OpenAI has taken a significant step into the world of scientific discovery by creating an artificial intelligence model designed to engineer proteins, marking its first foray into biological data. The model, called GPT-4b micro, has been hailed for its potential to revolutionize longevity science by significantly improving the efficiency of turning regular cells into stem cells.
The project began a year ago when Retro Biosciences, a San Francisco-based longevity research company, approached OpenAI. Retro’s ambitious goal is to extend human lifespan by 10 years, and its focus is on improving the effectiveness of Yamanaka factors—proteins capable of reprogramming human cells into stem cells. Stem cells are crucial for regenerative medicine as they can transform into any tissue type in the body.
Retro Biosciences’ CEO Joe Betts-Lacroix explained, “We threw this model into the lab immediately, and we got real-world results. The model’s suggestions led to improvements over the original Yamanaka factors in a substantial fraction of cases.”
Sam Altman, OpenAI’s CEO, had previously invested $180 million into Retro Biosciences, a link that led to this collaboration. Although the partnership raised questions about potential conflicts of interest, OpenAI maintains that Altman was not directly involved in the project.
Transformative Results
The Yamanaka factors have been a cornerstone of cell reprogramming research. However, their efficiency has been limited, with less than 1% of treated cells successfully completing the transformation into stem cells. OpenAI’s GPT-4b micro has shown the ability to redesign these proteins, making them more than 50 times as effective in preliminary experiments.
“Just across the board, the proteins seem better than what the scientists were able to produce by themselves,” said John Hallman, an OpenAI researcher. The AI model was trained on data from various species and examples of protein interactions, enabling it to propose significant redesigns of amino acid sequences in the proteins.
Harvard University aging researcher Vadim Gladyshev praised the model’s potential, saying, “For us, it would be extremely useful. [Skin cells] are easy to reprogram, but other cells are not. And to do it in a new species—it’s often extremely different.”
Scientific Implications
Unlike Google DeepMind’s AlphaFold, which predicts the structure of proteins, OpenAI’s model is tailored for re-engineering proteins. It offers a fresh approach to addressing inefficiencies in cell reprogramming, which researchers believe could pave the way for breakthroughs in regenerative medicine, organ regeneration, and cell replacement therapies.
“This project is meant to show that we’re serious about contributing to science,” said Aaron Jaech, one of the model’s developers at OpenAI. While the results are promising, the findings are yet to be published for peer review, leaving some questions unanswered about the broader implications of the model.
Also Read:
- Nigerian Hackers Breach UnitedHealthcare Database, Steal Data of 120 Million Americans
- US Supreme Court Upholds TikTok Ban, Set to Take Effect Sunday
- Blue Origin’s New Glenn Rocket Reaches Orbit on First Attempt
Path to Artificial General Intelligence
OpenAI views the success of GPT-4b micro as a milestone in its journey toward artificial general intelligence (AGI)—a form of AI capable of making true scientific discoveries. Sam Altman, OpenAI’s CEO, has expressed optimism about the potential of superintelligent tools to accelerate innovation beyond human capabilities. “Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own,” Altman said.
However, the project has also reignited discussions about the transparency of Altman’s wide-ranging investments in private tech startups, which some critics argue could present conflicts of interest.
Challenges and Future Prospects
While GPT-4b micro has shown impressive results, its mechanisms remain opaque. “How exactly the GPT-4b arrives at its guesses is still not clear—as is often the case with AI models,” Betts-Lacroix noted. He likened the challenge to understanding AlphaGo’s dominance in the game of Go, which took years to analyze fully.
For now, the model remains a bespoke demonstration rather than an official product, with OpenAI yet to decide whether it will integrate this capability into its broader AI tools. Jaech emphasized the importance of scientific collaboration, saying, “Whether these capabilities will come out to the world as a separate model or whether they’ll be rolled into our mainline reasoning models—that’s still to be determined.”
Despite these uncertainties, the collaboration between OpenAI and Retro Biosciences offers a glimpse into the transformative potential of AI in science. If successful, it could herald a new era in longevity research and regenerative medicine, with applications that extend far beyond the laboratory.