These engineers are becoming the real life equivalent of this Office Space scene:

https://www.youtube.com/watch?v=hNuu9CpdjIo

"I HAVE LLM SKILLS! I'M GOOD AT DEALING WITH THE LLMS!"

> Yes, the AI may have produced the recommendation but a human decided to follow it, so that human must be held accountable

It is common and a mistake IMO to rely on the AI as the sole source for answers to follow-up questions. Better verification would have humans sign off on the veracity of fundamental assumptions. But where does this live? Can an AI model be trusted to rely on previous corrections? This seems impossible or possibly adversarial in a public cloud.

The problem is the credit tends to go to LLMs. So there’s an imbalance. LLM did all the work. The person using it made all the mistakes.