Technology

Yeah you are likely correct. it's not hard to guess. all this hype around ai is to temporarily increase share price.

Whale is a term that became popular with the rise of online games, specially on mobiles, that rely heavily on microtransactions for their income stream

The companies behind such games internally use the term whales for players/customers who spend ludicrously and disproportionately high amounts of money on the game. Often these players are addicted to the game or there are other psychological factors at play.

Most importantly, income from whales accounts for a large portion of the total income. There is this talk by a mobile game dev that went viral because he very casually talks about the exploitative nature of his games with no self awareness: https://youtube.com/watch?v=xNjI03CGkb4

Since then the term is used in other businesses where applicable. Here a small number of customers are accounting for about half the revenue so they are being called whales.

I think it's less about models getting smaller more about the current approach to AI (transformer based LLMs) hitting a ceiling that will cause problems for them.

Meta trained Llama 3.1 405b model on 16 thousand H100s: https://ai.meta.com/blog/meta-llama-3-1/

So yeah, the scale of it can get wild. But from what I can tell, it seems like there's a clear diminishing returns on usefulness of throwing more processing power at model training and more breakthroughs are needed in the architecture to get much meaningfully further on general model "competence." The main problem seems to be that you need a ridiculous amount of decent data to make it worth scaling up. Not only in terms of the model showing signs of actually being "better", but in terms of the cost to run inference on it when a given user actually uses the model. And quantization can somewhat reduce the cost to run it, but in exchange for reducing overall model competence.

Right now, my general impression is that the heavy hitter companies are still trying to figure out where the boundaries of the transformer architecture are. But I'm skeptical they can push it much further than it has gone through brute forcing scale. I think LLMs are going to need a breakthrough along the lines of "learning more from less" to make substantial strides beyond where they're at now.