The whole LLM as a service business model has a fundamental flaw to it. The cost of operating the data centres is an order of magnitude higher than the profit.
But if models get efficient enough to bring the costs down, then they become efficient enough to run locally. So, either it’s too expensive to operate, or nobody will want to use it as a service because running your own gives you privacy and flexibility.
The fact that investors don't get this is frankly incredible.
Ah, but have you priced in the (what is likely technically impossible) possibility of reaching AGI with LLMs?
It's like a lottery ticket for the wealthy who don't mind setting a bit of their money and the economy and the planet on fire for the slimmest of chances to become obscenely wealthy, even by billionaire standards.
It mostly is. But there are some true believers out there who take AGI quickly followed by super intelligence as a near certainty.
The rationalist movement - the same asshats who spawned effective altruism and the likes of Sam Bankman-Fried - have strong pull in Silicon Valley and take the mytgology of AI super intelligence they created literally.
But you are spot on in thinking the monied interests regarding AGI is largely just a marketing con.
@FantasticalEconomics @yogthos if AGI is possible (and it really, absolutely isn't within the framework of LLMs/transformers), I still don't see the connection to making their money back.
1. Create single universal global super-intelligence
2. ???
3. Profit!
I think step 2 is: ask god-AI how to profit haha.
But you are both absolutely right. It's not a well thought-out strategy and AGI is almost certainly not coming out of LLMs. It's not the tech that can produce it and, at best, would be a distant stepping stone towards a very different looking tech.
Yogthos
Kyle Montanio
ahistorical immaterialist