More evidence AI is a bubble: Companies have stopped asking themselves whether new features and products empower their customers to achieve anything meaningful.
Instead, features and products are viewed through a single lens: “How well does this feature or product act as a conduit for marketing generative AI, lifting our prospects through hope?”
Exhibit 1: The attached image doesn’t actually show use cases: These are hand-wavey overgeneralized business functions. It’s not falsifiable.
Recognizing that this is a bubble is useful, because we can infer things about it from previous bubbles. Like… Companies will rush to get into the business of renting AI compute and tooling out to other companies, so they can collect rents.
If you are essentially reselling AI, and have little differentiation, you are—as we say about opening a store or restaurant without a sustainable plan—working for the landlord, not yourself.
Almost zero of these use cases are handled more efficiently, much less correctly, than by a human agent, and typically at much greater expense on every level of the whole shebang.
The one which might be more efficiently done by an algorithm is pricing and promotions, and that's just because that particular set of tasks is adjacent to sociopathic inclinations which reduce customers and clients to nothing more than numbers on a screen.
My favourite here is 'knowledge creation'. An LLM doesn't create knowledge. It eats what you feed it and then regurgitates a Markov chain which is engineered to look correct, without any consideration for content or meaning. And that to my mind illustrates, most clearly, why this bubble is going to pop, and soon: they're feeding the GDP of small nations into something which demonstrably puts out something much smaller -- and much worse.
@raganwald Notice which “use case” isn’t listed? Executive decision making. When a company finally outsources their C suite to AI, then and only then will I believe their AI hype.
Add comment