[bgwalter]

It's for the press and the general public who don't really understand the tech. The hype can usually be sustained for around two years.

The press nearly always obliges in the first year. It works for wars and pandemics, too. In the second year the first dissenters begin to show up. We are now in past the stage of opinion reversal, where the press and the public mostly hate "AI".

This is also why it is wrong to compare "AI" to the early Internet. After the Internet bubble burst, the public still liked the Internet.

[alephnerd]

> This is also why it is wrong to compare "AI" to the early Internet. After the Internet bubble burst, the public still liked the Internet

The comparison is from a business perspective, and I absolutely stand by the comparison with the early generation of Internet companies with the early generation of AI companies.

Anyhow, the general public doesn't matter. It's capital (private and public) along with business cycles that create markets, and most of the public doesn't have the knowledge or the capital to make a difference.

[piva00]

It's capital that can force markets to exist but without the public to consume it's just wasted capital if value doesn't materialise.

The crux of the whole AI boom is exactly how much value it can materialise given the capital expenditure being absurdly astronomical, if it doesn't become the next trillion+ US$ market it will be a huge misallocation of capital.

[alephnerd]

The AI/ML products that are FCF positive and seeing significant traction are those that are complementing workflows in some shape or form by reducing

AI is now a loose used term that is encompassing 3 loosely connected markets that have now fallen under the same umbrella:

1. Construction/Land Speculation: a large portion of the AI story you hear about is a DC construction story

2. Hardware: a large portion of the AI story is just a rebranding around GPU fab and design, especially due to issues around subsidy disbursement under the CHIPS act

3. SaaS/Applications: a lot of products being derisively called as "LLM wrappers" are not cool from a technical perspective, but from an RoI perspective are good enough - $30k for data entry automation that is 80% right is cheaper than hiring a data entry team of 4 who cost $60k each.

Much of the bubble is due to 1, but 2 and 3 are somewhat insulated because of FCF and adjacent markets and narratives to pivot to (eg. For 2 it was "Chip Wars" 2-3 years ago and before that it was "ML" for 1-2 years and before that it was "Precision Medicine"...)

> without the public to consume...

The AI story really isn't a B2C story no matter how much people try to shoehorn it. The Cs are perpetually broke and margins are shit. The value that arises from automation like AI is around workflow and workforce augmentation in some shape or form, which makes it a B2B play.