The Great AI Power Grab: Digital Dreams and Electric Nightmares

Nov. 29, 2024

Listen up, you digital dreamers and code cowboys. While you’ve been busy asking ChatGPT to write your love letters, something’s been cooking in those massive server farms - and I’m not talking about the midnight pizza runs for exhausted programmers.

I’m nursing my third bourbon of the morning, staring at these Goldman Sachs numbers, and they’re making my hangover seem pleasant by comparison. These fancy AI systems we’re all jerking off about? They’re about to jack up data center power demand by 160% by 2030. That’s not a typo, though I wish it was - my trembling hands don’t make that many mistakes.

The real gut punch? Training just one of these AI models burns through as much juice as 100 American households use in a year. Think about that while you’re asking an AI to generate pictures of cats wearing spacesuits. That’s enough power to run my entire apartment building, including crazy Dave’s illegal Bitcoin mining operation in 3B, for a full year.

And here’s where it gets really interesting, like finding out your ex is dating your therapist. Google - you know, the “don’t be evil” folks - has somehow managed to bump up their CO2 emissions by 48% in five years. That’s the kind of growth rate my liver doctor warns me about.

The suits are scrambling like cockroaches when the lights come on. Their solution? Carbon credits - the corporate equivalent of saying “sorry” after puking on someone’s shoes. It’s a band-aid on a bullet wound, folks, and about as effective as my attempts at moderation.

But wait, there’s more! These tech wizards are now pushing something called “digital twins” - virtual copies of real processes that can supposedly cut building emissions by 50%. Sounds great until you see the price tag: up to $1 million per twin. That’s a lot of zeros for what essentially amounts to playing The Sims with your office building.

Some companies are sobering up, though, pushing for “small language models” instead of the power-hungry behemoths we’ve got now. It’s like switching from Wild Turkey to near-beer - less impressive at parties, but you might actually remember what happened the next day.

The kicker is that investment in AI is expected to hit $200 billion by 2025. That’s enough money to buy everyone on Earth a decent bottle of bourbon, but instead, we’re dumping it into machines that need their own power plants.

Here’s the real punchline: while all this is happening, the bigwigs in charge of regulation are focused on making sure AI plays nice with ethics and privacy. That’s like worrying about your cholesterol while your house is on fire. The EU AI act and Biden’s executive order barely mention the environmental impact, probably because they were too busy making sure AI doesn’t hurt anyone’s feelings.

Microsoft is trying to play the good guy, making noise about sustainable AI use. That’s rich coming from a company whose operating system updates still manage to brick my laptop every other Tuesday.

The truth? We’re caught in a digital drag race, pedal to the metal, heading straight for an environmental brick wall. But hey, at least we’ll have AI-generated memes to keep us entertained while Rome burns.

Look, I’m not saying we should unplug everything and go back to counting on our fingers. But maybe - just maybe - we need to slow down and think about whether we need all these digital toys before we cook ourselves alive trying to make them smarter than us.

Time to wrap this up. My bottle’s running low, and these server farm emissions aren’t helping my smoker’s cough. Remember folks, the future’s coming whether we like it or not, but it might be nice if we could still breathe when it gets here.

Stay human, Henry Chinaski

P.S. If anyone needs me, I’ll be at O’Malley’s, teaching the jukebox AI to play better drinking songs.


Source: How Can Big Tech Companies Get To Grips With AI’s Energy Footprint?

Tags: ai ethics technologicalsingularity futureofwork aisafety