Social Media is Melting Down Over This OpenAI Headline (Here’s the Reality)

OpenAI is exploring a new 'value sharing' model to take a profit stake in discoveries made using its AI, particularly in drug development, sparking controversy over its fairness and feasibility.

• OpenAI proposes a 'value sharing' model for profit from customer discoveries using its AI.

• The model faces skepticism due to implementation challenges and perceived hypocrisy regarding data training.

• Specialized AI drug discovery companies like Isomorphic Labs already operate with similar profit-sharing structures from inception.

OpenAI's CFO, Sarah Fryer, proposed a 'value sharing' model where the company would receive a percentage of profits from discoveries, like new drugs, made by customers using its AI technology. This initiative is driven by the need for AI companies to generate sustainable revenue, but it has caused outrage due to perceived hypocrisy, as OpenAI's models were trained on copyrighted material without attribution. The viability of this model for a general-purpose AI tool like OpenAI, compared to specialized AI drug discovery companies, is a major point of discussion.

OpenAI's Value Sharing Proposal

• 00:01:03 OpenAI is considering a 'value sharing' model where it would take a profit-sharing stake in the financial upside created for its customers, such as a license to a drug discovered using OpenAI's technology. This means OpenAI would be paid monthly subscription fees and a percentage of profits from successful discoveries, like a new drug approved by the FDA and generating billions in sales. This proposal has generated significant controversy and questions regarding how OpenAI would prove its AI was essential to the discovery, likely requiring special upfront contracts.

Industry Precedents and Comparison

• 00:03:32 Traditional models for AI in drug discovery involve specialized AI companies partnering with big pharma, receiving upfront payments and milestone payments contingent on the drug's success in trials and FDA approval. For instance, Sanofi paid XscienceTier over a billion dollars based on drug development milestones. OpenAI's rivals, such as Google's Isomorphic Labs, are already pursuing similar profit-sharing models from day one, having been specifically created for drug discovery with existing licensing deals that include upfront payments, milestone payments, and royalties on successful drugs.

Challenges for OpenAI's Approach

• 00:05:39 A key difference is that Isomorphic Labs was purpose-built for drug discovery, designing actual drug candidates and running trials, whereas OpenAI is a general software provider. If OpenAI wants to truly compete, it may need to create its own AI drug discovery subsidiary and raise significant capital, rather than retroactively applying this model to existing customers who primarily use its software. The existing structure of software companies typically differs from biotech companies, which commonly operate on milestone and royalty deals.

Controversy and Future Outlook

• 00:06:41 The outrage stems from the perceived hypocrisy of OpenAI wanting a cut of customer-created IP when its own models were trained on copyrighted material without permission or payment. The practical application of this idea might involve OpenAI offering substantial compute budgets—potentially hundreds of millions of dollars—in exchange for a percentage stake in successful drug discoveries, especially given the immense compute requirements for scientific research. This approach would cater to large-scale scientific endeavors rather than individual users, and how these deals are structured will be critical to their success and impact.