Sparkling LLMs

When the internet first arrived, it was painted as something it never was.

It was this mystical cyberspace -- a separate world you connected to, like in Johnny Mnemonic (1995). You didn’t just browse -- you surfed the web, like Crash Override in Hackers (1995).

Fast forward to today. The internet isn’t a place you go anymore -- it’s everywhere. It surrounds us: on every screen, in every pocket, woven into daily life. It enables amazing things and, like everything in life, some terrible ones too. But the internet is magnificent. It’s a miracle, really. But it’s nowhere close to the cyberspace we were promised, or the metaverse Mark Zuckerberg had in mind just a few years ago.

I think the same is happening with AI now.

It’s even in the name. It’s being presented as artificial intelligence -- something autonomous, conscious, maybe even self-aware. A mind, humming away somewhere in a data center. But that’s not what it is.

It’s statistics. It’s math. It’s prediction. It’s token probability at scale. Yes, it can write, draw, answer, correct your grammar, code. Sometimes it even feels human. But it doesn’t understand.

I don’t know what we’ll end up with many years from now. It’ll probably be useful -- like the internet is -- and in many ways, it already is. But I don’t think this current wave of LLM technology is going to get us to true artificial intelligence. Not HAL 9000 from 2001: A Space Odyssey (1968), or the Entity from Mission: Impossible – The Final Reckoning(2025). So please stop pretending it is.

Because what you're building now isn’t the champagne of AI, and building a Dyson Sphere with investor money won’t change that. What you're building is sparkling LLM.

And some of it is just flat.