Is generative AI the best use of the world's electricity?
In a clip that starts a recent episode of the podcast Tech Won't Save Us, Sam Altman acknowledges generative AI's Achilles' heel: energy. Generative AI needs a ton of energy, at a time when tech companies are supposedly pushing to curb carbon emissions to address climate concerns. Seeing tech companies go all-in on generative AI when climate change very much remains a problem may seem like a disconnect, a misalignment of ideals and reality. But this disconnect isn't a problem to Altman, because he sees the contradiction as a feature, not a bug. According to Altman, the energy needs of generative AI are just the thing that will push us to find a much-needed breakthrough in energy generation.
Because, you know, we weren't properly motivated before generative AI hit the scene. Maybe tech will save us after all!
Altman's assertion highlights a problem I have with the prophets of AI: they're quick to tell us what they need for generative AI to succeed, but they can't explain why we, the masses, need generative AI to succeed. The closest they come is when they tell us how artificial general intelligence (AGI) will benefit us.
But I have a couple of issues with such claims:
- Despite what the hypemen tell us, we're nowhere near reaching AGI.
- If AGI replaces us, then we're screwed, so I don't want to help out if I have a choice.
Let's focus on objection #1. We're well beyond the point of selling generative AI on its potential. The technology is stalling, with no signs of significant gains anywhere on the horizon. So, when we promote generative AI, we must do so in terms of what it can do now, not what it can do in a future that may never come to be.
Let's say we somehow find a way to cleanly increase our energy generation to a level that meets the needs of generative AI. This hypothetical new power source is plentiful and places no extra strain on our power infrastructure. Why the hell should generative AI be first in line for this new source of energy? Surely there must be better ways we could use the resulting energy.
How many areas of the world still don't know firsthand the wonders of electricity? There must be other places where the extra energy could be used to produce something of value. Can't that energy go toward enabling manufacturing or keeping critical infrastructure online during bad weather? Generative AI has value only for those who sell it. The technology offers little value to those paying to use it.
The prophets of AI act as if everyday people need generative AI to be successful, but only those invested in generative AI have such need. For the rest of us, generative AI is merely something which exists. Sure, it can make for a fun digital toy or two—but that's the thing: when you account for generative AI's hallucinations and its lack of reasoning skills, you realize its greatest use is as something you use to kill time. Okay, that's not completely fair—generative AI is plenty useful if you want to spam and scam at scale. Scale is, after all, what generative AI can do better than humans. The only problem is that generative AI doesn't scale anything worth a damn.
You can't blame Altman for looking out for his investment. But we don't have to validate his quest. Altman is full of shit. He has every right to be full of shit. And we have every right to ignore him when he speaks, something I hope more people in power will start doing soon.
Altman and his ilk are out of touch, and it's obvious when they talk about how generative AI and AGI will make our lives better. The way they tell it, we'll all have more time to sit around and ponder life's biggest questions. Some, like Altman, will use this 'inevitability' to push for universal basic income (UBI). But technology alone won't bring about the adoption of UBI, because UBI isn't a technological problem; it's a political problem.
Even the need for more energy for generative AI isn't a purely technological problem. On one hand, it is, because we lack the technology to cleanly generate the needed extra energy. But, on the other hand, even if we were to solve the technological part, we'd soon run into the problem caused by a lack of investment in critical infrastructure. As Molly White has pointed out, tech alone can't solve all our problems. Unfortunately, tech may in fact be the very thing keeping us from finding the solutions we desperately need.
