OpenAI reportedly considering for-profit plans, but what would that be good for? Once upon a time, OpenAI was purely a non-profit. Really. It was established in December 2015 as a non-profit AI research organization. Now it appears increasingly likely that OpenAI will become a for-profit company. OpenAI, of course, has declined to comment on multiple reports on any moves to become a for-profit entity, but …
... say it again ... It ain't nothing but a heart-breaker; (AI) Friend only to The Undertaker; Oh, AI it's an enemy to all mankind; The thought of AI blows my mind ... AI, huh (good God y'all) ... AI (uh-huh), huh (yeah, huh)! (with many thanks to Edwin Starr, Steven, and Captain ... long live the 70s!)
AI seems to work well at anything trivial. Surely, the most likely AI answers to "How do I make a profit with this stuff? are
1, Buy a printing press and the cheapest scanner you can find. Beg, borrow or steal a "Benji" (A US 100 dollar note). Then ....
2. Put together a cryptocurrency scam. You will need access to a web server, an internet domain, and a spokesman with some amount of name recognition (Aside: A squeaky clean reputation would be OK, but a few modest felony convictions might be adventageous). ...
One thousand days is approximately 2.7 years. A "few" means three, or thereabouts. Three thousand days is roughly 8.2 years.
Or in other words Mr AltMan is telling us AI is 5 - 10 years away. Just like the flying car, cold fusion and many other science fiction pipe dreams.
I was surprised he was so honest in that essay.
The article implies that authors who created material that chatbots are trained on should be compensated. Legal precedent says otherwise: when the authors of "Holy Blood Holy Grail" sued the copyright owners of "The Da Vinci Code", the case failed because while many of the ideas in Da Vinci Code were very obviously copied from Holy Blood Holy Grail, no actual text was.
https://en.wikipedia.org/wiki/The_Da_Vinci_Code#Lawsuits
Btw, I also think that in today's world, copyright and patent periods should be shortened: no ideas are truly original - they're all derivative in some way.
Many are suing because their content was copied from their servers to "insert a AI company" servers, then manipulated, aka tagged and categorise and then expose to their AI to learn from.
That copying and reusing which is violating copyright law, perhaps.
However if the AI simply learn by browsing the web, even at a billion pages per hour, i don't see that violating copyright, especially if it did it via web cam staring as a screen. Are we close to having such a free learning AI, I don't know, probably not, but who knows.
@NewThought "The article implies that authors who created material that chatbots are trained on should be compensated. Legal precedent says otherwise: when the authors of "Holy Blood Holy Grail" sued the copyright owners of "The Da Vinci Code", the case failed because while many of the ideas in Da Vinci Code were very obviously copied from Holy Blood Holy Grail,no actual text was."
Actual text is copied to make the data sets used for training, so not the same as above. Where's the legal precedent?
It would be "good" for giving Sam Altman billions of dollars basically just for existing. With the investment round they want to do at the same time, it'd be "good", for giving Microsoft direct control. It'd also be good for appealing to the naive market-based religion of a lot of the people who work there.
Difficulty: If Altman's hand-picked board does this, it'll be in blatant, criminal breach of its fiduciary duties under the charter of the existing non-profit.
Difficulty with the difficulty: It's not obvious that anybody can or will do anything about that.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
