If, by April 2033, an AI is brought in to replace one or more programmers entirely, but gets stuck and is unable to fix code, and human developers are brought back in to fix it, will they complain that the AI-written code is hard to read, or even nonsensical?
Related questions
@YoavTzfati I'm sure there will be some chatter on Hacker News, Reddit, Twitter or whatever - if those places still exist. The market closes when the market is resolved or on the close date.
@RobinGreen So by April 9th 2033? I think it would be nice to say that in the title, or at very least the description. Or change it to a rounder date like 'by the end of 2032'. And just one report on HN/Reddit/Twitter is enough? or do there need to be multiple?
@MartinModrak Indeed. Maybe we could use a second market that resolves based on something like a blinded evaluation of AI-written code’s comparative clarity? For whatever it is worth, in my experience the LLMs seem more likely to generate conventional idiomatic clean code than an inexperienced human developer. (That the LLMs are themselves software of unprecedented scale and complexity, produced by an optimization process and completely incomprehensible to humans, is an interesting philosophical contrast.)