It's because other industries are buying expensive tokens, and there's no real ROI yet.

I have a suspicion that, outside of tech, AI WILL speed up the mundane aspects of people’s jobs, that part which they spend 80% of their time getting through, so they can then do the 20% where they actually add value.

BUT the new time they unlock will be spent doing more of the mundane work that previously just didn’t get done because it was overtaken by events or it got done to the absolute barest minimum level.

As a result the 20% where they add value never really grows or changes and in fact might be less valuable because of all the extra noise created by doing more tasks in the 80%.

This is working theory and likely full of holes.

The economy has many different resources with different supplies available. Making the supply of one resource infinite may sound like a huge win, but nothing is dependent on a single resource, everything requires a combination – just like in board game economies, you can't win by only producing wood.

AI represents some weird intelligence-like resource. It clearly does something of value, but that must be combined with other resources. In tech there's a lot of that, so making it (roughly) infinitely accessible has had a nice impact and we've run into our next limiting factor: human time. Companies that aren't "pure tech" though will run into much earlier limiting factors – hardware, compliance, regulation, artistic value, human time, and so on.

The cost of training/developing a frontier model will continue to rise. The cost of inference will continue to fall. Companies don't buy models as such, they buy inference. As inference becomes a larger and larger business expense companies will optimize that spending, routing work to the "capable enough" least cost tokens.

There is a lot of human work that can be offset with tokens. On the other hand, the amount of work for the highest intelligence models will remain limited - corporate strategy, major decisions, etc. are just not as frequent.

If mainstream corporate America thought it would turn increasingly cheap intelligence into higher revenue growth and higher profit margins, then mainstream corporate America needs to go back to school.

If you want to increase your profit margins, you'll have to innovate! Adopting AI tools that somebody else created, and are equally available to your competitors, won't increase your revenue or margins. The ones who will see increased revenue and profit margins are the ones who actually innovated, by creating the AI.

This is exactly what Apollo reports: the Magnificent Seven increased their profit margins, while mainstream corporate America didn't.

Previously: https://news.ycombinator.com/item?id=50224463

609 points by the-mitr 1 day ago | hide | past | favorite | 645 comments