Register Blindness: What a Failed Prompt Reveals About AI

Despite recent bad press and an increasingly shaky economic outlook for AI, I’ve remained focused on what seems to me beyond doubt or serious criticism: the idea that working with an AI chatbot can be productive, profitable and enriching if done in the right way. Most reasonable people agree with that. Yet the central question remains surprisingly open: What, exactly, is the right way? There’s been no lack of commentators reminding us that, whether for work or leisure, we tend to use AI the wrong way. And there happen to be several wrong ways. Perhaps the most common and laziest missteps is to ask direct factual questions and fall victim to hallucinations. Another is our inclination to anthropomorphize the interface, seeking emotional solace for loneliness. In terms of professional use, where the aim is increased productivity, blindly delegating tedious tasks risks creating cognitive dependency on an unaccountable subordinate. Other examples of dysfunction exist.

This excerpt comes from the named source. Read the original for the full reporting.