Navin posts a delightfully dystopian article about a job candidate who got interviewed… by an AI.
- I remain unclear about whether it was a human being asking questions as decided by the bot, or whether it was a bot throughout. As you can above, Paddyumn first says that “the panel was the same person who sent me this mail”, but it was ACTUALLY an AI bot. And later on, he goes on to say that “except for the interviewer’s voice, every part of the conversation felt pretty real”. So was a human there on the other side at all, or not?
Second, Praddyumn, are you really sure that a person sent you that first email? - Hey, if interviewers can be AI bots…
- No, but seriously, think about it. If, at the end of the interview, the interviewee can show that it was a bot all along, what other proof do you want about the awesome prompting skills that the interviewee possesses?
- Navin points out that this is unlikely to work well. He has run his fair share of experiments on ChatGPT4, and Navin wouldn’t really want AI to take decisions on behalf of humans.
- Well, yes, but how about building out a two stage filtration process? Conduct, say, a hundred such interviews, and select the bottom and top 10% for human verification. This is, of course, a variant of an excellent idea proposed by Navin himself…
- We have essay graders already, by the way.
- What becomes scarce in this framework? What becomes more valuable? What becomes less valuable? How does one think about the answers to these questions, and optimize for the labor market accordingly?
- How many of these eight questions do you think I came up with? <wink, wink>
(All eight were by me, I assure you. Unless specified otherwise, every single word on EFE is by me. I take the help of ChatGPT to edit and improve longer non-EFE pieces that I write, but the blog is mine and mine alone. So far.)
EconForEverybody 