All the content, and all the pieces, are complete. Now I just have to follow GPT’s instructions — which will be my first post on the site.
Before publishing anything a machine helped write, it’s worth knowing whose it is. OpenAI has posted terms of service that give you full rights to use the content under most circumstances:
“Input and Output are collectively ‘Content.’ As between the parties and to the extent permitted by applicable law, you own all Input, and subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output.”
Read that twice, because it’s doing two different jobs. You own what you type in. Whatever rights OpenAI has in what comes out, it hands to you. That’s generous, and it’s clear.
What the terms can’t do is manufacture a right that doesn’t exist, and as far as I can tell that’s the open question. The US Copyright Office has been consistent that copyright needs a human author. A year ago this month its Review Board upheld the refusal to register a picture called “A Recent Entrance to Paradise”, whose listed author was a “Creativity Machine.” The Office’s finding was that the work “lacks the human authorship necessary to support a copyright claim,” and the Board restated the rule: under the statute, “a work must be created by a human being.” More recently the Office registered a comic, Zarya of the Dawn, illustrated entirely with Midjourney — and then opened a proceeding to revoke the protection.
So the contract gives you everything OpenAI has. What OpenAI has, in a paragraph its model wrote unaided, may be nothing at all — and a paragraph that belongs to no one is a different thing from a paragraph that belongs to you. For a blog post, that’s a tolerable position. For a product, a brand or a book, it’s a question to settle before the money arrives.
And one more thing, because it changes what the tool is good for. GPT has been updated with much more recent data — I just got facts from November 20th, 2022. OpenAI’s release notes for January 30 say only that they’ve “upgraded the ChatGPT model with improved factuality and mathematical capabilities,” with nothing about how far forward its knowledge now runs. Whatever they changed, it seems the model’s memory now reaches later than it did.
