[Kelvin506]

For me it's that it's all so obviously a grift. With crypto there was at least some legitimate value to blockchains. ML is a valid science and has led to major breakthroughs in other sciences.

But AGI/genAI LLMs come across like just another scam from the Worst Guys You Know who were all cryptobros in the 2010s. The output integrity problems get dismissed. The traceability/repeatability problems get dismissed. The data ownership problems get dismissed. There's the endless problems with aggressive feeder bots causing real (financial/operational) harm to others. All of the people selling it have the smell of timeshare salesmen.

And now with China releasing open source models that seemingly embarass their U.S. counterparts, it makes the whole thing feel even more like a ponzi scheme.

LLMs as a whole have a severe credibility problem.

[jcgrillo]

It's infecting everything. The other day I overheard someone saying "we don't need to expose a query interface, we'll just give users a text box and let them use natural language". Total brain rot.

[sensanaty]

My PM has been talking about converting half our product into a "conversational UI", meaning what used to be buttons would be turned into a textbox that sends a request to an AI API that will in turn trigger the same action as before. Of course, it shits itself 80% of the time, but AI!!!11!

It's total lunacy, I feel like I'm in a mental asylum sometimes.