Stanford CS25: V4 I Behind the Scenes of LLM Pre-training: StarCoder Use Case

[@for-ever-22]

Thank you Stanford

[@perpetual989]

Great presentation. Thank you

[@zeeshanmehdi3994]

thanks guys, love this

[@IkhukumarHazarika]

Insightful can we get the slides 🙏🏼

[@StynerDevHub]

🌸🌸🌸🌸🌸🌸

[@shariqkhatri4657]

God damn professor fine as helll

[@vocesanticae]

Scrape first, filter later, opt out last is the precise opposite of "open & responsible research". It is mass copyright violation as a matter of first principle.

[@mjms3815]

Agreed. And you need to open a GitHub issue to opt out.

[@yannickpezeu3419]

Go show us the way 😂

[@Patristica]

@@yannickpezeu3419 Watch what unfolds when AI companies are sued left and right by media and publishing companies, as well as authors' and creators' unions, not to mention governments in Europe. Watch what happens when "scrape-first, ask-later" methods and their related server logs become subject to legal discovery. ""We're too big to sue" or "we're too innovative to sue" is a reckless business strategy, but it will give lots of lawyers lots of work in the years to come, and it will destroy many AI startups in the longer term.

[@mikezooper]

Artists are inspired by other artists. That’s why LLMs are legally safe. If you’re inspired by Van Gogh you’re not breaking copyright.