# OpenAI Redirects GPUs to AI Agents for Scientific Breakthroughs: Details from Sam Altman
OpenAI has shut down the video generator Sora and terminated its $1 billion partnership with Disney to focus all its computing power on developing automated AI researchers. Sam Altman explained on the Mostly Human podcast that the GPU shortage forced them to abandon integrating Sora into ChatGPT. This decision is similar to halting work on GPT-3 earlier—the company is sacrificing current projects for breakthroughs in fundamental science.
Automated AI agents are designed to compress decades of research into one year, revolutionizing the approach to scientific discoveries. Altman emphasized: it all comes down to computing resources, which are now directed exclusively toward this task.
Split with Disney: Negotiation Details
The partnership with Disney involved generating AI versions of over 200 characters—from Iron Man to Mickey Mouse—for placement on Disney+. Altman personally called Disney's new CEO Josh D'Amaro to inform him that the deal was off. "It was awful," he noted, but D'Amaro responded understandingly: "I understand".
Such a decision reflects OpenAI's tough priorities in resource allocation. Instead of entertainment content, the company is focusing on tools capable of automating the scientific process.
The Future of AI in Science and Business
Altman described a vision where AI agents completely transform research. These systems will be able to independently formulate hypotheses, conduct experiments, and analyze data, accelerating progress in biology, physics, and other fields.
He also mentioned the first $1 billion company with just one human employee, built on AI agents. This demonstrates the potential of autonomous systems in business.
The key term in Altman's mission—"abundance"—implies exponential productivity growth through AI.
- Advantages of AI Researchers: 10x acceleration of discoveries; autonomous operation with minimal oversight; integration with existing scientific pipelines.
- Risks and Challenges: GPU shortage; need for ethical oversight; dependence on data quality for training.
- Applications in Development: mid/senior specialists can integrate such agents into ML pipelines for automating feature engineering and hyperparameter optimization.
Technical Implications for AI Development
Redirecting resources to AI agents implies a focus on multi-agent systems with long-term planning. These agents use reinforcement learning and LLMs for hierarchical task decomposition, similar to OpenAI's o1 models.
For developers, this means growing demand for tools like LangChain or AutoGen for agent orchestration. In scientific applications, integration with simulators and databases like PubChem or arXiv is expected.
The scale of computations for training such systems requires clusters with thousands of GPUs—a typical scenario for frontier models.
Key Takeaways
- OpenAI is sacrificing Sora and the Disney deal due to GPU shortage, focusing on AI researchers to accelerate science.
- Automated agents compress decades of discoveries into a year, operating autonomously.
- Altman personally terminated the partnership; Disney's CEO expressed understanding.
- Emergence of $1 billion companies on AI agents without a staff of employees.
- Mission: creating "abundance" through AI productivity.
— Editorial Team
No comments yet.