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

midian182

Posts: 11,971   +183
Staff member
A hot potato: AI advocates have long said that the technology will change our lives at an unprecedented pace, disrupting the status quo and thrusting us into an unpredictable and scary future. That hasn't happened, at least not yet. In fact, anti-AI sentiment seems to be growing. Even OpenAI boss Sam Altman admits that "we've all been too ambitious on timelines."

A recent survey found that almost 49% of US adults now use AI chatbots. While that is up from 33% two years ago, it's not as all-encompassing as, say, the number of people who use the internet every day.

That figure is likley a lot lower than Altman and other AI evangelists expected. Speaking to podcaster David Senra, the OpenAI boss seemed to blame some of this on people's stubbornness, though he thinks we fleshy humans will eventually see the light and abandon our tried and trusted old ways.

"I think I was wrong about a few things," Altman admitted. "One of them is just [the fact that] the economy just has so much inertia. People keep doing the same things...they're buying from the same company, they keep wanting to use their tools in the same way. I think that's actually a positive in many ways, and it's going to make this big transition in front of us go smoother and slower, and I'm grateful for it. But I think it means we've all been too ambitious on timelines."

Offering an example of this old-habits-die-hard behavior, Senra said he recalled being surprised that people still used Blockbuster even after Netflix started offering DVD delivery – a seemingly superior service it originally provided before becoming a streaming giant. Altman agreed.

The number of people using AI has grown alongside the pushback against it. The anger against new data centers and the problems they introduce is causing plenty of headaches for tech giants. There are also the tens of thousands of workers who've lost their jobs as a result of automation or AI investment. There are concerns about AI causing major damage, too, amplified by the recent incidents of agents going rogue. And online AI slop has become so prevalent that it's making people fall out of love with the internet.

Altman thinks this sort of resentment is partly normal and healthy, arguing that we are always afraid of rapid socioeconomic change and comparing the current situation to the Industrial Revolution.

The CEO argued that the AI industry has amplified those fears by warning of existential risks and mass job losses while failing to explain the technology's benefits and how the downsides can be mitigated.

Promises of universal basic income or an end to compulsory work also miss what many people value most, he said: autonomy, personal freedom, and a say in their future.

Altman called the prospect of a few companies controlling AI in exchange for providing safety and abundance an "anti-human" sales pitch. Instead, he believes wider adoption will depend on empowering individuals, including making entrepreneurship more accessible and triggering what he predicts will be an unprecedented boom in small businesses.

Altman's view that AI's slower-than-expected transformation is partly due to old habits – and the industry's failure to explain the technology's benefits – is debatable. A recent survey showed most people don't trust CEOs to act responsibly on AI, so would they even believe the spiel?

Altman's been offering some pretty controversial views recently. He said in July that AI won't shorten the workweek because we humans secretly enjoy being busy. More recently, he said four years is too long for college courses.

Permalink to story:

 
The sociopaths are making ill-thought-out tone-deaf pronouncements from their bunkers again.
The figure for the amount of people using chat bots is really misleading. When you ask people if they would pay to use a chat bot the number dwindles to near-zero.

Also as was widely predicted the models have plateaued massively in the last year and are still as unreliable as ever, making very dogmatic pronouncements which are just plain wrong. As an example LTT recently retried to use Chat GPT to help build a PC - here the video of how badly it went. Given what an on-rails process a PC build is it compared to the vagaries of modern life in general, it gives you some idea of how the devil-in-the-detail failings of LLMs are still all there.
 
Last edited:
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).
 
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).
Again, the devil is in the details. People are only using AI because it is heavily subsidized and extremely expensive. if it wasn't free for most people then they wouldn't be using it. There is incredibly efficient hardware around the corner, but we have thousands of datacenters already filled to the brim with current gen hardware with massive loans attached that will take several years to pay off. They won't be able to upgrade to the latest hardware without trillions more in investment. Public AI debt is 1.5trillion but there are estimates that there is between 5 and 7 trillion in shadow debt created by the AI industry.

So for now what you said is true, but people are either going to have to start paying the real cost of AI, there is so magic round of funding to pay off the old hardware AND upgrade to the latest chips or there is a crash where the hardware infrastructure is bought up for pennies on the dollar in bankruptcy and allows someone to offer AI for extremely cheap.
 
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).
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...)
 
Last edited:
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).
Is that including forced things like googles overview though
 
Also as was widely predicted the models have plateaued massively in the last year and are still as unreliable as ever, making very dogmatic pronouncements which are just plain wrong. As an example LTT recently retried to use Chat GPT to help build a PC - here the video of how badly it went. Given what an on-rails process a PC build is it compared to the vagaries of modern life in general, it gives you some idea of how the devil-in-the-detail failings of LLMs are still all there.
I watched the video and he did not have ChatGPT help build a PC. He had it build the entire PC and did no real thinking for himself that actual people would do. Normal people will use a brain and do things themselves, going to AI to automate tasks or for help for things they don't know. For most things, AI is a useful tool but not the entire solution.
 
Don't be so cruel to Sam people ! After all he has invested so much money in AI. He wants that money back and he is now acting like a crybaby, blaming others for his inadequacy to proliferate AI adoption. Since I live in Greece, I can reassure you that he recently "suddenly" appeared online, during a meeting regarding AI adoption in Greek Schools, to promote AI culture, while the Prime Minister and the Minister of Education were in awe and amazement for his participation. After that meeting the mandatory adoption of AI tools in Greek Schools of all grades was announced triumphantly. For Sam's sake, a very minor mission was accomplished, but it is nowhere near enough though...
 
Is that including forced things like googles overview though
Forced, seriously? What do you think Google's website should look like? The same as 20 years ago? Maybe sorting the results alphabetically instead of by relevance?

Nobody is forced to do anything, it's perfectly possible to use classic search and sift through dozens of irrelevant websites, in case you enjoy wasting time. Google's AI overview will accurately explain to you how to get rid of it.
 
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...)
What's the purpose of these childish insults, except to show you have no real arguments?
 
Forced, seriously? What do you think Google's website should look like? The same as 20 years ago? Maybe sorting the results alphabetically instead of by relevance?

Nobody is forced to do anything, it's perfectly possible to use classic search and sift through dozens of irrelevant websites, in case you enjoy wasting time. Google's AI overview will accurately explain to you how to get rid of it.
If you search anything in Google you get an AI overview regardless of if you want it or read it. If that counts as a user then it artificially pumps up the numbers.

Google had actively made the normal search worse to push people to the overview. You now get ads, the AI, then ads then maybe what you searched for. At least they don’t bury Wikipedia to the bottom of the page like they did when AI first started
 
Again, the devil is in the details. People are only using AI because it is heavily subsidized and extremely expensive. if it wasn't free for most people then they wouldn't be using it. There is incredibly efficient hardware around the corner, but we have thousands of datacenters already filled to the brim with current gen hardware with massive loans attached that will take several years to pay off. They won't be able to upgrade to the latest hardware without trillions more in investment. Public AI debt is 1.5trillion but there are estimates that there is between 5 and 7 trillion in shadow debt created by the AI industry.

So for now what you said is true, but people are either going to have to start paying the real cost of AI, there is so magic round of funding to pay off the old hardware AND upgrade to the latest chips or there is a crash where the hardware infrastructure is bought up for pennies on the dollar in bankruptcy and allows someone to offer AI for extremely cheap.
I think if AI wasn't free for most people they would still be using it and paying, because it's incredibly useful and actually saves you money. Rising prices will create a real push towards efficiency, of which nobody really thinks yet but the reserves are huge. Simply using a router can slash the cost by more than half.
I have no idea where the notion that only the newest hardware is useful comes from. Older hardware and not so powerful but very cheap models can do 80-90% of the tasks just fine, because the vast majority of tasks do not require state of the art.
 
I have yet to trust ai to provide me with technical solutions to car problems. Every result is almost correct but that's not good enough for performing a proper repair or diagnosis.
 
I think if AI wasn't free for most people they would still be using it and paying, because it's incredibly useful and actually saves you money.

Then why is noone monitizing it? A
nthropic and OpenAI have both state that their subscription plans actually lose money.
I have no idea where the notion that only the newest hardware is useful comes from. Older hardware and not so powerful but very cheap models can do 80-90% of the tasks just fine, because the vast majority of tasks do not require state of the art.
AIs biggest problem isn't a lack of hardware, we actually have an over abundance of hardware and datacenters have setup markerts where they rent unused capacity. The specific problem that I'm talking about is the electricity and water costs associated with generating tokens. nVidia is saying that Vera Rubin has a 30X improvement in token generation perwatt of electricity used. The problem is that these things are being financed 5-7 years out and nVidia is still Shipping blackwell AI hardware into 2027 with datacenters currently being built with Blackwell hardware. Noone will be able to buy the new hardware and they won't even be able to make a case for the new hardware as the current gen still isn't profitable. It actually likely never will be profitable because we over built so much infrastructure using debt that people are selling compute at a lose in order to maintain loans.

Just a final note, inferencing and LLMs is not AI. These businesses highjacked the term AI for marketing purposes to raise capital and everyone fell for it.
 
Last edited:
I watched the video and he did not have ChatGPT help build a PC. He had it build the entire PC and did no real thinking for himself that actual people would do. Normal people will use a brain and do things themselves, going to AI to automate tasks or for help for things they don't know. For most things, AI is a useful tool but not the entire solution.

I don't know. I've been seeing plenty of lazy people, especially younger zoomers, completely turn off their brains and blindly follow ChatGPT advice for PC building.

Playing devil's advocate, it gives decent advice more often than not, even if not always the best/most optimal. One single hallucination could have disastrous consequences or lots of money wasted on incompatible/suboptimal parts though.
 
I don't know. I've been seeing plenty of lazy people, especially younger zoomers, completely turn off their brains and blindly follow ChatGPT advice for PC building.

Playing devil's advocate, it gives decent advice more often than not, even if not always the best/most optimal. One single hallucination could have disastrous consequences or lots of money wasted on incompatible/suboptimal parts though.
Like putting glue on pizza to help the toppings stay on.

Issue is it also gets a lot of stuff from places like Reddit which are often just wrong.
 
Then why is noone monitizing it? A
nthropic and OpenAI have both state that their subscription plans actually lose money.

AIs biggest problem isn't a lack of hardware, we actually have an over abundance of hardware and datacenters have setup markerts where they rent unused capacity. The specific problem that I'm talking about is the electricity and water costs associated with generating tokens. nVidia is saying that Vera Rubin has a 30X improvement in token generation perwatt of electricity used. The problem is that these things are being financed 5-7 years out and nVidia is still Shipping blackwell AI hardware into 2027 with datacenters currently being built with Blackwell hardware. Noone will be able to buy the new hardware and they won't even be able to make a case for the new hardware as the current gen still isn't profitable. It actually likely never will be profitable because we over built so much infrastructure using debt that people are selling compute at a lose in order to maintain loans.

Just a final note, inferencing and LLMs is not AI. These businesses highjacked the term AI for marketing purposes to raise capital and everyone fell for it.
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?
 
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?
I don’t think you’re reading it correctly. The majority of AI use currently is on free licences which operate at a massive loss, the paid tiers also operate at a massive loss. Using a lot of tokens just means the company providing the service is losing more money as the tokens cost more to generate than they cost to make.

If your inefficient GPUs aren’t profitable now they’re not going to be when competitors can offer a lower price and maintain the same margin.
 
Last edited:
I keep browsing my slop social media platforms like TikTok and Reels and they keep serving me AI slop.

Altman, stop the development of AI right now or I'm calling Bernie. I heard on TikTok he's going to sort all of this out. I have been swiping past all the AI slop posts and now I see a ton of posters that share my feelings, we are many.

I'm looking for likeminded, highly sophisticated individuals on TechSpot. Together, we can take down the AI. Though we might not even need to, this 5 year bubble is going to burst any time now.


 
Back