The video clarifies that OpenAI’s proposed value-sharing model—where it would receive a share of intellectual property or revenue in exchange for providing AI compute—is intended specifically for drug discovery partnerships, not for general use of its AI models. It also raises concerns about the potential concentration of power if access to advanced AI becomes essential for innovation across industries, but remains cautiously optimistic about the future.
The video discusses recent controversy and speculation around OpenAI’s approach to intellectual property (IP), particularly in the context of scientific and medical research. The catalyst was a comment by OpenAI’s CFO, Sarah Friar, at Davos, which led to viral reactions online. Many people misunderstood her remarks, fearing that OpenAI would claim ownership over any ideas or products developed using ChatGPT, including code or business concepts. However, the video clarifies that OpenAI’s value-sharing model is being considered specifically for drug discovery and medical research, not for general use of their AI models.
The presenter explains that OpenAI is exploring partnerships with pharmaceutical companies, where OpenAI would provide AI compute resources to accelerate drug discovery in exchange for a share of the resulting IP or revenue. This model is compared to previous industry examples, such as the 2018 partnership between GSK and 23andMe, where genetic data was leveraged for drug development and profits were shared. The rationale is that drug discovery is resource-intensive and time-sensitive, so companies may be willing to share IP in order to access powerful AI tools and speed up their research.
The video also examines how IP sharing is common in academic settings, using Stanford University as an example. At Stanford, any potentially patentable invention developed by faculty or students using university resources must be assigned to the university, even if the idea was conceived outside of work hours. This is contrasted with the private sector, where employees typically retain ownership of inventions developed independently and without company resources.
Returning to OpenAI, the presenter suggests that their proposed model is akin to cutting out the financial middleman. Instead of drug companies raising capital from investors to pay for AI compute, OpenAI could directly provide compute in exchange for a stake in the resulting discoveries. This arrangement would likely be opt-in and negotiated on a case-by-case basis, rather than imposed on all users of OpenAI’s models. The video notes that other AI labs, such as Google DeepMind and Anthropic, are also exploring similar partnership models in biotech.
Finally, the presenter raises concerns about the broader implications if this model expands beyond drug discovery. If access to advanced AI and compute becomes a prerequisite for innovation across industries, those controlling the compute could accumulate disproportionate power and wealth, potentially leading to a “permanent underclass” scenario. The video ends on a speculative note, questioning the incentives for companies to sell artificial general intelligence (AGI) if they could instead use it to dominate markets, but ultimately expresses optimism that such dystopian outcomes are not inevitable.