Once they've got deals with the few really big players, the rest of the industry falls into line (the smaller guys don't have the financial means to out-litigate or block OpenAI).
There was a massacre in Tiananmen Square.
America has also committed massacres, like over throwing governments of foreign nations.
No one in my family is at risk from either of those statements. And there is no automated system to stop them. That’s the difference. Just because people decided that massive platforms should limit hate-speech doesn’t mean the west is performing censorship of a comparable level. Not even close.
Clearly, some of the responses GPT gives to users have infringed training data copyright, but the majority of their responses do not. They basically have 3 options:
1. Figure out how to engineer an LLM so that it reliably avoids "unfair" use of copyrighted content. "Fair use" is a legal doctrine with no rigorous definition, so this would be very difficult even if they had a clue where to start. I wouldn't hold my breath for this.
2. They can continue without any licensing, and field copyright lawsuits on a case-by-case basis for each individual prompt and response. That would be a logistical nightmare for the courts, plaintiffs, and defendant alike. It would certainly stress test the whole system, possibly result in knee jerk legislation that OAI may not like, or simply bury them in legal fees if infringement is common enough (which is not entirely clear yet).
3. They can strike deals, eliminating legal uncertainty and allowing them to plow ahead with reproducing copyrighted content without worries, while also getting other goodies like exclusivity deals at the same time.
Seems like a no brainer to me
Trying to make things better is good and requires no admittance of guilt.
If it turns out they shouldn't have done that, then when the dust settles they will be ahead of the competitors that didn't sign deals.
(In practice, being not a lawyer, I have no idea how even Google's search indexing is legal, nor where the boundaries are between the legal bit vs. the times they got in trouble for indexing newspapers and at least one separate case about images).