Hacker Newsnew | past | comments | ask | show | jobs | submit | marcus_holmes's commentslogin

We used to have a holiday home in Cornwall when I was young. My father used to have a pint or two at the local pub most evenings when we were there, and we were there 3-4 weeks every year. Eventually we sold the house and stopped going.

My father happened to go back there about 15 years later, randomly on a road trip through Cornwall and thought he'd stop in for a pint for old times' sake. On walking through the door, the barman said "hello Simon, good to see you again, pint of the usual?".

I wonder how many people call in to have a pint at the pub purely because of that barman. I wonder if that pub has survived the cull of British pubs purely because of that barman.


Except the PM is better at the internal politics needed to survive the drama.

Won't that just create another channel for social engineering to delete a victim's account?

> But if you imagine a sci-fi world where machines can build and do everything humans can do, the concept of a human-centric economy would be pointless.

In Iain Banks' The Culture novels, the machines provide the How, humans provide the Why.


I mean, there is a read of the culture that is pretty dystopian. If Minds are able to, essentially, predict the future, what is the purpose of the humans, other than as a sort of abstract pet acting out the Minds' great plans.

Something something Bora Horza Gobuchul was right all along.


No, that's a read from people who equate suffering and toiling with meaning. The Culture is a true utopia, Iain says it explicitly.

Minds keep humans as pets.

I can see why you might think that - but people and machines and groups are free to leave if they want - at least one of the Culture stories is based on this: e.g. "A Gift from the Culture".

Free-to-roam pets are still pets. The Minds are benevolent dictators, and generally seek not to be seen as such by humans.

What would the Minds do if all humans wanted to leave?

Probably sublime... who knows!

> when the founder receives pay for their work, in the form of cash or exercised options or whatever else.

No. Earned income is salary, and taxed as such. Unearned income is capital gains, exercised options, etc, not taxed as income. It's pretty easy.

If we just use this tax office definition, it's effectively impossible to earn a billion dollars because you'd need to pay yourself ~$1.6 billion dollars in salary (depending on where you live), and very, very, few people can afford to do that


Being paid in stock is a form of income, which is taxed when given and which is separate from the additional capital gains tax levied when the stock is sold. Regardless, even if different forms of compensation are taxed different, the point is that they’re still earned.

> You're in the 0.001%. Your asks are arcane and orthogonal to most users of software, who just want their PC to do something neat and useful.

Right up until enshittification kicks in and suddenly everyone cares and there are shouts of destroying the evil techbros who are poisoning the minds of our youth to buy a new yacht.

Can you imagine the situation if Jobs hadn't killed Flash? Most of the commercial websites required a Flash blob to deliver full functionality even back then in the early 2000's. Adobe never even vaguely pretended to be the good guys, they would have enshittified as soon as they possibly could, as hard as they possibly could (as they have done with the rest of their software). The entire web would be held to ransom at this point.


> Most of the commercial websites required a Flash blob to deliver full functionality

Being a binary blob is not a strong argument all by itself. chrome.exe, firefox.exe, etc. are also binary blobs. I have no love for Adobe, but that specific criticism is weak.


I think there's a difference between a single-install binary blob to run the entire browser, and a binary blob required to view a single website in the browser. But I don't think it's a huge difference, for sure.

I can very easily see a licensing requirement coming soon. Running a higher-grade AI will require a govt-issued license, which involves a six-month application process, explanations of why you need to run it, where it's going to be stored and who will have access to it, pretty much the same as non-USA countries deal with firearms.

TACO Tuesday!

The ban on exporting cryptography in the 90's lasted for years, and got to be a major pain in the arse for the entire web industry in its early years. The US govt can be very stubborn about this stuff when it wants to be.

In the late 90s the US was in a position of power that it no longer holds.

It had just one the cold war and China wasn't even a shadow of what it is now.


Almost all of the major tech companies are either HQ'ed in the US or have a very significant US entity, and make up probably about half of the S&P500. The US's power has changed and is actively changing, but the US still holds all the cards in 2026.

> the US still holds all the cards in 2026

So, the last time an AI-related export control was imposed - NVidia chips deemed to be "too powerful" - how do you think that will work out? If the US is holding all the cards, why is China now refusing their chips?

Don't you think these types of restrictions will weaken rather than empower the US?


> Don't you think these types of restrictions will weaken rather than empower the US?

I do. But I think the US is currently in a position of strength that they are continually undermining.


They're undermining their position of strength precisely by using these types of restrictions. That's the point I'm trying to make.

For the sake of argument, assume everyone is working on good faith and at least believes and means the things they're saying.

The US government believes that Fable/Mythos is a weapon that needs to be export-controlled, and limited to only US customers. Presumably OpenAI/xAI/Google would face the same constraints, for the same reasons.

OS/foreign models are unaffected - OS because they cannot control who runs them, and foreign because they are not controlled by the US government. We could assume that China will implement the same policy controls, but they see the world differently so might not.

So US AI companies are then limited to the US market, effectively, after about six months (the lag between the current frontier models and the OS models). They have much less incentive to push the envelope to create better models, because the US govt might also ban those completely.

The investor froth around the race to AGI dies, so valuations shrink (the current IPOs may be affected), and presumably the bubble bursts. None of the AI companies can afford to continue building data centres, so that all dies immediately. US GDP drops by ~5% because of that alone.

In a year's time, the US is in a major recession because it gambled so hard on AI. Europe less so, only because it was such a distant follower in that race. China is more-or-less unaffected. The best models are now OS/foreign, and AI is moving forward more slowly, but still moving forward.

Any other scenarios?


> Any other scenarios?

Get to Tuesday, restriction is lifted. Get to Friday, restriction is back on. Confusion reigns.

Interesting scenario though: do the other labs attempt to “dodge” the import restrictions by claiming their models are “dumber and not a threat” thereby maintaining larger market access.

If so, doesn’t this basically force a stall in US-based development? EU will keep doing its thing at its pace. Chinese models will get a boatload more popular, but will probably slow down as they can drop to whatever pace they wish.

cynical follow-up: if they’ve plateaued, is this a clever way to avoid the negative consequences of the market implosion that a substantially substandard model release would cause, thereby giving them an “out”?


It is very hard to believe the US government is operating in good faith any more. Do I need to gesture more broadly at the open corruption?

Expecting “transparency” out of a government trying to protect national interests seems like a tall order. They have to withhold or obfuscate things to do that job.

Agreed. Going by patterns in the Iran war, members of Trump's family/in-crowd will invest in AI while it suffers from this decision, and then 15 mins later Trump will reverse the decision.

The thing is, that blatant market manipulation is playing with fire here, as so much of the US economy is invested in the AI bubble.


A Holmes indeed... your deductive powers are piercingly perceptive! (the event chain was a joy to follow, gave me ai2027 vibes, but slowdown like)

Of course, the world is not filled with rational actors, and the probability of the current administration allowing the market to tank like that seems next to null, so Occam's razor (or whatever) would point to another TACO inevitably incoming

I'd certainly bet on your scenario if it was reasonable to assume the US and China could get over the 'race to the top or die at the bottom' dynamic

so far ai2027 seems to be playing out to an almost uncanny accuracy, realpolitik obliterates the façade yet again


> OS models are unaffected

If the major nations that host companies that create those OS models implement export control on top models, there won't be any new OS models with top capabilities.


Can you expand? As far as I'm aware, the OS models are being created by researchers and organisations that are not hosting them commercially, so cannot be banned from serving them. Or am I missing something?

Assuming anyone involved in this crap is operating in good faith is foolish at best. The only thing any of them give a shit about is accumulating money and power.

Agreed, which is why I prefaced it with "for the sake of argument", an idiom you might not be familiar with (but is very useful).

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: