Article is behind paywall and is simply saying the same things that people have been saying about the post tech crash.
Now, what this sort of article tends to miss (and I will never know because it’s paywalled like a jackass) is that these models services are used by everyday people for every day tasks. Doesn’t matter if they’re good or not. It enables them to do less work for the same pay. Don’t focus on the money the models are bringing in today, focus on the dependency they’re building on people’s minds.
The issue though being that most people aren't paying and even those paying if they use it moderately aren't profitable. Nvidia "investing" 100B in one of its largest customers is a cataclysmically bright red flag.
A handful of the largest companies cyclically investing and buying from each other is propping up the entire economy. Also stuff like Deepseek and other open source models exist. Unless AGI comes from LLMs (it absolutely won't) then its foolish to think there wont be a bubble
I was thinking it's a bit like developing powered flight and saying steam engines won't work. It's true they didn't but the internal combustion engine was developed which did. It was still an engine machined from metal but with a different design. I think LLM -> AGI will go like that - some design evolved from LLMs but different in important ways.
AGI might require a nobel price level invention, I am not even sure it will come in my lifetime and I am in my 30s.. Although I would hope we would get something that could solve difficult diseases that have more or less no treatment or cure today, at least Demis Hassabis seems interested in that.
Now, what this sort of article tends to miss (and I will never know because it’s paywalled like a jackass) is that these models services are used by everyday people for every day tasks. Doesn’t matter if they’re good or not. It enables them to do less work for the same pay. Don’t focus on the money the models are bringing in today, focus on the dependency they’re building on people’s minds.