Posted by todsacerdoti 22 hours ago
Can we generate a huge amount of code, just compilable code, which is essentially just a trash. We seed the github, bitbucket, etc. and pollute the training grounds.
How many times have you had a coworker/boss/user ask for something totally nonsensical, impossible, or maybe even illegal? These people when they get unleashed on these agents and on large business problems are going to be a menace. They will create walls and heaps and mountains of working but totally stupid code as the AI attempts to work with their malformed attempts at a 'thought' and this will pollute github and the codebase for training...eventually it will outnumber serious, professionally written projects and you will reach an inflection point at which AI coding agents 'peak' and they begin to deteriorate.
Either that or you need to somehow ensure you are not training on AI written code but that too will cause a PR problem for the firms as it becomes obvious that AI cannot in fact replace developers?
What will/can HN do about it?
If that's worth the cost... probably not?
For now maybe all forums should require some bloody swearing in each comment to at least prove you've got some damn human borne annoyance in you? It might even work against the big players for a little bit, because they have an incentive to have their LLMs not swearing. The monetary reward is after all in sounding professional.
Easy enough for any groups to overcome of course, but at least it'd be amusing for a while. Just watching the swear-farms getting set up in lower paid countries, mistakes being made by the large companies when using the "swearing enabled" models and all that.
Maybe the em dash is the self censorship/deletion mechanism that we've all been waiting for. Better than having to write pill subscription ads, I suppose.
Sometimes I'd like to correct misinformation about something niche that I happen to be over-knowledgable about, or drop a 1-5 line code fix for the very thing they're talking about.
But if the cost of me getting that commenting access is also lowering the threshold needed for anyone (or anything) to comment, I'd rather keep the threshold high.
A good chunk of users on lobste.rs are bloggers, so if they get something not quite right, I can just contact them through their blog anyway.
Perhaps there needs to be some sort of voluntary ethical disclosure practice to disclaim text as AI-generated with some sort of unusual signifiers. „Lower double quotes perhaps?„