Generative AI needs tools to avoid copyright infringement, Databricks' Naveen Rao says — or more companies could meet Napster's fate

generative-aicopyrightdatabricksmosaicmlllm

Abstraction: Generative AI copyright risk parallels Napster; open-source training on proprietary data as solution

Key points:

Naveen Rao (formerly MosaicML, acquired by Databricks for $1.3B) compares generative AI copyright risk to Napster, which went bankrupt three years after launch due to copyright lawsuits
17 authors including George R.R. Martin sued OpenAI in September 2023 for "systematic theft on a mass scale" of training data — compared to Metallica's 2000 suit against Napster
MosaicML's thesis: companies that train on their own proprietary data in a secure, transparent, open-source environment avoid legal exposure and gain product differentiation
Databricks launched Dolly, an open-source LLM, in April 2023 to enable companies to fine-tune on their own data
Rao sees democratization of generative AI — more actors armed with the same tools — as the best defense against misuse
Differentiation from competitors like OpenAI requires companies to build their own LLMs rather than relying on shared foundation models

Connections: Databricks · Mosaicml · Openai · Generative AI · Copyright · Large Language Models