OpenAI has unveiled the GPT-4b micro artificial intelligence model in collaboration with Retro Biosciences, which could be a breakthrough in biotechnology, according to MIT Technology Review. It can help scientists redefine protein engineering by demonstrating the ability to improve proteins that turn ordinary cells into stem cells, an important step in cellular reprogramming.
This collaboration marks OpenAI's first step into the field of biological data and the company's first public statement on the use of artificial intelligence for scientific development. This is in line with OpenAI CEO Sam Altman's recent statement that AI can "massively accelerate scientific discovery and innovation," bringing the company closer to achieving its goal of creating artificial general intelligence (AGI).
The GPT-4b micro model was trained to suggest modifications for these proteins, optimizing their functionality. According to OpenAI researchers, the AI suggestions increased the efficiency of certain Yamanaka factors by more than 50%, a significant breakthrough compared to human-designed projects.
Retro Biosciences CEO Joe Betts-Lacroix praised the model's success in laboratory testing, noting its ability to deliver practical improvements in a short time frame. "We immediately brought this model into the lab and got results in the real world," he said. The improvements proposed by AI could make cellular reprogramming more efficient, opening up new opportunities for regenerative medicine, organ creation, and anti-aging therapies.
Despite the promising results, this collaboration has reignited discussions about Sam Altman's personal investments. Earlier, The Wall Street Journal called his involvement in private tech startups an "opaque investment empire," citing possible conflicts of interest. Altman is the largest investor in Retro Biosciences, having provided the company with $180 million.
OpenAI emphasized that Altman was not directly involved in the GPT-4b micro project and confirmed that the collaboration with Retro Biosciences was a joint initiative with no financial agreements.
Although the GPT-4b micro model has not yet been released as a commercial product, its success marks a new era of AI in scientific research. Unlike models such as Google's AlphaFold, which predict protein structures, GPT-4b micro focuses on designing proteins for specific functions. This has the potential to revolutionize not only regenerative medicine, but also other areas of biology.