Recently, Jacob Coxon posted on X that he believes frontier AI models will likely kill us all by the end of the decade. And as always - the media and the internet has picked it up and ran with apoloptic Skynet like scenarios.
Honestly, I’m getting pretty tired of this kind of AI doom and gloom.
The idea that we are somehow hurtling toward an inevitable superintelligence that will wipe out humanity makes for a great headline, but I don’t see much evidence for it. What we actually have today are large language models — probabilistic token-prediction systems - that can produce remarkably impressive results while also making spectacularly stupid mistakes.
That’s it.
They’re software. And, like all software, they’re buggy.
They hallucinate. They lose context. They confidently make things up. They struggle with basic reasoning tasks. They can produce brilliant answers one minute and complete nonsense the next. Calling these systems “superintelligence” or “artificial general intelligence” doesn’t magically turn them into something they aren’t.
What bothers me most is how the conversation has shifted from “here’s what these systems can actually do” to increasingly elaborate predictions about what they might become.
And there are plenty of financial incentives behind that story. If you’ve spent the last three years selling investors on the idea that AI is going to fundamentally transform everything, it’s probably not a great time to start admitting that the technology has some pretty significant limitations. “We’re building powerful but imperfect software” doesn’t quite have the same Wall Street appeal as “we’re creating superintelligence.”
So now we’re supposed to believe that the same companies that have been hyping AI as the greatest technological revolution in history are also racing toward machines that will destroy civilization.
These stories are designed to keep the money flowing.
Maybe I’m wrong. Maybe something radically different will emerge. But I have lived through multiple iterations of the Silicon Valley tech bro’s so called revolutions. The Metaverse, VR, Crytopcorrency and NFTs are just some of the things we have gone through in the last decade - the latest of which is AGI. I’m not convinced we’re watching the birth of some digital god.
We’re watching the development of some very impressive—and and very buggy software. Skynet - it is not.