> Fundamentally, generative AIs are… generative. They make content. But the value add of most jobs is not the content itself; it’s the judgment. It’s knowing the right questions to ask, the right people to talk to, the right sources to trust, etc. Until an AI can take meaningful initiative, it will be at most a tool in the hands of a more competent human worker.

This is the ONLY part about "AI" that worries me at all.

AI is a computational tool. Like all such tools, the value is automating what was previously receptive, mindless tasks. Some people will use GPT to craft emails or document outlines for them and then will "massage" the output for correctness. That's a great application.

But the second that people start asking AI to make VALUE JUDGMENTS for them then we have a problem.

Even pre-GPT, we saw news articles about certain government agencies and businesses using AI to make policy or hiring decisions easier. Conversations have arisen about introduced biases into the models etc.

Where I work we are exploring using GPT to produce summaries of documents and messages. I can see how this will introduce conflict when the CEO insists that something was communicated but the LLM didn't consider that critical piece of information to be "relevant" or "important." Deciding relevance or importance is making a value judgement.