Text only | Text with Images

The Core

Why We Are Here => Hardware & Technology => Topic started by: rcjordan on April 04, 2025, 11:25:28 AM

Title: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: rcjordan on April 04, 2025, 11:25:28 AM
Debbie: Governments poorly regulate things that move too fast or too slowly....

https://www.theregister.com/2025/04/03/openai_copyright_bypass/
Title: Re: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: ergophobe on April 04, 2025, 08:07:27 PM
Interesting methodology.

We used to always talk about Google tweaking the dials. I wonder if OpenAI tweaking the dials will consist of spending more compute resources and engineer time on figuring out how to obfuscate sources.
Title: Re: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: rcjordan on April 04, 2025, 08:51:58 PM
>methodology

Beg forgiveness rather than ask permission.
Title: Re: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: ergophobe on April 04, 2025, 08:55:15 PM
I meant the methodology used by O'Reilly. Multiple-choice exam and see how often it knows exactly which text came from the books.
Title: Re: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: rcjordan on April 05, 2025, 02:55:58 PM
Judge Rejects Bid to to Dismiss New York Times Lawsuit Against Open AI, Microsoft

https://www.thurrott.com/microsoft/319033/judge-rejects-bid-to-to-dismiss-new-york-times-lawsuit-against-open-ai-microsoft
Title: Re: Study suggests OpenAI isn't waiting for copyright exemption • The Register
Post by: ergophobe on June 14, 2025, 08:44:49 PM
Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book
https://www.understandingai.org/p/metas-llama-31-can-recall-42-percent
Text only | Text with Images