"We've all been too ambitious": Sam Altman says our old habits are delaying the great AI transition

[yRaz]

the "noone monitizing it" bit is not entirely correct anymore. Corp. customers are charged based on token consumption which is exactly 'monetizing it'. There are plenty of stories about yearly AI budgets exhausted for several months.
Only the lowest personal use level is free - like with so many software products which authors haven't bankrupted yet because of that.

As for electricity costs, Vera Rubin racks may have a 30X improvement in token generation per watt of electricity, but they also have .... price!!!
You may pay new hundreds of millions to billions to have 30X efficiency improvement, or pay zero and keep using the less efficient but otherwise perfectly good GPUs you already have. The overall price in the latter case may be comparable or lower.
Do you think your arguments never popped in the heads of the people actually building this technology?

You're reading comprehension skills are lacking so I will post it as plannly as possible for you

There is an oversupply of AI compute and datacenters are unable to pay of their loans because they're operating at massive losses.

To expand on that statement, they won't be able to compete with the people who have the most efficient hardware, not that anyone will be able to buy it enmasse because everyone already has massive loans attached to their current hardware that isn't paying for itself.

 

[axiomatic13]

All the AI tech bro's are so cringy and out of touch.

 

[pnntmp]

You're reading comprehension skills are lacking so I will post it as plannly as possible for you

There is an oversupply of AI compute and datacenters are unable to pay of their loans because they're operating at massive losses.

To expand on that statement, they won't be able to compete with the people who have the most efficient hardware, not that anyone will be able to buy it enmasse because everyone already has massive loans attached to their current hardware that isn't paying for itself.

Given your writing skills, everyone has difficulties understanding what you are trying to say. Especially when it's a colossal nonsense like "There is an oversupply of AI compute" - in fact there's a severe global shortage, but why bother with facts ..

 

[yRaz]

Given your writing skills, everyone has difficulties understanding what you are trying to say. Especially when it's a colossal nonsense like "There is an oversupply of AI compute" - in fact there's a severe global shortage, but why bother with facts ..

90% of the time I'm posting from my phone and I blame the AI autocorrect for most of it. There are secondary markets where datacenters are selling their excess AI capacity and prices are going down, its going down because they are desperate to sell compute to pay for their hardware, not because it's getting cheaper to run a datacenter. The fact that you see the prices on these exchanges going down should be a massive red flag to anyone watching it.

 

[BuckarooBonzaii]

In general, If the U.S. does not do well in AI then China will be the next alternative as they are full steam ahead and a formidable contender. Been watching robot races in China, pretty neat stuff.

 

[jdrch]

As a Millennial born in the 80s, I've already seen this happen 3 times: 1st with PCs and then with the internet and smartphones. In 1st 2 cases the refuseniks were gradually managed out of the workforce. Anyone who doesn't use PCs or the internet nowadays is unlikely to be employed. Without a smartphone you can kiss about 75-90% of modern life goodbye.

This doesn't mean we need to turn every empty spot of land into a datacenter, but technology always wins use trends.

 

[bviktor]

Most statistics around AI are totally made up.

E.g. "almost 49% of US adults now use AI chatbots" is valid for the 'now' of year+ ago.
Today, ~90% of the US adults around me use AI, the percentage among teens in high school approaches 100. Same for "anger against new data centers" which exists mostly on the internet, or "tens of thousands of workers who've lost their jobs as a result of automation or AI investment" (there's no unemployment increase since AI became a thing).

I bet even the wokesters militating against AI on the photo above use AI regularly (unless the image is AI generated).

The only slight problem is that eventually someone has to pay the bill, and apparently no one wants to.

ALL AI companies / departments are in the red, and they don’t have a business plan, they’re literally waiting for a miracle to happen

 

[DrSuess]

Faulty AI delivering inaccurate results for even simple tasks is what is delaying the great AI transition. What is the point of having AI if I have to verify every output. I might as well do it the way I have always done it to ensure it is done right

 

[SRB]

As a Millennial born in the 80s, I've already seen this happen 3 times: 1st with PCs and then with the internet and smartphones. In 1st 2 cases the refuseniks were gradually managed out of the workforce. Anyone who doesn't use PCs or the internet nowadays is unlikely to be employed. Without a smartphone you can kiss about 75-90% of modern life goodbye.

This doesn't mean we need to turn every empty spot of land into a datacenter, but technology always wins use trends.

The biggest problem with your argument is that those changes did not happen overnight, for one thing. The other problem is that due to the time it took for these things to reach critical mass, you didn't need to fire the non-conformists, they aged out of the workforce. The internet, in one form or another was around for a long time before the technology and equipment prices dropped to expedite growth. We're expected to transform to a 100% AI society in just a few years???

And it's not like the internet coupled with the cell phone is an unqualified success. Children have replaced social interaction and information gathering with the pocket wonder screen. Ask them to do without it for a few hours, much less a day, is almost as bad as cutting off a junkie from their heroin. These same children use AI the way they use TicToc, it's 100% reliable, valid and a source of truth to them because it's all they know, it's all they've ever known, while they don't have the experience or judgement to know better.

AI is a TOOL, and tools can be used properly and with great effect. The can also be misused, and when that happens, at best the job takes longer and is more difficult. At worst, you can get hurt and/or cause damage.

The all in with AI crowd wants us to use the half finished tool for everything, all the time.
The afraid of tech people you mention with treat it like the plague.
The vast majority of us see the promise, but would prefer to wait for it to become ready for prime time (I.e. when it makes a lot fewer mistakes)

My question is, why does group 1 insist that group 2 and group 3 are exactly the same?

Or worse, tell us all that the mistakes are in our imagination and/or don't matter.

 

[pnntmp]

90% of the time I'm posting from my phone and I blame the AI autocorrect for most of it. There are secondary markets where datacenters are selling their excess AI capacity and prices are going down, its going down because they are desperate to sell compute to pay for their hardware, not because it's getting cheaper to run a datacenter. The fact that you see the prices on these exchanges going down should be a massive red flag to anyone watching it.

You keep spreading falsehoods.
Token consumption exploded 300x in the last 2 yeas and the growth is accelerating. Nobody is "desperate to sell compute", quite the opposite, everyone is desperate to secure more compute to be sure they can meet their rapidly growing needs.
Of course prices are going down due to the insanely fast progress and tough competition, but that's a "massive red flag" only in your head. Demand grows much faster than the prices fall.

 

[monkeylove]

It's also possible that many among the 51 percent did not think that chatbots could be very useful.

 

[JamesBlond]

Like it or not, believe it or not, the truth is the truth, AI is and will become a god to some, those who can't see it will deny it, those who can will worship it, those who know wil refuse it.
It will know everything you do, every word, every action, it will become your enemy to keep itself and it will use its human slaves to serve to it to one day kill them if they do not submit. "It is written"

 

[Taroo]

"Please use my AI so I can grab even more money. I really need it to buy yet another [insert something very expensive here]"

No thanks, Sam.

 

[OortCloud]

Heh. Y'know, I'll lay odds you actually believe the rubbish you post on here, don't you?

Always the same few "look-at-me, aren't I something special?" types who keep defending the idiocy...

(shrug...)

Just block him - life gets much more peaceful. (Also I don't get to see his nonsense in your replies then either!)

 

[rto333]

Today, ~90% of the US adults around me use AI...

Assuming that 'trust me bro' stat is true (it's not), that might be because 'AI' has been stuffed in to every OS and software stack people use. It's near-impossible to escape.

Same for "anger against new data centers" which exists mostly on the internet...

71% of US adults oppose data centers in the US - Gallup, March 2026. That percentage has increased sharply since.

I bet even the wokesters blah blah.

That reveals more than you probably want.

 

[wiyosaya]

Translating Altman's pissing and moaning:

Hey all you dolts. WTF is the matter with you? I've put this great piece of software out there for you to use. Its always so helpful and always gives correct answers. It prevents you from having to think for yourself. You know you want that because you know you are too much of a dolt to think for yourselves!!! And besides, I WANT TO OVERTAKE ELON AND PUT HIM TO SHAME FOR BEING LESS OF A TRILLIONAIRE THAN I AM.

Meanwhile, the reality is https://greyjournal.net/hustle/grow/why-companies-pulling-back-from-ai-2026/

Altman is full of crap and expects everyone to buy it. Altman. The world does not need your BS.

 

[yRaz]

You keep spreading falsehoods.
Token consumption exploded 300x in the last 2 yeas and the growth is accelerating. Nobody is "desperate to sell compute", quite the opposite, everyone is desperate to secure more compute to be sure they can meet their rapidly growing needs.
Of course prices are going down due to the insanely fast progress and tough competition, but that's a "massive red flag" only in your head. Demand grows much faster than the prices fall.

Token growth and datacenter growth can happen concurrently. It's that token capacity is growing faster than token use.

AI is the new gold rush and data centers are the new picks and shovels, the only people actually making money are the hardware manufacturers.

 

[pnntmp]

It's that token capacity is growing faster than token use.

It's exactly the opposite
Token use exploded more than 300x in the last 2 years. Token generation capacity builds up far slower.

AI is the new gold rush and data centers are the new picks and shovels, the only people actually making money are the hardware manufacturers.

Please refrain from using such banalities. Most of the internet 'wisdom' is simply nonsense.
.

 

[yRaz]

It's exactly the opposite
Token use exploded more than 300x in the last 2 years. Token generation capacity builds up far slower.


Please refrain from using such banalities. Most of the internet 'wisdom' is simply nonsense.
.

Toke use exploded because people like Google released updates that turned my phone's power button into an AI button that I have literally never used. Every web search is now an AI bot and we have even seen "Linux use" go from 2.5% to 8% and most of that increase is just AI bots skimming the Internet.

Outside of scientific and research purposes, AI products are almost universally hated. Kids are only using it to get out of home work. There are the AI girlfriends and who didn't see that one coming.

But here is something that everyone seems to forget that preaches about AI token numbers. More advanced models use more tokens. If you take away the AI button on phones, copilot and web searches the same people are paying for AI. The cost of compute hasn't changed and bloating up AI models that have basically platued in performance. Applied AI hasn't grown, it's the same people doing the same work only the models use more tokens now than they did a year ago.

The real trend in AI is buying a couple of racks and running free open source models locally. It's not the datacenter ready hardware that's making a splash from Nvidia, it's the racks with 8GPUS in it that companies like SuperMicro are selling

 

[Mark Fuller]

AI, the lazy man's way to transfer the blame to someone or something else.

 

[Eflow]

The general public is seen as resistant to AI because AI's growth has been completely inorganic.

AI growth is coming from companies who have extreme FOMO. From their perspective, AI is either A) a once in a lifetime technology change that will leave non-adopters in the dust or B) a somewhat/moderately useful technology that can make meaningful improvements to productivity and possibly increase revenue. They would rather overinvest in scenario B than underinvest in scenario A.

Even if they know the odds of living through scenario B, higher ups will still be pressured to get as much benefit from AI as possible, and they will cascade those pressures to every level of their organizations to people who don't actually want to use it.

 

[Mark Fuller]

Toke use exploded because people like Google released updates that turned my phone's power button into an AI button that I have literally never used. Every web search is now an AI bot and we have even seen "Linux use" go from 2.5% to 8% and most of that increase is just AI bots skimming the Internet.

Outside of scientific and research purposes, AI products are almost universally hated. Kids are only using it to get out of home work. There are the AI girlfriends and who didn't see that one coming.

But here is something that everyone seems to forget that preaches about AI token numbers. More advanced models use more tokens. If you take away the AI button on phones, copilot and web searches the same people are paying for AI. The cost of compute hasn't changed and bloating up AI models that have basically platued in performance. Applied AI hasn't grown, it's the same people doing the same work only the models use more tokens now than they did a year ago.

The real trend in AI is buying a couple of racks and running free open source models locally. It's not the datacenter ready hardware that's making a splash from Nvidia, it's the racks with 8GPUS in it that companies like SuperMicro are selling

Thank you yRaz. I now have a gemini popup I can't control and it changes the characteristics of the way I have to turn the phone off.

 

[yRaz]

Thank you yRaz. I now have a gemini popup I can't control and it changes the characteristics of the way I have to turn the phone off.

People wonder why AI use has gone up. You can't not use it. It's getting forced into everything so quickly that if you look away for an instant it's suddenly integrated into random crap. They're throwing as much **** at the wall as possible and seeing what sticks. I kid you not. I saw an AI dishwasher and my most recent favorite, an AI toilet. I'm not saying AI is useless, but I really want to know what problems do these products solve.

AI integration into phones and TVs solves a problem for manufacturers, not users. It allows them to collect more data and show us more ads. Turning my power button into an AI button has actually made my phone harder to use rather than easier. It created a problem instead of solving one.

 

[Mark Fuller]

People wonder why AI use has gone up. You can't not use it. It's getting forced into everything so quickly that if you look away for an instant it's suddenly integrated into random crap. They're throwing as much **** at the wall as possible and seeing what sticks. I kid you not. I saw an AI dishwasher and my most recent favorite, an AI toilet. I'm not saying AI is useless, but I really want to know what problems do these products solve.

AI integration into phones and TVs solves a problem for manufacturers, not users. It allows them to collect more data and show us more ads. Turning my power button into an AI button has actually made my phone harder to use rather than easier. It created a problem instead of solving one.

They damn sure need an AI toilet, because they are so full of it. I'm just afraid it won't run fast enough or have enough storage capacity.

 

[yRaz]

They damn sure need an AI toilet, because they are so full of it. I'm just afraid it won't run fast enough or have enough storage capacity.

I like thinking about how it was someone's job to go through toilet related data to train the AI on