This whole case is about the letter of the law—did the model see the exact string of words bypassing the tons of the NYT paywall. And does the mod output precisely the article and not paraphrase it or some other use.
What you didn’t touch on, Zvi, was that copyright law exists to encourage the creation of new works by a “temporary” license.
Capable AI tools make creation of new work cheaper, dramatically so for new works that fit the current limitations of the tools.
Basically the current laws aren’t appropriate to the new situation and the NYT wants to get paid.
Another thing you didn’t touch on was what value is the NYT offering. Now that “NYT style” articles are cheap to write, and you could likely scaffold together more model instances to provide fact checking, the main value the times offers is 2 things:
The name and url/physical paper and continuous reputation. In a world of cheap misinformation credibility is everything. The Times won’t usually post outright made up information. No entity that hasn’t existed for 172+ years can claim this.
New information. When the Times directly collects new information with reporters and interviews, this is something current AI cannot generate.
This whole case is about the letter of the law—did the model see the exact string of words bypassing the tons of the NYT paywall. And does the mod output precisely the article and not paraphrase it or some other use.
What you didn’t touch on, Zvi, was that copyright law exists to encourage the creation of new works by a “temporary” license.
Capable AI tools make creation of new work cheaper, dramatically so for new works that fit the current limitations of the tools.
Basically the current laws aren’t appropriate to the new situation and the NYT wants to get paid.
Another thing you didn’t touch on was what value is the NYT offering. Now that “NYT style” articles are cheap to write, and you could likely scaffold together more model instances to provide fact checking, the main value the times offers is 2 things:
The name and url/physical paper and continuous reputation. In a world of cheap misinformation credibility is everything. The Times won’t usually post outright made up information. No entity that hasn’t existed for 172+ years can claim this.
New information. When the Times directly collects new information with reporters and interviews, this is something current AI cannot generate.