Maybe people simply don't want math PhDs but something else? Since 1-2 years ago I started doing consulting/freelancing in the ML space, but more on the infrastructure, deployments and similar stuff, as a general purpose developer, and I have a waiting list of clients interested in more work, some of them even trying to recruit me to work for them full-time as well. I'm based in continental Europe, fwiw.
Also bunch of past workplaces who've adopted AI in various ways who reach out once they find out what my current focus lies, but that's harder for others to replicate unless you've already had a career as a developer.
I don't think the "incredibly small minority of companies in the world do any real training or optimisation" part is necessarily as true, as some parts of the work I do get is about helping them optimize training and infrastructure around training. Mind you, none of this is for building LLMs from scratch, it's 99% fine-tuning existing checkpoints.
I'd also agree with "paulg is in somewhat of a bubble" regardless of this, which is worth remembering whenever you read his content. Same goes for any person living in SF, and dare I say the US. But also, YMMV, I live and work in Europe, probably why I have this perspective.
Worst of both worlds - no casual gossip feed in the Bay Area, no big dog meetings in the UK. (Which mostly has no idea he exists.)
As for the question - what are the odds LLMs will be anywhere near the top of the tech tree five years from now?
The trend seems pretty clear to me - local/offshore models are snapping at the heels of the big names in the US, and the current investment arc is insane.
I wouldn't bet on Anthropic or OpenAI being leaders five years from now. Longer term, I especially wouldn't bet on the US build-yourself-a-monopoly corporate model surviving AI at all.
Is it possible to see some of your old works? Personal research?
Citation needed. I'm guessing you're not counting children.