Ask HN: AI projects that have hit the trifecta of Benefit, Cost and Time

[KuriousCat]

Looking at many of the ongoing AI projects, I wonder if they are consistently hitting the trifecta of Cost, Time and Benefits. Even something like chatGPT is a loss maker.

https://finance.yahoo.com/news/chatgpt-cost-bomb-openais-losses-125101043.html

Are there any strong examples I can cite where AI has delivered/over delivered? Or majority of them vaporware?

[al_borland]

Most of what I see that seems like good AI/ML is stuff being added to existing products and make it better, not stand alone LLMs or startups.

For example, all the ML involved in the image processing on modern smartphone cameras. Or ML models used to recognize and extract text from images or video. I find these things extremely useful, and with them happening on-device, there isn’t a giant hungry datacenter required to use it.

The other day I was sent a screenshot of a table in a website by my boss and needed to add information for each row. I was table to highlight it all out of the image and paste it into Excel and it required only very minor cleanup thanks to the ML. If this was 5 years ago I probably would have had to re-type it all by hand, or just hand back a much worse product.

Sure, this is just better OCR, but I find these types of things have a bunch bigger positive impact on my day-to-day life than whatever the newest AI startup is pitching.

[usgroup]

Google/Bing/Yahoo/Yandex/etc search and Meta/Google/AMZ/Linkedin engagement algos seem like fairly clear front runners for world changing ML projects, but very many others come to mind too.

[mikewarot]

Speech recognition seems to actually be useful these days, it's been a long time coming.

[sircastor]

I think most AI projects are thriving off of investor money and media hype. I don’t genuinely believe any of them are making any kind of profit.

[solumunus → sircastor]

I don't think anyone is suggesting they are.

[jyu]

there's a lot of hype, but you already see it helpful at the fringes. alphazero, alphafold, and recently alphageometry is almost gold medalist level already.

it's turning learning and search into a compute bound problem accessibly by SWEs and DSs.

[muzani]

By definition, every company raising VC money is losing money. Maybe look at AI projects that aren't VC backed? Lots of open source stuff is also indirectly VC funded too, if you trace the money.

It's not like they'll lose money forever. Some end up like telcos, just a part of the background. Some end up like FAANG, utilizing an advantage until they become rich and die from complacency. Some end up enshittifying like Twitter/X and Reddit; they hit the brakes hard to cut costs but users still don't leave.