Gary Marcus

Rank 14 of 47
|
Score 89
3/15/2024, 4:54:45 PM
X
In reply to:
ClockworkRelativity
@Clockrelativity
·
575d
@GaryMarcus outright change the 1976 copyright act to make sure that while you can, in fact, train AI models (even with copyrighted works), proper royalty fees are paid in due course.
2/2
ClockworkRelativity
@Clockrelativity
·
575d
@GaryMarcus I'm sure that the courts will, but there are 2 potential hurdles:

1) The Authors Guild v. Google case from 2015, which is a core Fair Use defense AI firms love.
2) HiQ v. LinkedIn pertaining to data scraping in the 9th circuit (that one got settled out of court).

We need to
1/x
Gary Marcus
@GaryMarcus
·
575d
Can we just stop saying “public data”, and start to distinguish (as the law does) between “public domain”, where license has been affirmatively granted, from “shit I found on the internet that probably isn’t licensed”?

Y’all can pretend not to understand the distinction, but I…
Romeo Lupascu
@RomeoLupascu
·
575d
@kaptankaos @GaryMarcus I agree with you just one difference. LLMs were/are trained on some public data and a training cycle is not cheap in both time and power. So every they may have access to internet they don't search it (like Google). Here is something you may find useful: https://www.romeolupascu.net/202…

The statement is part of a complex conversation about copyright law, AI, and the use of public data. The tone is informative and contributes to the dialogue without any harmful or disrespectful language. It suggests caution in interpreting a legal ruling, which is a constructive approach to the discussion.

  1. Principle 4:
    I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
    The statement engages in constructive criticism by advising careful reading of a legal ruling, which promotes a thoughtful and analytical approach to the conversation. [+1]