[wpietri]

I definitely wonder if industry is any better here. At least in academia, the papers are public, so there's an opportunity for scrutiny. But I've heard tell of "AI" boondoggles in both large companies and small. E.g., the large corporate "AI" efforts burning millions without making any real improvements. And I wonder how many startups out there have standards that are in effect lower than academia, but instead of writing papers they are shipping products than harm people's lives when they go wrong.

[Aperocky]

As a previous Data 'Scientist' I can tell you that it absolutely is.

Garbage in garbage out is the norm, the models are long established, but they can't mine gold from dirt. But nobody except the engineers seem to understand this.

I'm so glad I left to become a regular software engineer. My code does not depend on a blackbox that is fed with crap, and can be reliably tested.

[jerzyt]

It is rampant in consulting business. At least in an internal project it's possible to pull the plug when the results are not promising. In a consulting engagement, when the results are garbage, there's no revenue and no potential for an upsell. In effect the pressure to "find" significant results is enormous.