It's too late for monopolizing large language models. The technology for making them is known and the cost keeps coming down.
There are things to worry about, but this doesn't seem to be one of them.
There are things to worry about, but this doesn't seem to be one of them.
1. GPU prices aren't as high as they were during the mining peak, but you can still buy a car with the money it takes to buy a pod of A100s. Further to that point, Google et al. are making the next generation of ML accelerators and they won't even sell the best ones.
2. Compute is only half the problem. Access to data (e.g. in the form of email and smart assistant queries and ring door bells) is a much under appreciated second half of that equation.
You may argue that the knowledge required to develop LLMs is open source and will be accessible to all (and in practice this is not-quite but mostly true). However the federal government is currently in the process of making it so only certain people are authorized to do so legally.
It may be very hard to train cutting edge models if you only have consumer hardware
Not so sure non-institutional folks can catch up.
OpenAI is trying to regulatory capture LLMs to prevent the Amazons and the well-funded unicorns from following. Everyone left in the lurch after these players crystalize is SOL.
You, average HN reader, cannot possibly compete with OpenAI. They're trying to stem venture dollars from standing up more well-funded competition so they take as much of the pie as possible.
The hype/funding cycle for LLMs is mostly over now. You're not Anthropic and you're more than likely too late to get in on that game.
It's not the hammers, it's the building permits.
Edit: drugs and licensed services like law and healthcare also apply