cross-posted from: https://lemmy.ml/post/5400607
This is a classic case of tragedy of the commons, where a common resource is harmed by the profit interests of individuals. The traditional example of this is a public field that cattle can graze upon. Without any limits, individual cattle owners have an incentive to overgraze the land, destroying its value to everybody.
We have commons on the internet, too. Despite all of its toxic corners, it is still full of vibrant portions that serve the public good — places like Wikipedia and Reddit forums, where volunteers often share knowledge in good faith and work hard to keep bad actors at bay.
But these commons are now being overgrazed by rapacious tech companies that seek to feed all of the human wisdom, expertise, humor, anecdotes and advice they find in these places into their for-profit A.I. systems.
Only if you consider AI-supercharged misinformation to not be harmful.
Only if you consider the entropy of human interaction on the internet to not be harmful.
Only if you consider being unable to know who is real to not be harmful.
None of those things directly harm the resources being “grazed”, and none of them are inevitable consequences of AI. If you think they are then you’re actually arguing against AI in general and not the specific way in which they’ve been trained.
You think the internet being flooded with articles, comments etc. all being written by AI whose only goals are selling shit, disseminating misinformation, and manipulating elections and opinions - with no way to know what is human and what is AI - is going to be a great environment to continue to train your AI?
You might be interested to read about Model Autophagy Disorder.
That is not a problem caused by “overgrazing” those open resources. It’s a separate problem with AI training that needs to be addressed anyway. You’re just throwing out random AI-related challenges regardless of whether they’re relevant to what’s being discussed.
Simply put, quality control is always important.