is that in reality not only does nobody bother to lock the superintelligent cat in room and leave it no available actions, but you're lucky if they don't hook the cat up directly to the internet. It doesn't matter whether you could maybe control a superintelligence, if you were very careful and treating it very seriously, when nobody is even trying, much less being very careful.

One man controls entire country. Yet we call that freedom and democracy.

Are you talking about the president? Because if so lol

Having one guy in charge of the entire executive branch was possibly the biggest blunder made by the framers of the US Constitution.

name checks out

Or Congress and the courts made the blunder of allowing the executives to accumulate so much power.

Fair, but one is structural, the other is a series of bad choices that could be undone without upending the constitutional order.

https://ai-2027.com does a solid job of demonstrating the existential risk of the singularity. If it is actually approaching, we need leaders who will give potential black swan events the severe caution they are due.

I sure hope the theoretical timeline is compressed because the singularity under Donald Trump likely means that we're all dead due to misalignment.

Singularity is something we might never see. The question is whether we crossed the event horizon.

"Insider is supported by Anthropic"

“Insider is supported by ANTHROPIC“ get their money and act like independent? What a joke

Reply on news.ycombinator.com