IMO his comments are blown way out of proportion. He says "from a global allocation perspective" it's better to have RAM in the cloud where it can be used much more efficiently. I think there's an argument to be made that this is more than just economics to most people, it's about ownership and freedom, but he's clearly just talking economics.
Absolutely, I agree that money is better centralized in the goverment where it can be put to use for the benefit of all society instead of being concentrated in billionares.
While I agree there's certainly a better in between, if you go the other way you have moral hazard. If people aren't bearing the risks (or rewards) for how they allocate resources, history shows they allocate them poorly.
How? Elon and Dario aren’t responsible? It’s all AI companies and hyper scalers. Let’s not forget how 3 companies in memory industry have got caught in colluding to keep things high.
Because he (on behalf of Open AI) made deals with 2 of the 3 major manufacturers of RAM to buy a significant proportion of their output, in the same week, without either of the manufacturers knowing that he was also making a deal with the other one. This directly led to the current RAM shortage.
This problem is far bigger than one individual. He's making rational decisions is his capacity as CEO of an AI company. If he didn't make large compute deals, all his customers would be upset when they're hitting usage limits 10 times a day and all his investors would be doubting his decisions, especially when competitors are also trying to hoover up all the compute.
Is there actual evidence that these deals caused the RAM shortage? Seems like a stretch to pin a global supply problem on one person when demand is coming from across the entire industry.
And if one person can indeed cause it then blame the system? I go back to the proven legal case of them fixing prices. Now there is just a scapegoat because it feel good?
The system shares some blame, but the individual is also the one exploiting a confluence of factors. Should the system prevent one person from purchasing 40% of global RAM production with fake money, yes, it should. But our systems are glacial to respond to edge cases they were not prepared for, and this is an extreme edge case - people who are not Sam Altman could not pull off purchasing 40% of the world's RAM production capacity on an IOU. Only a very few individuals in the world could have exploited this gap in the system, and the one who actually did deserves blame for it.
Sure, but he's also insinuating that the economics of the situation is more important than the other, unmentioned aspects. In short, that his interests weigh more in what "best" means.
Everybody believes their interests weigh more. Fortunately we have capitalism, so at the end of the day you have to put your money where your mouth is and pay for what you think your interest is worth.
I don't. When I make decisions that might benefit me but affect others, I consider the impact on others before deciding what to do. Sometimes (often?) I choose the option that benefits me less (or is neutral, or hurts me), if I believe the negatives to others are too high.
> I don't. When I make decisions that might benefit me but affect others, I consider the impact on others before deciding what to do. Sometimes (often?) I choose the option that benefits me less (or is neutral, or hurts me), if I believe the negatives to others are too high.
I didn't say that means they don't consider others, and I haven't taken a survey of everybody on the planet.
Obviously people believe their interests weigh more.
I don't. I believe I have to weigh my benefits to collective ones. Sometimes I do things that offer me no benefit whatsoever but does for others. Just the other day I offered to take a picture for a couple who was trying to get a selfie together. Last week I bought an intern lunch. Why? Because they were an intern (grad student) and I'm sympathetic to that.
Fun fact: I'm not a "rare outlier". People like me are pretty common and can be found all over. We usually just don't openly talk about it because we're not looking for a reward. I'm not speaking up now because I'm looking for self-gratification or anything. I'm speaking up to push back against your claims. I'm speaking up because we live in an increasingly selfish period of time. I'm speaking up because we're also at cultural turning point and many people are saying "enough is enough." I sure as hell don't care how rich you are, but I do give a damn about where the economic floor is. If you take from that floor to lift yourself up, I'm upset. If you do this to lift yourself when you already have more money than you could possibly spend in several generations? I'm furious.
I believe you. I'd challenge you though to consider Sam Altman would make the same claim about himself. The difference between you is not that you are good and he is bad, it's that you likely have very different beliefs of what the "good" course of action is. He likely believes successful adoption of AI means advances in science, medicine, freeing people up for more meaningful things, etc.
Fair, maybe I should say I think he wouldn't see his actions as taking from the economic floor, he would argue he is raising it. E.g. more effectively utilising resources (like RAM) means RAM is cheaper, which means more people can use it for their businesses and ideas.
Maybe a billion people would view you as having and owning and spending more than several generations of their families, maybe more. Yet you see wealth accumulation as a problem because there are those with much more than you. Are you the villain of your own story?
Well I congratulate you for finally having the courage to speak up about offering to take a picture for some people that's very commendable, but that doesn't really address what I wrote. I didn't say everybody weighs their own interests at 100% and everybody elses' at 0%.
Sure. Better the capitalism that has Altman complaining about the peasants getting the RAM that he deserves better, than having the corrupt incompetent government and bureaucrats bought and paid for by him decide to allocate it "for the greater good" (i.e., his data centers).
Esp. for ML inference he's right, and not just for RAM. I think this should be thought out seriously instead of mocked, or else open models and local inference won't be able to compete with big data centers with exclusive contracts with hyper scalers. There ought to be some sweetspot where one pooled server could serve 50-100 people, with let's say 25 doing concurrent inference and seeing good throughput.
If DeepSeek-sized models will be required in the future to get Opus 5.5 quality, they're not feasible to run for a single person. Pooling usage is necessary to make it efficient.
But even that is pretty painful to sustain efficiently with current hardware prices:
The efficiency of centralization is not enough excuse to deny individual agency.
The argument for saying that ram is put to better use collected into central hands vs distributed among individuals is the same as for any and all other resources, and it's invalid.
Humans are not ants and should not aspire to the efficiency of ants at the expense of having to be ants.
You get some centralization, and you get some distribution, and it's completely invalid to be jealous of the portion of all resources that is in distribution.
This should indeed be "thought out seriously" and the result is either mockery or torches and daggers. Think that out seriously.
I think there's a divide here in people who hear the words "it's more efficient than" and are interpreting that statement on its own (it's a true statement), and people who are assuming the statement also implies "we should only give it to data centers" (which I don't think anyone agrees with)
You misinterpreted my comment. I'm advocating for a middle ground where we may pool hardware in small groups for attacking the hyperscaler moat, specifically for ML inference.
...just economics that benefit him. The economics of rental for consumers are horrific. It doesn't even have to be about an abstract sense of ownership and freedom, even if it's "just economics" this is still an insanely self-serving to say because he's talking about economics that benefit 00.01% at the expense of the other 99.99%.
It's not necessarily "at the expense of" if people want to use those products every day.
Imagine if we had a shortage of cars: your city only had 20% of its current vehicles. It would be accurate to say "in this global shortage, new cars are more efficiently used by groups/taxis/uber than sitting idle in an individuals home". A car used by a group share would allow more individuals in the city to have access to vehicle transport. If cars instead all went to individuals, only a privileged 20% would have transport.
There simply isn't enough ram in the world for every individual to do local inference right now: one person having a 512GB RAM system means others don't. Data centers can batch the processing and serve many, many more at once.
That's not to say individuals shouldn't have it! individuals can buy a 512GB mac right now. Or that it should go to Open AI. Your org or group could do the same batching (OpenAI is only a minority of compute, still). But saying "it's more efficiently used" is trivially true
This falls apart because of the simple fact that the people who had to contribute to create the pooled resource do not get to say how the pooled resource gets used.
This argument that only considers the one positive dimension of centralization and pooling is never valid.
What model are you suggesting that gives them more access/control?
Currently AI providers are hyper competitive and almost all of the world has access to it quite cheaply. If people had to buy hardware for local inference, it would only be a tiny percentage of the world's affluent tech enthusiasts with access to AI, wouldn't it?
We're currently managing the shortage with a hybrid model, where enthusiasts can still buy (more expensive) individual systems and the broader shared hardware goes to many providers. OpenAI is something like 15% of data center power in the US I think? That's a lot but we still have choice.
kinda bad economica though - the ai bubble eating all the RAM is bad for many industries worth of useful production, including the infrastructure that makes gaming and movie industries possible nowadays
That went out the window a few years ago. These tech oligarchs are getting more and more comfy with their optics being full psychoticism. They don't even know how to hide it anymore because all they think about and talk about and plan long short and medium term is extracting capitol at all costs for themselves.
Okay? So what's your argument, that we should just ignore them when they are being openly pathological? Sounds like that just let's them continue being crazy.
This is typically the best argument for using cloud - it is the environmentally beneficial thing to do. And yet, knowing what I know of spotify, I do believe that me serving my own music library is a tad less wasteful than most commercial music streaming operations are.
RAM in the cloud ends up being used by GitHub actions runners that fetch and install the world anytime a shmuck pushes a commit to his confobulator feature branch for the paper clip factory. Multi-cloud deployments of a service nobody asked for. Enterprise architects deploying a service mesh to deliver what can be done in 50MB of memory on a raspi. Running some vectors representing natural language through gigabytes of noise only made useful by some arcane math only to be unable to count letters in words.
I am sure there's also lots of actually efficient use of memory out there. But when enterprises are willing to burn cash, efficient processes or methods are nowhere to be found.
Maybe we should be talking about the global allocation perspective of wealth then. Since Altman is so keep to optimize allocation, perhaps he should start with this aspect first, and then move to RAM, as wealth would impact a wider area of economics than RAM.
If we the people collectively owned "the cloud", then sure. But since we don't, I can only take it as him tryng to become a RAMlord seeking to extract rent from the rest of us
Yeah, Altman and the people in that world are utilitarians, which you would think they would have learned by now doesn't translate well to public conversation (or reality if we are being honest).
Nah. OpenAI themselves are wasting tons of RAM by doing frivolous stuff like providing ChatGPT for free for millions upon millions of people so they can ask what rocks go well with pizza. And wtf was Sora for.
No, if compute/memory was _that_ scarce that you begin thinking about what a waste it is that a decent chunk of it is sitting in people's homes, then why aren't you dedicating every atom of silicon that you already own to more productive efforts first.
Ah yes, I'm sure my RAM would be more economically useful if it were used to generate slop videos for Sora and inappropriate child content bty Grok rather than running VS Code and postman on my laptop.
the ridiculous "no ad blocker" full-screen modal that's unclosable led me down a rabbit hole there - turns out there's some obfuscated script from "Ad Shield" which replaces the whole site with an iframe if you dare to block their domain, "html-load.com", which proxies ad networks.
fortunately one can just run debugger; to stop it early
I didn’t think we’d get to the point where all compute is headed for the data centre and normal people don’t get to have anything bigger than a phone but here we are.
Kind of inevitable when you realize how much people are naturally willing to give up computational sovereignty for the sake of convenience and psychological safety espousing values claiming the opposite, and it's still the case despite everything that's happened so far
I don't think there is a lot of conscious choice here: If at some point, anything more complex than a phone would be phased out or become unaffordable for consumers, that's not because consumers had somehow collectively voted they only want phones from now on.
Well sure someone could argue that the market isn't an actually reasonable set of choices that people can vote on, but how can it not be considered at least some form of voting? There's been at least 30-40 years in which consumers could try to find out. Unfortunately there's only a finite amount of resources in the market and consumers have to compromise on what takes priority in the tech industry, like getting work done or tackling issues that are visible only in front of them. When people collectively vote for politicians they also have to compromise
I don't feel extremely on either end, but I would prefer if consumers could own their choices and not try to pretend otherwise
Well, ok, what would you do in the current situation to "vote" for affordable RAM?
In a perfectly free and fair market and in a world where companies would plan their product strategy exclusively based on current consumer data, this would be true. But neither holds in the real world: The current AI gold rush is now something that is largely decided between investors, boardrooms and hardware vendors (*) with little input from small everyday consumers. And I wager this has been the case for many "tech trends" before that seem more forced on users than originating from them - such as forced updates, mandatory accounts, subscriptions, in-app purchases etc.
Even if it were possible to vote in the market (I deeply disagree though), the ultra rich have waaaaaaay more "votes" in the market to a point where anyone's vote doesn't matter.
Its interesting that you put it as "don't get to", I am pretty sure "normal people" don't care about having local computational power. We in this forum are not normal people, 80% of people just care about the end value computation brings. The future is likely devices with network chips attached to screens ans that's probably a better outcome for 80% of the world. There is no reason to think we pay the same amount of money to get worse results in the future, we can work towards one where we pay less because centralized computation and devices that are only screens and WiFi chips will be cheaper for everyone
To rub Altman's nose in the fact that we have RAM that he wishes he had.
Or, more likely (since I doubt he's ever going to see one of these emails), to make us feel a bit better about ourselves after Altman said (in effect) that what we want doesn't matter.
"A lot of people want local inference. And somehow, they assume that like the problem is solved now. Because we are going to have open-source models. We're going to be able to run them on our computers. And so we're not gonna need all the stuff in the data center. But from a global allocation perspective, actually, it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer. And being used 5% of the time."
I've seen the video where he says those words out loud, on mic.
Is this a script-kiddy DDOS (when many users run it) of someone's email? This doesn't seem very HN.
Edit: No, the email is blank, so it's just a regular DOS, unless his email is public somewhere and everyone will fill in the same email. Still not very HN, but just bad, not malicious.
We should allow it. I have around 15 computers across my home lab, desktop computer and laptop(excluding the toys like pi zeros). Letting him know that I'm not utilizing several hundred GBs of ram seems reasonable.
I am an old-school guy. I don’t send them emails with unused RAM, I just fax them a printout of the script so they have to physically look at it and remove it before it goes in the trash. Bonus points if you send it to legal. Legal always has a fax machine.
With what people at OpenAI make per hour, every fax costs them thousands of dollars in labor costs, wasting all their precious resources one toner cartridge at a time in the process.
I am just making proper use of all that underutilised toner.
If you do not store a Group on any of your Brother machine's One Touch or Speed Dial locations, you can Broadcast a fax to as many as 272 different numbers.
Which comes in especially handy for keeping their SI from exfiltrating via fax to a remote location one baud at a time. You just keep all their landlines busy until the army flattens the building.
It wasn’t nukes or bombs that saved mankind from Skynet.
It was a single fax machine keeping all of OpenAIs landlines busy for 25 minutes, sending hundreds of pages of hand drawn A4 sized dicks.
So, the guy who built an AI company that thrives on your personal information to sell you ads and your idea of fighting the said guy is voluntarily emailing him everyday more personal information?
IMO his comments are blown way out of proportion. He says "from a global allocation perspective" it's better to have RAM in the cloud where it can be used much more efficiently. I think there's an argument to be made that this is more than just economics to most people, it's about ownership and freedom, but he's clearly just talking economics.
Absolutely, I agree that money is better centralized in the goverment where it can be put to use for the benefit of all society instead of being concentrated in billionares.
While I agree there's certainly a better in between, if you go the other way you have moral hazard. If people aren't bearing the risks (or rewards) for how they allocate resources, history shows they allocate them poorly.
> IMO his comments are blown way out of proportion
The context for people's reaction to this is that he has single-handedly caused the current spike in RAM prices.
How? Elon and Dario aren’t responsible? It’s all AI companies and hyper scalers. Let’s not forget how 3 companies in memory industry have got caught in colluding to keep things high.
Because he (on behalf of Open AI) made deals with 2 of the 3 major manufacturers of RAM to buy a significant proportion of their output, in the same week, without either of the manufacturers knowing that he was also making a deal with the other one. This directly led to the current RAM shortage.
This problem is far bigger than one individual. He's making rational decisions is his capacity as CEO of an AI company. If he didn't make large compute deals, all his customers would be upset when they're hitting usage limits 10 times a day and all his investors would be doubting his decisions, especially when competitors are also trying to hoover up all the compute.
Is there actual evidence that these deals caused the RAM shortage? Seems like a stretch to pin a global supply problem on one person when demand is coming from across the entire industry.
And if one person can indeed cause it then blame the system? I go back to the proven legal case of them fixing prices. Now there is just a scapegoat because it feel good?
The system shares some blame, but the individual is also the one exploiting a confluence of factors. Should the system prevent one person from purchasing 40% of global RAM production with fake money, yes, it should. But our systems are glacial to respond to edge cases they were not prepared for, and this is an extreme edge case - people who are not Sam Altman could not pull off purchasing 40% of the world's RAM production capacity on an IOU. Only a very few individuals in the world could have exploited this gap in the system, and the one who actually did deserves blame for it.
Sure, but he's also insinuating that the economics of the situation is more important than the other, unmentioned aspects. In short, that his interests weigh more in what "best" means.
Everybody believes their interests weigh more. Fortunately we have capitalism, so at the end of the day you have to put your money where your mouth is and pay for what you think your interest is worth.
> Everybody believes their interests weigh more.
I don't. When I make decisions that might benefit me but affect others, I consider the impact on others before deciding what to do. Sometimes (often?) I choose the option that benefits me less (or is neutral, or hurts me), if I believe the negatives to others are too high.
> I don't. When I make decisions that might benefit me but affect others, I consider the impact on others before deciding what to do. Sometimes (often?) I choose the option that benefits me less (or is neutral, or hurts me), if I believe the negatives to others are too high.
I didn't say that means they don't consider others, and I haven't taken a survey of everybody on the planet.
Obviously people believe their interests weigh more.
Fun fact: I'm not a "rare outlier". People like me are pretty common and can be found all over. We usually just don't openly talk about it because we're not looking for a reward. I'm not speaking up now because I'm looking for self-gratification or anything. I'm speaking up to push back against your claims. I'm speaking up because we live in an increasingly selfish period of time. I'm speaking up because we're also at cultural turning point and many people are saying "enough is enough." I sure as hell don't care how rich you are, but I do give a damn about where the economic floor is. If you take from that floor to lift yourself up, I'm upset. If you do this to lift yourself when you already have more money than you could possibly spend in several generations? I'm furious.
I believe you. I'd challenge you though to consider Sam Altman would make the same claim about himself. The difference between you is not that you are good and he is bad, it's that you likely have very different beliefs of what the "good" course of action is. He likely believes successful adoption of AI means advances in science, medicine, freeing people up for more meaningful things, etc.
Sure, most people don't view themselves as the villain of their own story, but that's not what my comment was addressing. My comment was narrower
Fair, maybe I should say I think he wouldn't see his actions as taking from the economic floor, he would argue he is raising it. E.g. more effectively utilising resources (like RAM) means RAM is cheaper, which means more people can use it for their businesses and ideas.
Maybe a billion people would view you as having and owning and spending more than several generations of their families, maybe more. Yet you see wealth accumulation as a problem because there are those with much more than you. Are you the villain of your own story?
Well I congratulate you for finally having the courage to speak up about offering to take a picture for some people that's very commendable, but that doesn't really address what I wrote. I didn't say everybody weighs their own interests at 100% and everybody elses' at 0%.
Not sure if your arguments here is for capitalism. Having 1 vote = $1 is not a system most people would agree with today.
Sure. Better the capitalism that has Altman complaining about the peasants getting the RAM that he deserves better, than having the corrupt incompetent government and bureaucrats bought and paid for by him decide to allocate it "for the greater good" (i.e., his data centers).
Because these are literally the only two options?
Esp. for ML inference he's right, and not just for RAM. I think this should be thought out seriously instead of mocked, or else open models and local inference won't be able to compete with big data centers with exclusive contracts with hyper scalers. There ought to be some sweetspot where one pooled server could serve 50-100 people, with let's say 25 doing concurrent inference and seeing good throughput.
If DeepSeek-sized models will be required in the future to get Opus 5.5 quality, they're not feasible to run for a single person. Pooling usage is necessary to make it efficient.
But even that is pretty painful to sustain efficiently with current hardware prices:
Concurrent requests 25
Aggregate generation 1,000–1,250 tokens/s
Model DeepSeek V4.1 Flash, 552B MoE
Active parameters during decoding 16B
Candidate hardware 4 × H200, 564 GB HBM
Continuous rental at $2.90/GPU-hour $8,352/month
Cost per registered user $84–167/month
The efficiency of centralization is not enough excuse to deny individual agency.
The argument for saying that ram is put to better use collected into central hands vs distributed among individuals is the same as for any and all other resources, and it's invalid.
Humans are not ants and should not aspire to the efficiency of ants at the expense of having to be ants.
You get some centralization, and you get some distribution, and it's completely invalid to be jealous of the portion of all resources that is in distribution.
This should indeed be "thought out seriously" and the result is either mockery or torches and daggers. Think that out seriously.
I think there's a divide here in people who hear the words "it's more efficient than" and are interpreting that statement on its own (it's a true statement), and people who are assuming the statement also implies "we should only give it to data centers" (which I don't think anyone agrees with)
You misinterpreted my comment. I'm advocating for a middle ground where we may pool hardware in small groups for attacking the hyperscaler moat, specifically for ML inference.
> he's clearly just talking economics
...just economics that benefit him. The economics of rental for consumers are horrific. It doesn't even have to be about an abstract sense of ownership and freedom, even if it's "just economics" this is still an insanely self-serving to say because he's talking about economics that benefit 00.01% at the expense of the other 99.99%.
It's not necessarily "at the expense of" if people want to use those products every day.
Imagine if we had a shortage of cars: your city only had 20% of its current vehicles. It would be accurate to say "in this global shortage, new cars are more efficiently used by groups/taxis/uber than sitting idle in an individuals home". A car used by a group share would allow more individuals in the city to have access to vehicle transport. If cars instead all went to individuals, only a privileged 20% would have transport.
There simply isn't enough ram in the world for every individual to do local inference right now: one person having a 512GB RAM system means others don't. Data centers can batch the processing and serve many, many more at once.
That's not to say individuals shouldn't have it! individuals can buy a 512GB mac right now. Or that it should go to Open AI. Your org or group could do the same batching (OpenAI is only a minority of compute, still). But saying "it's more efficiently used" is trivially true
This falls apart because of the simple fact that the people who had to contribute to create the pooled resource do not get to say how the pooled resource gets used.
This argument that only considers the one positive dimension of centralization and pooling is never valid.
What model are you suggesting that gives them more access/control?
Currently AI providers are hyper competitive and almost all of the world has access to it quite cheaply. If people had to buy hardware for local inference, it would only be a tiny percentage of the world's affluent tech enthusiasts with access to AI, wouldn't it?
We're currently managing the shortage with a hybrid model, where enthusiasts can still buy (more expensive) individual systems and the broader shared hardware goes to many providers. OpenAI is something like 15% of data center power in the US I think? That's a lot but we still have choice.
kinda bad economica though - the ai bubble eating all the RAM is bad for many industries worth of useful production, including the infrastructure that makes gaming and movie industries possible nowadays
People at his level are typically mindful of how their words are perceived when doing interviews to the public at large.
That went out the window a few years ago. These tech oligarchs are getting more and more comfy with their optics being full psychoticism. They don't even know how to hide it anymore because all they think about and talk about and plan long short and medium term is extracting capitol at all costs for themselves.
Okay? So what's your argument, that we should just ignore them when they are being openly pathological? Sounds like that just let's them continue being crazy.
They watched Trump being able to say absolutely anything comes to his mind and figured that maybe they don’t have to pretend to be humans either.
This is typically the best argument for using cloud - it is the environmentally beneficial thing to do. And yet, knowing what I know of spotify, I do believe that me serving my own music library is a tad less wasteful than most commercial music streaming operations are.
RAM in the cloud ends up being used by GitHub actions runners that fetch and install the world anytime a shmuck pushes a commit to his confobulator feature branch for the paper clip factory. Multi-cloud deployments of a service nobody asked for. Enterprise architects deploying a service mesh to deliver what can be done in 50MB of memory on a raspi. Running some vectors representing natural language through gigabytes of noise only made useful by some arcane math only to be unable to count letters in words.
I am sure there's also lots of actually efficient use of memory out there. But when enterprises are willing to burn cash, efficient processes or methods are nowhere to be found.
Maybe we should be talking about the global allocation perspective of wealth then. Since Altman is so keep to optimize allocation, perhaps he should start with this aspect first, and then move to RAM, as wealth would impact a wider area of economics than RAM.
If we the people collectively owned "the cloud", then sure. But since we don't, I can only take it as him tryng to become a RAMlord seeking to extract rent from the rest of us
Yeah, Altman and the people in that world are utilitarians, which you would think they would have learned by now doesn't translate well to public conversation (or reality if we are being honest).
Nah. OpenAI themselves are wasting tons of RAM by doing frivolous stuff like providing ChatGPT for free for millions upon millions of people so they can ask what rocks go well with pizza. And wtf was Sora for.
No, if compute/memory was _that_ scarce that you begin thinking about what a waste it is that a decent chunk of it is sitting in people's homes, then why aren't you dedicating every atom of silicon that you already own to more productive efforts first.
Ah yes, I'm sure my RAM would be more economically useful if it were used to generate slop videos for Sora and inappropriate child content bty Grok rather than running VS Code and postman on my laptop.
Economically speaking, I’m sure it would be better if all energy were routed to him directly.
Probably water too, even food, money and air.
Obviously better.
the ridiculous "no ad blocker" full-screen modal that's unclosable led me down a rabbit hole there - turns out there's some obfuscated script from "Ad Shield" which replaces the whole site with an iframe if you dare to block their domain, "html-load.com", which proxies ad networks.
fortunately one can just run debugger; to stop it early
uBlock Origin + NoScript does the trick.
Using uMatrix/nuTensor to block JS from respawnfirst.com also lets you ignore the nasty script.
Looks fine to me.
Firefox, ublock origin. (on linux but I don't think that matters)
No noscript or special filters.
if you're on ubo lite, move from 'optimized' to full for that page and it shoooould dump adshield out.
I'm not getting it either, nor have I seen it in other sites. I wonder if this is because of Firefox not having neutered uBO?
That is correct, uBO Lite is A Chrome-only thing since they finally neutered uBO recently.
i've seen it personally on Orion on iOS, their adblocker seems to be unable to get around it.
Interesting. I'm using Firefox Focus and the page loads and is ridiculously free of distractions, err, ads.
I noticed it on several completely unrelated sites since a few days. I've never seen it before.
I was wondering if they're was some kind of consolidation/buyoff of pro-ad-protection services.
Or one could find the original post here <https://news.ycombinator.com/item?id=50412766>, which leads to <https://ram-altman.lol>
Or... one run noscript
Ah, I was complaining about the modal slide-in and other pester-widgets.
I am glad to not be the only one who hates that. Modern websites are so scummy.
Dupe? https://news.ycombinator.com/item?id=50414839
That's a dupe too. This is the original https://news.ycombinator.com/item?id=50412766
I would hesitate to say the memory used by the filesystem cache is unused.
I didn’t think we’d get to the point where all compute is headed for the data centre and normal people don’t get to have anything bigger than a phone but here we are.
Kind of inevitable when you realize how much people are naturally willing to give up computational sovereignty for the sake of convenience and psychological safety espousing values claiming the opposite, and it's still the case despite everything that's happened so far
I don't think there is a lot of conscious choice here: If at some point, anything more complex than a phone would be phased out or become unaffordable for consumers, that's not because consumers had somehow collectively voted they only want phones from now on.
Well sure someone could argue that the market isn't an actually reasonable set of choices that people can vote on, but how can it not be considered at least some form of voting? There's been at least 30-40 years in which consumers could try to find out. Unfortunately there's only a finite amount of resources in the market and consumers have to compromise on what takes priority in the tech industry, like getting work done or tackling issues that are visible only in front of them. When people collectively vote for politicians they also have to compromise
I don't feel extremely on either end, but I would prefer if consumers could own their choices and not try to pretend otherwise
Well, ok, what would you do in the current situation to "vote" for affordable RAM?
In a perfectly free and fair market and in a world where companies would plan their product strategy exclusively based on current consumer data, this would be true. But neither holds in the real world: The current AI gold rush is now something that is largely decided between investors, boardrooms and hardware vendors (*) with little input from small everyday consumers. And I wager this has been the case for many "tech trends" before that seem more forced on users than originating from them - such as forced updates, mandatory accounts, subscriptions, in-app purchases etc.
(* if you want to be provocative, in the US mostly by the people in this photo: https://apnews.com/article/ai-safety-openai-tech-week-32064f... )
Even if it were possible to vote in the market (I deeply disagree though), the ultra rich have waaaaaaay more "votes" in the market to a point where anyone's vote doesn't matter.
Its interesting that you put it as "don't get to", I am pretty sure "normal people" don't care about having local computational power. We in this forum are not normal people, 80% of people just care about the end value computation brings. The future is likely devices with network chips attached to screens ans that's probably a better outcome for 80% of the world. There is no reason to think we pay the same amount of money to get worse results in the future, we can work towards one where we pay less because centralized computation and devices that are only screens and WiFi chips will be cheaper for everyone
[dupe] https://news.ycombinator.com/item?id=50412766
Just what is the point?
To rub Altman's nose in the fact that we have RAM that he wishes he had.
Or, more likely (since I doubt he's ever going to see one of these emails), to make us feel a bit better about ourselves after Altman said (in effect) that what we want doesn't matter.
Have you not seen the endless posts pointing out that the RAM story was fake exaggerated nonsense?
"A lot of people want local inference. And somehow, they assume that like the problem is solved now. Because we are going to have open-source models. We're going to be able to run them on our computers. And so we're not gonna need all the stuff in the data center. But from a global allocation perspective, actually, it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer. And being used 5% of the time."
I've seen the video where he says those words out loud, on mic.
I have no unused RAM; if any shows up I spin up some nested Alpine VMs. I got 20 layers deep the other day.
Fucking Cry Harder, AI bros.
Watch out in case someone wants to SSH into those nested Alpine VMs using any 257 character password.
Joke's on them, there's 20 of them and they're really fucking slow. It'll take them a week to type a 257 character password.
Is this a script-kiddy DDOS (when many users run it) of someone's email? This doesn't seem very HN.
Edit: No, the email is blank, so it's just a regular DOS, unless his email is public somewhere and everyone will fill in the same email. Still not very HN, but just bad, not malicious.
Is you determining what is and isn’t HN, simply by if it’s “many users” vs “many companies” spamming someone?
Based on your argument, I would assume YC-funded AI crawlers spamming a website with requests is “very HN” and “not DOS.”
Isn’t this kind of an illegal crowd-sourced DDoS?
We should allow it. I have around 15 computers across my home lab, desktop computer and laptop(excluding the toys like pi zeros). Letting him know that I'm not utilizing several hundred GBs of ram seems reasonable.
No
> Your country is not allowed to access this resource
Well, maybe the country where I currently stay in is not allowed to access this resource. But I don't see any connection of this message to me.
What are the chances that they already do this :) Well, probably not given how much memory the harnesses use. (Claude too in fairness)
All your RAM are belong to us
I am an old-school guy. I don’t send them emails with unused RAM, I just fax them a printout of the script so they have to physically look at it and remove it before it goes in the trash. Bonus points if you send it to legal. Legal always has a fax machine.
With what people at OpenAI make per hour, every fax costs them thousands of dollars in labor costs, wasting all their precious resources one toner cartridge at a time in the process.
I am just making proper use of all that underutilised toner.
This guy knows what's up.
No toner on fax machines, just forever chemical paper that slowly muders their employees.
Obviously you do not know my trusty battle fax. The Brother Laser Fax 2840.
22 direct-dial and 200 speed-dial entries. Broadcast to up to 272 participants. Sexy.
Lots of companies still use Laserfax. dr2200 toner cartridge is 40 bucks for 12k pages.
> "Broadcast to up to 272 participants."
256 + 16? How do you get a number like that? Some trick with variable-width binary codes?
Not a binary encoding trick. It's just the sum of the fax machine's recipient capacities.
22 One-Touch numbers + 200 Speed-Dial numbers + 50 manually dialled numbers = 272
If you do not store a Group on any of your Brother machine's One Touch or Speed Dial locations, you can Broadcast a fax to as many as 272 different numbers.
Which comes in especially handy for keeping their SI from exfiltrating via fax to a remote location one baud at a time. You just keep all their landlines busy until the army flattens the building.
It wasn’t nukes or bombs that saved mankind from Skynet. It was a single fax machine keeping all of OpenAIs landlines busy for 25 minutes, sending hundreds of pages of hand drawn A4 sized dicks.
This guy faxes.
If you're going to that trouble you may as well create a DOI number for it so they think it's a rare book and scan and destroy it.
I think my signature will do for now.
Regards,
https://fbi.cults3d.com/uploaders/15196480/photo-file/19c974...
Seller of cloud inference warns you should not have your own local inference
I am sad that it is not configured to actually email him by default
I think Sam owes us money.
Also - this will probably age as well as Bill Gates and the 640kb statement, which he never stated in that way, but will forever be linked to him.
Sam RAM-boy Altman. Or Sam-RAM-man.
Wouldn't unused RAM infuriate him more ?
That’s the point.
More like SCAM Altman.
So, the guy who built an AI company that thrives on your personal information to sell you ads and your idea of fighting the said guy is voluntarily emailing him everyday more personal information?
It doesn't actually email anyone anything, does it?
It's satire.