What is a bit weird about AI currently is that you basically always want to run the best model, but the price of the hardware is a bit ridiculous. In the 1990s, it was possible to run Linux on scrappy hardware. You could also always run other “building blocks” like Python, Docker, or C++ easily.

But the newest AI models require an order of magnitude more RAM than my system or the systems I typically rent have.

So I’m curious to people here, has this in the history of software happened before? Maybe computer games are a good example. There people would also have to upgrade their system to run the latest games.

Like AI, there were exciting classes of applications in the 70s, 80s and 90s that mandated pricier hardware. Anything 3D related, running multi-user systems, higher end CAD/EDA tooling, and running any server that actually got put under “real” load (more than 20 users).

If anything this isn’t so bad: $4K in 2025 dollars is an affordable desktop computer from the 90s.

The thing is I'm not that interested in running something that will run on a $4K rig. I'm a little frustrated by articles like this, because they claim to be running "R1" but it's a quantized version and/or it has a small context window... it's not meaningfully R1. I think to actually run R1 properly you need more like $250k.

But it's hard to tell because most of the stuff posted is people trying to do duct tape and bailing wire solutions.

I can run the 671B-Q8 version of R1 with a big context on a used dual-socket Xeon I bought for about $2k with 768GB of RAM. It gets about 1-1.5 tokens/sec, which is fine to give it a prompt and just come back an hour or so later. To get to many 10s of tokens/sec, you would need >8 GPUs with 80GB of HBM each, and you're probably talking well north of $250k. For the price, the 'used workstation with a ton of DDR4' approach works amazingly well.

If you google, there is a $6k setup for the non-quantized version running like 3-4 tps.

Indeed, even design and prepress required quite expensive hardware. There was a time when very expensive Silicone Graphics workstations were a thing.

Of course it has. Coughs in SGI and advanced 3D and video software like PowerAnimator, Softimage, Flame. Hardware + software combo starting around 60k of 90's dollars, but to do something really useful with it you'd have to enter 100-250k of 90's dollars range.

> What is a bit weird about AI currently is that you basically always want to run the best model,

I think the problem is thinking that you always need to use the best LLM. Consider this:

- When you don't need correct output (such as when writing a blog post, there's no right/wrong answer), "best" can be subjective.

- When you need correct output (such as when coding), you always need to review the result, no matter how good the model is.

IMO you can get 70% of the value of high end proprietary models by just using something like Llama 8b, which is runnable on most commodity hardware. That should increase to something like 80% - 90% when using bigger open models such as the newly released "mistral small 3"

With o1 I had a hairy mathematical problem recently related to video transcoding. I explained my flawed reasoning to o1, and it was kind of funny in that it took roughly the same amount of time to figure out the flaw in my reasoning, but it did, and it also provided detailed reasoning with correct math to correct me. Something like Llama 8b would've been worse than useless. I ran the same prompt by ChatGPT and Gemini, and both gave me sycophantic confirmation of my flawed reasoning.

> When you don't need correct output (such as when writing a blog post, there's no right/wrong answer), "best" can be subjective.

This is like, everything that is wrong with the Internet in a single sentence. If you are writing a blog post, please write the best blog post you can, if you don't have a strong opinion on "best," don't write.

I'm not sure Linux is the best comparison; it was specifically created to run on standard PC hardware. We have user access to AI models for little or no monetary cost, but they can be insanely expensive to run.

Maybe a better comparison would be weather simulations in the 90s? We had access to their outputs in the 90s but running the comparable calculations as a regular Joe might've actually been impossible without a huge bankroll.

Or 3D rendering, or even particularly intense graphic design-y stuff I think, right? In the 90’s… I mean, computers in the $1k-$2k range were pretty much entry level, right?

The early 90's and digital graphic production. Computer upgrades could make intensive alterations interactive. This was true of photoshop and excel. There were many bottle necks to speed. Upgrade a network of graphic machines from 10mbit networking to 100mbit did wonders for server based workflows.

Adjusting for inflation, $2000 is about the same price as the first iMac, an entry level consumer PC at the time. Local AI is still pretty accessible to hobbyist level spending.

Raytracing decent scenes was a big CPU hog in the 80s/90s for me. I'd have to leave single frames running overnight.

well, if there was e.g. a model trained for coding - i.e. specialization as such, having models trained mostly for this or that - instead of everything incl. Shakespeare, the kitchen sink and the cockroaches biology under it, that would make those runable on much low level hardware. But there is only one, The-Big-Deal.. in many incarnations.

In the 90's it was really expensive to run 3D Studio or POVray. It could take days to render a single image. Silicon Graphics workstations could do it faster but were out of the budget of non professionals.

Read “masters of doom”, they go into quite some detail on how Carmack got himself a very expensive work station to develop Doom/Quake.

We finally enter an era where the demand for more memory is really needed. Small local ai models will be used for many things in the near future. Requiring lots of memory. Even phones will be in the need for terabytes of fast memory in the future.

How were you running Docker in the 1990s?

> you basically always want to run the best model, but the price of the hardware is a bit ridiculous. In the 1990s, it was possible to run Linux on scrappy hardware. You could also always run other “building blocks” like Python, Docker, or C++ easily

= "When you needed to run common «building blocks» (such as, in other times, «Python, Docker, or C++» - normal fundamental software you may have needed), even scrappy hardware would suffice in the '90s"

As a matter of facts, people would upgrade foremostly for performance.

Heh. I caught that too, and was going to say "I totally remember running Docker on Slackware on my 386DX40. I had to upgrade to 8MB of RAM. Good times."

I just finished a blog post with some thoughts on AI’s future [1] and the surprising conclusion was that most big tech companies probably have much bigger problems than whether researchers leave or not.

As Taleb and DeepSeek’s CEO point out, usually when you have a disruptive technology, then the incumbents will be left behind. Cursor AI and DeepSeek are a sign of new players coming out of nowhere and beating the incumbents.

[1]: https://huijzer.xyz/posts/ai-learning-rate/

blancolirio (active pilot who regularly makes videos about crahses with the aim of improving safety) just released a new video on this accident: https://youtu.be/_3gD_lnBNu0

Summary: The helicopter might have been a little bit too high. The night vision goggles might have negatively affected the vision. The two aircraft were probably on different frequencies so didn’t hear each other. And the helicopter might have focused on the wrong plane to avoid because they didn’t see the right one thanks to the city skyline. It looks like the classic swiss cheese where multiple problems stacked up to cause the collision. See the video for details.

Wouldn’t the aircraft see the helicopter?

Unlikely; the pilots are (correctly) following the approach controller's direction. It's approach's responsibility to keep them clear of other traffic. Also they just got handed a different runway to land on that's a lot shorter and in one of the more complex airspaces of the planet; they're busy.

> That said, I’m still skeptical about how practical this really is for most people.

I'm running Open WebUI for months now for me and some friends as a front-end to one of the API providers (deepinfra in my case, but there are many others, see https://artificialanalysis.ai/).

Having 1.58-bit is very practical for me. I'm looking much forward to the API provider adding this model to their system. They also added a Llama turbo (also quantized) a few months back so I have good hopes.

Oh I love Open WebUI as well!! But glad to hear the 1.58bit version could be helpful to you!

> Is the claim that it only took $5M to train generally accepted?

Based on Nvidia being down 18% yesterday I would say the claim is generally accepted.

Because the markets are rational, all-knowing, and have never been wrong?

No, because the market is an aggregate of opinions, so it’s entirely fair to say it’s “generally accepted.” That has nothing to do with whether something happens to be true or not.

It may provide a financial opportunity for someone who disagrees with that aggregated opinion though.

That was not the question.

It is if you're using market movements as evidence of anything factual. If markets aren't rational, you can't use them that way.

do you only take advice/learn from all-knowing people?

Do you know any?

But here's my advice: drop the fallacious arguments and try something more honest.

my argument isn’t fallacious - it is logical: we can learn/use evidence from something without presuming it is all knowing. you are putting words in others mouths that they did not say

I'm sorry, I thought you introduced the "all-knowing" out of nowhere, but this was indeed mentioned by willsmith72. I'd missed that.

Still, his implied assertion that markets that markets can often behave irrationally, and can't be used as evidence of technical matters, seems pretty valid to me.

But I suppose you could see it as a sign that something is at least temporarily "generally accepted" among investors. That doesn't mean it's generally accepted among AI researchers, though.

Although I thought it was $6M rather than $5M, and that that was only the last step, and not the total investment. What does seem to be generally accepted among investors that this isn't good news for NVidia's profits, but that still doesn't mean that all the specific facts are generally accepted.

Don't worry, NVDA will bounce back and you will get a chance to get out.

Efficient market hypothesis is for nerds!

It is still unconfirmed since no one outside of deepseek reproduced it.

If confirmed, Nvidia could go down even more

based on information and background they thoroughly gave when releasing their research its pretty easy to put together that it did take them significantly less resources to train this model. only having specific parameters available at a time instead of activating everything all at once is pretty ingenious.

that and they just happened to be undergoing a large scale "cyber attack"

I'm not sure I see the bear argument for NVidia here. Huge AI models certainly drive NVidia sales, but huge AI models are also widely thought to be untrainable and nearly un-runnable save for large datacenters.

To me, this is ripe for an application of the Jevons paradox. If architectural improvements make similar models cheaper, I would expect to see more of them trained and deployed, not fewer, ultimately increasing the market for GPU-like hardware.

While Deepseek was an instigator in the price movements I would not say its accepted.

> Nvidia being down 18%

The only part of DeepSeek-R1 I do not like. I hope it's over, but I am not holding my breath.

Nvidia is now up only 1906% over 5 years. What a disgrace

It crashed all the way back to June 2024 levels, eons of progress wiped out

> If anything, they're likely trying to commoditize their complement in a way not all that dissimilar from Meta's approach.

Thanks. Great observation. Sounds indeed extremely plausible that they use the LLM for automated data cleaning.

I wonder if they shorted NVDA before releasing the model?

wouldn't that be outsider trading ?

'... that's right people, forget SPARKs, this season the 'it crowd' are spinning up companies solely to create turmoil so that they can short stocks.'

This sounds like maybe it's in the training data? Based on Elon going on about Wikipedia, I have been more carefully reading it and yes maybe it does have a bias (I'm not saying the bias is wrong, I'm not saying Elon is right, I'm only saying that maybe there is a bias).

For example, the page talking about blogs is for 20% about "Legal and social consequences" including "personal safety" [1]. And again, I think that's fine. Nothing wrong with discussing that. But I don't see any arguments why blogging is great such as it being useful for marketing, that you possibly have platform independence, and generally lots of freedom to write what you want to express.

Put differently, here on Hacker News we have a lot of links pointing to blogs and I think generally they are great. However, if I would not know about blogs and read the blog Wikipedia page then I could conclude that blog's are very dangerous, which they shouldn't be.

And just to be sure. I'm not saying Wikipedia is bad and I'm not sure whether it's a good idea that Elon takes control of it. I think Wikipedia in the current form is great. I'm just saying maybe there is indeed a bias in the source data, and maybe that ends up in the models.

[1]: https://en.wikipedia.org/wiki/Blog

Wiki Is Open and has tons of money why would anyone buy it? There's already "unbiasing" or "bias shifting" projects for Wikipedia, but regardless the data is CC licensed just make a new one for a couple million and hire real editors and experts for $10mm/yr and get to it.

Yeah, that's definitely an option. It would be interesting to know for sure, though.

I think the blunder is firmly with the people who caused the blowback. I don’t think it’s a good idea to all be up in arms whenever someone says something you disagree with.

For context, people were upset about using the official accounts to endorse a party wholesale, in direct violation of their charter.

A subset were also upset with the nature of the endorsement.

And then you could also discuss whether a CEO of a company has to consider what they say, regardless if they're using an official account or not. This is more of a gray area, and has people divided.

I'm personally of the opinion that, as a CEO, you're always representing your company when speaking to the public to some extent.

Andy did not endorse Trump, just the Republican nominee for antitrust. Nevertheless, we have retracted the statement, which was put out by mistake due to an internal miscommunication. Proton is not controlled by any single person but by the nonprofit Proton Foundation, which has neutrality in its governing principles, and that remains the case today.

To give more context:

Andy, or someone else, posted what can be read as an endorsement[1].

> Until corporate Dems are thrown out, the reality is that Republicans remain more likely to tackle Big Tech abuses.

Putting aside that this was a divisive statement, the problem emerged when this was posted by the official company account.

For the duration of time that this was the official response, it's not unreasonable for the public to assume that Andy, or someone else, endorsed a political party using the official company account.

This is the core of what caused the incident to escalate.

It was later retracted, and followed up with a statement[2] on the incident.

[1]: https://archive.ph/2yWGz

[2]: https://www.reddit.com/r/ProtonMail/comments/1i2nz9v/on_poli...

The irony is that this was already changing: anti-trust has been massively empowered under Biden and Lina Khan. Trump is certainly attempting to bandwagon this tide, but the appearance of so many Big Tech leaders in VIP seats at his inauguration does not lend a favourable hindsight view to Andy Yen's statement.

Is that really true? What did Lina Khan do to thwart big tech and enforce antitrust laws to a greater extent? I feel like there was mostly blockers for M&A that allowed startups to exit, but little real impact on big tech.

She led the FTC to bring lawsuits against Meta, Amazon, and Microsoft. Although two of the cases have been blocked in the courts, this alone was unthinkable in previous administrations. She's also helped enforce policies like right-to-repair and click-to-cancel which threaten big tech's market share. Like it or not she's played a massive part in shifting the narrative on antitrust. Trump's dismissal of her along with Big Tech's front-row seats at his inauguration and general sycophantism are not good signs for antitrust.

JD Vance vocally supported her, then the administration he is part of fired her as soon as it was able.

By contrast, Democrats nominated her to be commissioner, officially it was Biden who did so but Elizabeth Warren, another Democrat, is the one who really pushed for it.

Words don't mean shit compared to action.

With your kind of reasoning its pretty obvious why America is currently driving itself off a cliff.

Sorry, your CEO tainted any respect I have for your organization. I was planning on move to your service this year. Now I am looking else where.

The statement maybe retracted but the message and statement and ideology remain. I cannot trust your organization based on your CEO.

The irony is the collection of big-tech that showed up for and financially supported Trump. This has a great profound outlook on real antitrust action towards those individuals and their organizations. Actions are more verbose than words but often ignored.

In this case though... the actions you are referring to are the words Andy Yen used?

As a long-time user and a volunteer on the translation team, I can't begin to express my disappointment with this.

I've subscribed to Proton for several years, and I now deeply regret my decision. Regardless of any explanations about the removal of the statement or internal miscommunication, the damage is already done.

In my opinion this is clearly a display of a character flaw. I simply don't trust Proton now that it's CEO revealed itself to be of such flawed thought. Isn't that the whole job of a CEO?

In my opinion, this is clearly a display of a character strength. I strongly trust proton now that its CEO revealed itself to be of such strong character. Isn’t that the whole job of a ceo? For every one of you, there is one of me to offset you and increase my support of the product. You probably don’t even use them and never intended you, and the rest of the user base is better for it.

> I don’t think it’s a good idea to all be up in arms whenever someone says something you disagree with

Of course not, but there's a difference between disagreeing with which is the best recipe for a cheesecake, the best castle in France, or whether or not it's acceptable to invade your neighbouring countries, if the government should be following the law/constitution, or basic biology, etc.

Some opinions one disagrees with are very worth to be up in arms. I'd even say that people have a civic duty of being up in arms against certain egregious topics. Say if a politician says that they want to allow for 9 year olds to be married to adults; or a rich guy backing multiple politicians all over the world Sieg Heils on national television; or a politician that just got elected is asking for money for favours from companies; or there's talk of "other"ing significant swathes of the population.

No, I think a guy doing a Nazi salute on TV, twice, at an official ceremony, that's tantamount to Nazism. And the fact that everyone who identifies with his side of the isle tries to deflect and excuse and explain how the Nazi salute he didn't wasn't actually a Nazi salute only helps with that. And frankly, anyone associated with Nazi salute guy and those who defend him... is associated with Nazis, which isn't a good thing.

It reminds me of this sketch: https://youtu.be/zvgZtdmyKlI

Gail Slater did that? That’s crazy, man. But it says on Wikipedia that Gail Slater is a woman?

I don’t think it’s good optics for CEOs to be simping for POTUS without a clear rationale.

Why is a privacy focused company from Switzerland kissing the ring? It’s a relevant question given the business they are in.

I’m getting more and more convinced that this is a main problem all over the West. In Germany, a business owner called Marco Scheel is becoming more and more popular by being very outspoken about how bureaucracy is hindering him. His company is called Nordwolle by the way. They make clothing out of sheep wool. They spend a lot of effort finding the right type of wool so they don’t need chemicals to paint it.

One major example which Marco first became popular with was that he owned a barn but wasn’t allowed to use it for the factory since it was a farm on paper. The government told him to move to a designated factory area. He argued that it made no sense since he was living in a very remote area, and the barn was of high quality. What else should he do with the barn? Why would he need to build something new somewhere else? The barn was there already and stood already for hundreds of years.

His most popular quote is something along the lines of “we can’t all sit with a Chai latte and a MacBook in a coworking space in Berlin and make the 5th dating app. We need some people who do that but not everyone. Some people need to make things with their hands! And for that I need space! I don’t need glass fibre. I need space!”

I associate this attitude with criticism of Wikipedia and narcissism. I don't think it's a coincidence that it's often around editing things they are related to that this comes up.

I tried to do something and they stopped me. This is wrong, I should be able to do this and write my own story.

Which is a perfectly normal feeling. But if you end up saying that loudly in public without ever thinking, well what if the rule of "let this person do whatever they want" applied to people other than yourself, then that seems to indicate some lack of a wider view.

> ... well what if the rule of "let this person do whatever they want" applied to people other than yourself, then that seems to indicate some lack of a wider view.

It sounds like Mr. Scheel is applying exactly that view. The idea is everyone should be able to use their remote farm shed for industrial purposes. Indeed, most of the intellectual foundation of the pro-freedom view is precisely that when you take a wider view freedom is generally better for everyone than authoritarianism right up until it becomes a threat to personal safety (even then, pushing the dial a little further towards freedom generally gets better results). If people can't do what they want, then how are things supposed to get done? If we're all doing things in ways that are believed to be impractical then it is going to waste an unreasonable amount of resources and be stupid.

> I associate this attitude with criticism of Wikipedia and narcissism.

I never said anything about Wikipedia. For the record, I'm a big fan of Wikipedia and I'm skeptical about the new US government.

Please don't assume that because someone holds opinion X that they also hold opinion Y. With the current levels of polarization, it's probably a fair assumption to make, but I think we all as individuals have a responsibility to counter that.

I brought up Wikipedia because it's something I'm interested in.

And it's an example of somewhere I'd seen this exact argument against rules/regulation regularly made on HN stories and the comments on them in what I thought was a mostly non-political context.

> I associate this attitude with criticism of Wikipedia and narcissism.

> I brought up Wikipedia because it's something I'm interested in.

> One major example which Marco first became popular with was that he owned a barn but wasn’t allowed to use it for the factory since it was a farm on paper. The government told him to move to a designated factory area. He argued that it made no sense since he was living in a very remote area, and the barn was of high quality. What else should he do with the barn? Why would he need to build something new somewhere else? The barn was there already and stood already for hundreds of years.

Of course it makes sense. Farmland is dedicated to farming and producing food/related things. It lacks connectivity, has fertile soil, prices are cheaper. If anyone can just build a factory there, they will have a negative ecological impact (interrupt animal flows, pollute in areas that are supposed to be cleaner, etc). It's the same reason why you can't farm in an industrial zone, nor can you set up a factory in the middle of the city.

Yes, it can be taken too far and abused, but absolutely 100% makes sense and must exist.

The guy's not building a gigafactory in his garden, is he?

> If anyone can just build a factory there, they will have a negative ecological impact (interrupt animal flows, pollute in areas that are supposed to be cleaner, etc).

All true of farming. FWIW as a fellow NIMBY myself, I use the excuse of 'animal flow' (in particular the flow of bats) to prevent anyone from putting anything more than a fence up within 150m of my house. It's great!

> The guy's not building a gigafactory in his garden, is he?

How could this possibly be known without a review in your opinion?

> fellow NIMBY myself

Unless you're American, things don't have to be so binary. The choice isn't between nothing gets built or anyone can just do whatever. We need a balance.

> All true of farming

I'm pretty sure birds and bees and what not prefer having plants than factories.

> How could this possibly be known without a review in your opinion?

By all means, have planning applications and a system to process them. Things don't have to be so binary.

You're going to end up in a position where you're telling a farmer how to manage & value farmland. That'll lead to more misses than hits.

> You're going to end up in a position where you're telling a farmer how to manage

Funny you say that. Not only does that actually happen in pretty much all developed country, it's actually needed for a variety of reasons. There are subsidies to incentivise the "correct" crops (you don't want all farmers only doing cash crops for export, rendering your country very vulnerable to import markets to sustain itself), there are also rules/policies to rotate crops to avoid top soil erosion which could be devastating, there are rules on what types of pesticides can be used, etc etc etc etc.

> Not only does that actually happen in pretty much all developed country...

"Everyone does it" isn't much of an argument when it comes to economics, the field is littered with a long history of group-think episodes where most people do things in a way that was, in hindsight, a mistake. And being steamrollered by more economically productive societies that don't ban progress. The modern policies developed countries adopted have resulted in vast investments in China (and Asia more broadly) to dodge the regulatory states that were built.

And the rest of your comment is straightforwardly telling farmers how to farm. On average, I bet they know all that stuff better than the legislators. They're farmers! If we can't trust them to farm then putting regulators in charge isn't going to save us. That attitude of mother knowing best is still going to result in more misses than hits, even if confidently repeated a few times.

> And the rest of your comment is straightforwardly telling farmers how to farm. On average, I bet they know all that stuff better than the legislators. They're farmers

Strongly disagree. The incentives are just not the same. If farmers use pesticides which will kill all bugs and pollute nearby rivers to increase their yield a tiny bit, that's not good for everyone else. If they decide they're only going to do tobacco because it's very lucrative to export, that's not good either. If the techniques they're using are obsolete (and thus inefficient and resulting in them barely being able to survive against foreign competition) or very bad for the soil/environment.

Farmers produce food, it's one of the most critical things in a country. If things go wrong, there are famines or economical crisis (cf. Egypt, Sri Lanka in the last few years, Soviet Russia in the past century). Hell, many countries were couped to take over control over their farming sectors for commercial interests (Hawai, Central America and the Caribbean, cf. the Banana Wars).