Sam Altman hates Pc users using and hoarding Ram

It's better to be thought of as a fool then to open your mouth and remove all doubt.

I remember sam was on joe rogan show and said years ago that Ai would first replace blue collar jobs. I knew right then he was a moron.
 
I had somewhat mixed opinions of Sam Altman prior to this, but wow.... His comments remind me of that Pluribus TV show, where when the hivemind took over, everyone abandoned their homes and started sleeping next to each other on the floor in the middle of large warehouses instead, simply because it was more efficient. Sam Altman would probably be fine with that also, as it would save a lot of power for his data centers. If he doesn't understand why people want their own computers, he probably doesn't understand why anyone values any personal possessions or personal freedom either. Sam Altman is so detached from reality, I believe that he would be genuinely confused if people did not rush to pay a subscription just to smell his farts.
 
I'm kind of a traditional free-market guy, but I really would like to do something about exploitative rent-seeking.

We are on a path to a very bad place.
 
Much in the same way that IBM could do little to prevent the downfall of the dominance of mainframe computers. Sam, despite his attempts at monopolising AI access by buying all available RAM supply and then storing it in a warehouse, will not be able to prevent the downfall of the centralised AI subscription monopoly.
 
well, i would much more prefer him nowhere near me...

but if that mouth seems like something you wanna go for - i won't stand in your way :)
 
LukeTbk said:
And that a complete different subject, Altman point is so simple and true it is an strawman I think, no one disagree with it, the solution to the AI ram shortage isobviously not every AI consumer building is own way less efficiant AI infrastructure that sleep doing nothing when they do not work, yours of Ai consummer should consume 100 to 10,000 time less AI, so many time less than it can run on an RTX A1000, well of course, if we diminish IA use by a factor of 1,000 that would help.
Click to expand...
I very much disagree with Sam Altman. I'm sure large scale Data Centers are more efficient at AI than someone's computer at home, but that's retarded. That's like saying buying a new car is more efficient than driving an old car. It simply isn't true.

  1. My computer does other things when AI isn't one of them, and so far it isn't. Sam Altman's computers are converting electricity into shareholder confidence and nothing more. The demand for AI is not enough to justify the electricity, let alone the DRAM usage in Data Centers.
  2. You still need real people who make real content to train the AI on. Which means you still need real computers that have ram. Training AI on AI is akin to incest. It works for a while until it doesn't.
  3. Cloud AI has real limits. Go ahead and see which AI allows you to make fun of Israel, because a lot of them won't. Some cloud AI isn't letting people use them to make game decomps. LLM's are the only way to go if you want true freedom when using AI.
anthropic cracking down on game decomp.jpeg
 
DukenukemX said:
I very much disagree with Sam Altman. I'm sure large scale Data Centers are more efficient at AI than someone's computer at home, but that's retarded.
Click to expand...
so you 100% agree, not only they are more efficiant when they run, but they do run more hours by days.

Sam Altman said nothing at all, no one actually disagree with the very simple statement he made, but as you said it is a bit of a strawmen, outside very uninformed or people that did not though much, no one actually believe that evreyone building his own AI infracstructure to run all their wanted ai workload on 2027 themslve, would help ram become cheaper.

DukenukemX said:
My computer does other things when AI isn't one of them, and so far it isn't.
Click to expand...
Yes, as long that the datacenter usage rate is near 100% that a bit irrelevant and you do not need to buy a lot of memory or sram specialized ASIC ai chips for those others things.

DukenukemX said:
Sam Altman's computers are converting electricity into shareholder confidence and nothing more. The demand for AI is not enough to justify the electricity, let alone the DRAM usage in Data Centers.
Click to expand...
How did Anthropic sold for about 17 billions of AI service at a 9.9 billions compute&infracstructure&electricity cost to deliver those service if that the case ?

The demand in AI not being enough to justifiy the electricity seem a bit strange, almost all the electricity used is now inference, which is simply the expression of the demand for AI.

DukenukemX said:
You still need real people who make real content to train the AI on. Which means you still need real computers that have ram. Training AI on AI is akin to incest. It works for a while until it doesn't.
Click to expand...
Less and less, AI is not that much trained on premade content-data anymore (on why it is getting better), they give an AI an objective and it trains by trying to do the task, scoring euresthic by doing a lot of trying and error, which way gave the best results, cheapest way to do it and so on, nobody debate you still need real computer to run ram, all Sam Altman user are using its product from a computer with ram, it is about buying a bunch of Spark to have 1TB of ram at home to run your own model being a way that will make the global ram shortage situation better or worse than using a more efficiant and shared with other cloud.

Yes an anthropic-meta-openai has a lot of user that accept to pay less for their service in exchange of training on what they do and they do it with a computer, but Sam altman of course has no problem with everyone having a computer, that not what this is about at all.

DukenukemX said:
Cloud AI has real limits. Go ahead and see which AI allows you to make fun of Israel, because a lot of them won't. Some cloud AI isn't letting people use them to make game decomps. LLM's are the only way to go if you want true freedom when using AI.
Click to expand...
You mistake cloud AI with commercial solution provider ai like chatgpt, if I run Deepseek 4.1 on my computer and make it available for you over the web, you are using cloud AI now and you will have the exact same result as if you ran Deepseek on your own computer at home, when you rent microsoft-AWS-google cloud infracstructure to run AI, they have no idea what you are doing, which model you are running and so on.
 
LukeTbk said:
so you 100% agree, not only they are more efficiant when they run, but they do run more hours by days.
Click to expand...
No, because all they do is try to make AI less retarded. ChatGPT isn't even the top dog in AI anymore. Sam Altman has spent so much resources to not even have the best AI.
Sam Altman said nothing at all, no one actually disagree with the very simple statement he made,
Click to expand...
That sentence...
suction cup man eyes.gif

but as you said it is a bit of a strawmen, outside very uninformed or people that did not though much, no one actually believe that evreyone building his own AI infracstructure to run all their wanted ai workload on 2027 themslve, would help ram become cheaper.
Click to expand...
His problem is that people are using LLM's over cloud based AI like ChatGPT, which has made DRAM even more expensive even for Sam.
Yes, as long that the datacenter usage rate is near 100% that a bit irrelevant and you do not need to buy a lot of memory or sram specialized ASIC ai chips for those others things.
Click to expand...
100% doing what though? They don't make enough money to pay even a fraction of the debt they owe.

View: https://youtu.be/roe3SgezmmU?si=byGKEH6fRnSe5TFH
How did Anthropic sold for about 17 billions of AI service at a 9.9 billions compute&infracstructure&electricity cost to deliver those service if that the case ?
Click to expand...
Anthropic owes $900 Billion. $17 Billion is a drop in the bucket.
The demand in AI not being enough to justifiy the electricity seem a bit strange, almost all the electricity used is now inference, which is simply the expression of the demand for AI.
Click to expand...
Or they're training AI? It's a huge waste of resources.
 
You missing the point. AI as a whole is not need nor wanted. It is a huge drain on resources. Not to mention people are forcefully made supplement the cost through higher water, electricity, noise and so on form these DC. Also the cost of all electronics have been steadily rising in price with seemingly so end in sight on how high it will go. Fuck Sam Altman and his AI bullshit.
 
Altman is a rent seeker bozo, but to address his point about % utilization, I was just saying that it would be nice to pool our resources together with some sharing of compute with some peer to peer credit system. That way we can us the trillion+ param local models (in a distributed way if possible) without needing to use their stuff where they can control the pricing and can tell us what we can or can't do (like reverse engineer games).
 
SecretStash said:
Altman is a rent seeker bozo, but to address his point about % utilization, I was just saying that it would be nice to pool our resources together with some sharing of compute with some peer to peer credit system. That way we can us the trillion+ param local models (in a distributed way if possible) without needing to use their stuff where they can control the pricing and can tell us what we can or can't do (like reverse engineer games).
Click to expand...
Which is exactly why they won't let us have it. Look at 3d printers trying to be locked down they won't allow us to have stuff that might bypass their control
 
vegeta535 said:
You missing the point. AI as a whole is not need nor wanted. It is a huge drain on resources. Not to mention people are forcefully made supplement the cost through higher water, electricity, noise and so on form these DC. Also the cost of all electronics have been steadily rising in price with seemingly so end in sight on how high it will go. Fuck Sam Altman and his AI bullshit.
Click to expand...

Not wanted by who? Most normal people? Yeah, probably, though they like it too, but more in the sense of asking it banal questions. They are still a lot more skeptical about it overall but are mostly just NIMBYs. Overall they are OK with data centers... just not near them.

The large capitalists? Oh, they definitely want it. Human labor is their largest cost. They definitely want to eliminate that as much as they can, and if AI is the means to that end, then they want it and will 100% pursue it. Verizon gives less than 1 shit if your cost of water or electric goes up, that's your problem, not their problem. They'd love to scrap 90% of their support team though. If your electric rates goes up and your access to drinking water goes down, that's a price they are willing to pay.
 
mullet said:
Wait till the elites, start using robots everyday. Then they have no need for humans. Do the math on that.
Click to expand...
That is the part that really really really scares me. If they can have an AI that builds them stuff, wipes there ass, fixes them when they are sick and collects food for them ECT they have motivation to actively kill us off so they don't have to share the land with 8 Billion people. COVID was probably a test run to figure out the optimal way to spread something
 
tunatime said:
That is the part that really really really scares me. If they can have an AI that builds them stuff, wipes there ass, fixes them when they are sick and collects food for them ECT they have motivation to actively kill us off so they don't have to share the land with 8 Billion people. COVID was probably a test run to figure out the optimal way to spread something
Click to expand...
.tenor.com%2FTwTKD9LK7TUAAAAM%2Fyes-jack-nicholson.gif
 
IT WAS ME!
 

Attachments

  • RDT_20261031_211715326611627177449667.png
    RDT_20261031_211715326611627177449667.png
    315.2 KB · Views: 0
GotNoRice said:
If he doesn't understand why people want their own computers,
Click to expand...
This is so far away from his comments, all Sam altman clients need to own a computer to pay openAI to use its product.
This is all a click-bait for profit cycle (of course), and everyone want to not understand what going on for the fun of being mad here.

What he said was:
An example of this is people...
A lot of people want local inference.
And somehow, they assume that 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 going to need all the stuff in the data center.
But from a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer.
And, being used—Mm-hmm. 5% of the time.
You want maximum utilization in the cloud if you're trying to deal with this very supply constrained environment.

i.e. It is about people thinking that having 512gb-12 TB of ram to run lage local model with big context using it only 500 hours a year at full capacity, instead of people sharing that ram with each other for when they need it is a not a good global allocation.

Nothing to do with everyone wanting to or having both a phone and a Macbook
 
vegeta535 said:
You missing the point. AI as a whole is not need nor wanted.
Click to expand...
I am not sure what not wanted would mean here, it is the fastest adoption of a new tech by regular people all over the world of all time, the revealed preference is hard to deny. The most bottom up has well, pushed by regular people into the world more than anything we seen before.

Internet-railroad-electricity-PC, there was a long controlled moment from government-military-industriel cartel gatekeeping, generative AI was on github right away, with everyday consummer demands-usage orienting the industry aggressively right away.
 
Looks like I'll kick off my own local AI VM just to
LukeTbk said:
This is so far away from his comments, all Sam altman clients need to own a computer to pay openAI to use its product.
This is all a click-bait for profit cycle (of course), and everyone want to not understand what going on for the fun of being mad here.

What he said was:
An example of this is people...
A lot of people want local inference.
And somehow, they assume that 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 going to need all the stuff in the data center.
But from a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer.
And, being used—Mm-hmm. 5% of the time.
You want maximum utilization in the cloud if you're trying to deal with this very supply constrained environment.

i.e. It is about people thinking that having 512gb-12 TB of ram to run lage local model with big context using it only 500 hours a year at full capacity, instead of people sharing that ram with each other for when they need it is a not a good global allocation.

Nothing to do with everyone wanting to or having both a phone and a Macbook
Click to expand...
Playing devil's advocate for the big AI corpo overlords doesn't mean it's right regardless. The greater good is a BS philosophy used as a blanket to cover the continuous sins of those committing them & require more from others who refuse to go along with it.
 
westrock2000 said:
The current situation is that there are a lot of people duplicating small models to run locally. If you took that same amount of collective memory and instead used it to hold one gigantic model that many people could access, that would be a more efficient system and give better results. Does that idea go against every historical precedence of human nature? Of course. But it doesn’t mean it’s not accurate.
Click to expand...
I doubt that it is more efficient.

AI inference is not scalable, like the web was. You can host a thousand web sites on a single PC in a datacenter, but AI inference will have the same compute and RAM cost no matter the scale. So it's not like the bigger the data center the more efficient it runs, it might even be less power efficient to run in a datacenter than on a simple PC.

The small models people are able to run at home are perfectly suitable for most home use cases. You don't need a gigantic model to ask the weather. I'm confident that the majority of normies who pay for ChatGPT or Claude, wouldn't notice the difference if their queries were redirected to a small open weight mode that their PC can run. Worse, many normies use ChatGPT for image generation, churning out piss colored slop. ChatGPT image generation is stuck at about the level of SD1.5. There are open image models that are far more capable and efficient than their "gigantic" model.

And it doesn't need a gigantic data center, it doesn't need fresh water for cooling, it doesn't need gas burning generators for power. It doesn't drive up the cost of every consumer electronics on earth. Because mind you it's not only PCs that have RAM in them. I'm pretty sure the RAM in most PCs are better utilized than the RAM in the latest iphone, that they buy for bragging rights, not because they need it.
 
  • Like
Reactions: kac77
like this
CrimsonKnight13 said:
Playing devil's advocate for the big AI corpo overlords doesn't mean it's right regardless. The greater good is a BS philosophy used as a blanket to cover the continuous sins of those committing them & require more from others who refuse to go along with it.
Click to expand...

Its more about physical reality of what possible versus anything moral, it is not really just some greater good that everyone that built local AI capacity that they do not use fully most of the time to rent it (i.e. share it) with others if they can, until something change in ram supply or how AI work it is the only way to give access to capable AI to 4-5 billions people.

People can of course have different axis and think AI should not be commonly used at all, but those people do not want everyone to build a MacUltra 256/512GB or more capacity either anyway.
 
M76 said:
AI inference is not scalable, like the web was. You can host a thousand web sites on a single PC in a datacenter, but AI inference will have the same compute and RAM cost no matter the scale. So it's not like the bigger the data center the more efficient it runs, it might even be less power efficient to run in a datacenter than on a simple PC.
Click to expand...
Ram wise compute speed can affect it, if a task take 1 hours instead of 6 hours to do, you need 1/6 of the ram on perfect timing (it will never be that much being saved, but you get the point), Rubin GB200 system is so much more efficient than blackwell who was so much more efficient than Hopper and so on, the level of custom kernel-effort that go to optimise token per watt into those giant datacenter, the superiority of the dedicated hardware for it, make that almost impossible.

Rack scale architecture efficacy is magnitude higher than simple PC at home, usually in the 10-25x more token per watt than MacStudio's ultra.

M76 said:
The small models people are able to run at home are perfectly suitable for most home use cases.
Click to expand...
always hard to talk about closed door model efficacy, but GPT 6 luna is so cheap to use (very similar per token price than a small Qwen model) and more token efficient, that people that could use a small model at home can choose that (or their routing system picking it) and will be way cheaper, cloud or at home does not need to change anything on the nature of the model used (you can use Qwen3 30b on the cloud or at home), that a bit 2 different subject.
 
LukeTbk said:
Its more about physical reality of what possible versus anything moral, it is not really just some greater good that everyone that built local AI capacity that they do not use fully most of the time to rent it (i.e. share it) with others if they can, until something change in ram supply or how AI work it is the only way to give access to capable AI to 4-5 billions people.

People can of course have different axis and think AI should not be commonly used at all, but those people do not want everyone to build a MacUltra 256/512GB or more capacity either anyway.
Click to expand...
Four to five billion people don't need access. Supposing everyone does, then it only needs to reach the same point where it's good enough on a local machine for the average user.
 
westrock2000 said:
The over dramatic responses here are only eclipsed by the over dramatic article.

Individuals wanting their own stuff to use and stuff being efficiently optimized for usage are two separate things.

The current situation is that there are a lot of people duplicating small models to run locally. If you took that same amount of collective memory and instead used it to hold one gigantic model that many people could access, that would be a more efficient system and give better results. Does that idea go against every historical precedence of human nature? Of course. But it doesn’t mean it’s not accurate.
Click to expand...
It's not really that everyone is "sitting" on ram though. It's that they are sitting on untapped, and already purchased resources that they themselves could use. It's more the reality that no one has offered a product that fully allows on-prem installation of the same resources in an easy way. Just because it's in the cloud doesn't make it more efficient. Because of the amount of ram and GPUs being used it's harder to support AI in the cloud not less. Inference will use every bit of ram that you can give it and each random query takes up its own bit of memory which can cause growth in utilization. Compare that with say a server for home use. The number of queries are drastically reduced, plus they are usually similar because they're the same queries over and over. A GPU with 16 - 24 GB of RAM can serve a household. That's not out of the question at all. Before the AI boom all of us could easily assemble a computer that could handle AI quite easily for a pretty reasonable cost.

What's really crazy is that MANY laptops come with nVidia, AMD, or Intel GPUs, which means most people are sitting on their own AI assistants w/o realizing it because no one has designed a product to use it other than Microsoft. That's really the problem because people go out of their way not to use MS OS features because of the problems they cause. Local on-prem use of AI assistants could work if it wasn't Microsoft pushing it.

Then you have energy usage. It's much easier to deal with increases of energy demand that's spread out. Allocation happens gradually vs. a big bigger than Walmart space that uses 10X the electricity and water plopped down in the middle of no-where or near residential spaces.

I view this no differently than cloud services for companies that have already built out their infrastructure. If you've already purchased it, keep it. If you want new construction to move there, then fine, but don't pay 3 times the cost just to move something you already have that's working. That's crazy in an economical environment where your ideas, thoughts and talent can be sucked up and utilized w/o your permission.

Consolidation works when demand for each action is minuscule in comparison to a larger operation asking for the same demands. AI is NOT that at all. That's why no one started really complaining about data centers until the resource demands went through the roof. That's also why cloud gaming has taken so long to build. It's resource intensive just like AI.
 
Last edited:
You must log in or register to reply here.
Back
Top