It's this part.
Salaries and benefits are expensive. A computer program doesn't need a salary, retirement benefits, insurance, retirement, doesn't call in sick, doesn't take vacations, works 24/7, etc.
Throwing "ai" into it is a simple addition, if it works, great, if it doesn't well the market just wasn't ready.
But if they have to actually talk to their users and solve their real problems that's a really hard pill to swallow, extremely hard to solve correctly, and basically impossible to sell to shareholders because you likely have to explain that your last 50 ideas and the tech debt they created are the problem that needs to be excised.
Tomorrow it will be Agentic AI Blockchains.
I know what you are thinking: robots are coming for our jobs after that. Don't worry! Those AI robots will run on the Cloud.
But, I must try to have a little bit of self awareness here: if we all think it can do the jobs we don’t understand and don’t think it can do the job we’ve got experience in, then maybe that just indicates that it isn’t really very good at anything yet.
In this scenario, individuals without substantial capital could leverage AI to achieve outcomes that today require the resources and influence of wealthy founders. It might do the opposite of what CEOs seem to think: challenge existing power structures and create a more level playing field.
When someone uses an AI they do not own, they are (maybe) receiving a benefit in exchange for improving that AI and associated intellectual property / competitive advantage of the person or entity that owns the AI—-and subsequently improving the final position of the AI’s owner.
The better an AI becomes, the more valuable it becomes, and the more likely that the owner of the AI would want to either restrict access to the AI and extract additional value from users (e.g. via paid subscription model) or leverage the AI to develop new or improve existing revenue streams—-even if doing so is to the detriment of AI users. After all… a sufficiently-trained “AGI” AI could (in theory) be capable of outsmarting anyone that uses it, know more about its users than its users consciously know about themselves, and could act faster than any human.
While I share in your hope, I think it is unfortunately far more likely that AIs will widen the gap between the haves and the have-nots and will evolve into some of the most financially and intellectually oppressive technology ever used by humans (willingly or not).
Data centers will soon outstrip all other uses of electrical power, as for an AI calling in sick, no it needs full power 24/7. AI has no creativity, no initiative, no conscious, and absolutely zero ethics.
"In a middle-ground scenario, by 2027 new AI servers sold that year alone could use between 85 to 134 terawatt hours (Twh) annually."
Why yes, it does.
Even setting aside that most AI hype: Yes, automation is in fact quite sinister if you do not go out of your way to deal with the downsides. Putting people out of a job is bad, actually.
Yes. The industrial revolution was a great boon to humanity that drastically improved quality of living and wealth. It also created horrific torment nexuses like mechanical looms into which we sent small children to get maimed.
And we absolutely could've had the former without the latter; Child labour laws handily proved it was possible, and should have been implemented far sooner.
So, if it is true we’re on the cusp of an AI Revolution, AGI, the Singularity, or anything like that, then there’s precedent to worry. It could destroy our lives and livelihoods on a timescale of decades, even if the whole world really would be over all improved in a century or two.