there are two opposite-but-equal problematic approaches journalists keep using to cover AI and automation technology:
1) they give the technology some insanely difficult and subjective task like "translate this poem" or "psychoanalyze me" and then when it fails they say "lol, well the technology isn't there yet." in actual reality, these products are mostly going to be doing repetitive grunt work, and given the right parameters, they can be **very** good at that.
2) they look at the output in a field or area where they have zero experience and say "wow! that's impressive!" when they have no way of knowing whether it's impressive because they don't work in that field. i can put Python prompts into ChatGPT and get what look to me like incredible results, but it's silly to people who do code for a living. or i just heard a snippet of a podcast that Spotify auto-translated into Spanish. the journalists were like "sounds great!" folks, it was nonsense.
Oddbean new post about login | logout