5 Surprising Insights from OpenAI's Emails with Elon Musk

The tech community buzzed with excitement last night as OpenAI published a blog post addressing Elon Musk's recent lawsuit against the organization and its leaders, CEO Sam Altman and President Greg Brockman. Musk’s allegations include breach of contract, breach of fiduciary duty, and unfair competition, all centered around the concern that OpenAI prioritized profits and commercial interests in developing artificial general intelligence (AGI) over its responsibility to safeguard public welfare.

In a robust defense, OpenAI released emails exchanged among Musk, Altman, Brockman, and Chief Scientist Ilya Sutskever, highlighting five key insights:

1. Clarification on Openness

Sutskever emphasized that the "open" in OpenAI doesn’t strictly imply open-source. He stated, “As we get closer to building AI, it will make sense to start being less open.” He explained that while everyone should ultimately benefit from AI advancements, not sharing scientific details upfront could be acceptable.

2. Musk's Agreement on AI Openness

Responding to Sutskever, Musk simply stated “Yup,” indicating his agreement that AI developments should not necessarily be open-sourced.

3. Funding Strategies with Tesla

Musk suggested that OpenAI connect with Tesla to ensure funding, referring to OpenAI as “burning cash.” He noted that attaching OpenAI to Tesla as a “cash cow” could enhance its competitiveness against tech giants like Google.

4. For-Profit Discussions Since 2018

Although OpenAI is currently a nonprofit, discussions about a for-profit pivot—leading to its present capped profit structure—dates back to 2018. Musk forwarded an email highlighting that transitioning to a for-profit model could secure sustainable revenue and attract substantial investments, particularly through a partnership with Tesla.

5. Concerns About Safe AI Development

Sutskever expressed significant worries regarding the potential risks of open sourcing unsafe AI, especially in a 'hard takeoff' scenario. He cautioned that if building safe AI becomes more challenging than creating unsafe AI, then by making everything openly accessible, the risk of malicious actors creating dangerous AI systems increases.

OpenAI's decisive response and the insights from these emails spotlight critical considerations in the ongoing discourse surrounding AI safety and commercialization.