But what will ChatGPT use to dredge up answers once there's no stack overflow threads to fuel its training data? This is the wrinkle I don't see getting addressed.
Read an study today that filtered out new studies based on whether they used GPTisms. They found that people who used LLMs to help write their studies had significantly faster career advancement, but that the AI based studies seemed to break significantly less scientific ground, and had narrower subject matter, because it really only helps on already tread ground.
Same for derivative videos, derivative forum answers.
The slop will be fed slop and only get sloppier, but it is difficult enough to recognize that there will still be personal gain from using it.
It is why certain billionaires are pushing human recognition devices, so they can have an advantage at training against things less likely to be slop and can corner things more by using it to disallow scraping. They will likely fail, and be fed slop anyway.
Right now AI is still getting better because of increases in ram, but eventually that will hit a limit in that a larger window only helps if you have useful things to put in it, and things will get worse.
Too bad no one makes manuals that are worth a damn or kept up to date anymore. Go ask the sysadmin crowd about Microsoft's ever changing interface for their cloud infrastructure.
But what will ChatGPT use to dredge up answers once there's no stack overflow threads to fuel its training data? This is the wrinkle I don't see getting addressed.
The slopmageddon is nigh.
Read an study today that filtered out new studies based on whether they used GPTisms. They found that people who used LLMs to help write their studies had significantly faster career advancement, but that the AI based studies seemed to break significantly less scientific ground, and had narrower subject matter, because it really only helps on already tread ground.
Same for derivative videos, derivative forum answers.
The slop will be fed slop and only get sloppier, but it is difficult enough to recognize that there will still be personal gain from using it.
It is why certain billionaires are pushing human recognition devices, so they can have an advantage at training against things less likely to be slop and can corner things more by using it to disallow scraping. They will likely fail, and be fed slop anyway.
Right now AI is still getting better because of increases in ram, but eventually that will hit a limit in that a larger window only helps if you have useful things to put in it, and things will get worse.
RTFM. It will be back to the old days.
Too bad no one makes manuals that are worth a damn or kept up to date anymore. Go ask the sysadmin crowd about Microsoft's ever changing interface for their cloud infrastructure.
It's a massive problem with everything.
OpenAi is getting tons of training data from your daily coding queries vs what works + codex etc. But the knowledge is 100% private forever