3 ms·

What's wrong with that? And companies do this all the time with enterprise software...

leobg 4y ago
The problem with it is that copyright was never designed to protect ideas. Society has to strike a balance. You want to protect authors, so they do publish their stuff, and get paid for it. But you also want ideas to spread, so that innovation spreads from one to many. Hence legal institutions like patents and copyright. Both carefully balance the interests of the owner with those of the public. Patents give you the right to sue people who use your idea without a valid license, but only if you publish the idea, and only for a certain amount of time. Afterwards, the idea falls into the public domain. Copyright, on the other hand, has never never protected the ideas contained in a work. Rather, copyright gives you the right to sue people who re-publish your work verbatim. However, copyright never did and never can give you the right to sue somebody because they are learning from your material. Because that would 1) be a de-facto patent without the formal requirements of a patent and 2) run counter to the goal of free speech and the circulation of ideas for the common good. Besides, copyright laws were enacted in a time where machines learning from text did not exist. So even if a society wanted to put restrictions on such a use of published content, it would require new legislation to do so. (I am a lawyer, and the above represents my personal view of the legal situation here.)
leobg 4y ago
Besides, what always gets ignored in this discussion: What about those who already did harvest publicly available information and used it to train their models? Google. Facebook. OpenAI. Now draw the curtain for everyone else, making it a legal moat for those companies‘ profits? That‘s like allowing Google to build an index of the web and then to make it illegal for everyone else to scrape publicly accessible websites.

The problem with it is that copyright was never designed to protect ideas. Society has to strike a balance. You want to protect authors, so they do publish their stuff, and get paid for it. But you also want ideas to spread, so that innovation spreads from one to many. Hence legal institutions like patents and copyright. Both carefully balance the interests of the owner with those of the public. Patents give you the right to sue people who use your idea without a valid license, but only if you publish the idea, and only for a certain amount of time. Afterwards, the idea falls into the public domain. Copyright, on the other hand, has never never protected the ideas contained in a work. Rather, copyright gives you the right to sue people who re-publish your work verbatim. However, copyright never did and never can give you the right to sue somebody because they are learning from your material. Because that would 1) be a de-facto patent without the formal requirements of a patent and 2) run counter to the goal of free speech and the circulation of ideas for the common good. Besides, copyright laws were enacted in a time where machines learning from text did not exist. So even if a society wanted to put restrictions on such a use of published content, it would require new legislation to do so. (I am a lawyer, and the above represents my personal view of the legal situation here.)

leobg 4y ago

Besides, what always gets ignored in this discussion: What about those who already did harvest publicly available information and used it to train their models? Google. Facebook. OpenAI. Now draw the curtain for everyone else, making it a legal moat for those companies‘ profits? That‘s like allowing Google to build an index of the web and then to make it illegal for everyone else to scrape publicly accessible websites.