OpenAI and Retro Biosciences have launched a new artificial intelligence model, GPT-4b micro, with the goal of using AI technology to overcome the problem of aging and add ten years to human health. This model focuses on "Yamanaka factors" and aims to improve the efficiency of cell reprogramming, which is of great significance for delaying aging and treating age-related diseases. This collaboration combines OpenAI’s deep accumulation in the field of artificial intelligence and Retro Biosciences’ expertise in the biomedical field. It marks a major breakthrough in the application of artificial intelligence technology in the biomedical field and may change human understanding of aging and disease in the future. and treatment.
OpenAI announced the launch of its first artificial intelligence model focused on biological data - GPT-4b micro, which is designed to help humans extend their lifespan, with the goal of allowing humans to live 10 more years. According to MIT Technology Review, this model not only provides a new perspective for biomedical research, but may also lead to unexpected scientific discoveries.
OpenAI has partnered with Retro Biosciences, which was founded in 2022 and focuses on developing new drugs to combat cellular aging. The company's co-founder, Professor Ding Sheng, once served as dean of Tsinghua University School of Pharmacy. His team achieved a breakthrough result in 2022, successfully using chemical small molecules to induce the stable culture of mouse totipotent stem cells. This result was also published on Published in the famous journal "Nature".
The collaboration focuses on "Yamanaka factors," a group of proteins proposed in 2006 by Nobel laureate Shinya Yamanaka that can transform skin cells into younger-looking stem cells. However, existing methods are not efficient, often taking several weeks, and have a success rate of less than 1%. Therefore, OpenAI developed the GPT-4b micro model, aiming to improve the efficiency of cell reprogramming by generating methods to improve the function of Yamanaka factors.
The training data of this model includes protein sequences of a large number of species and information on their interactions. It is significantly different in function from Google's AlphaFold, which is mainly used to predict protein structures. Retro scientists used "few-shot" prompts to guide the model to generate possible redesigns, showing that about one-third of the amino acids suggested by the model had significant changes.
Preliminary experimental results show that the effect of the two mountain factors modified based on the model recommendations is increased by more than 50 times. Retro CEO Joe Betts-Lacroix says the recommendations provided by the model are indeed significantly better than traditional methods. Although these research results have not yet been published publicly, and outside scientists cannot verify their authenticity, OpenAI emphasizes that the success of this research may open up new paths for future cell reprogramming technology.
In the future, it has not been determined whether GPT-4b micro will be released as a standalone product or integrated with other OpenAI models. It should be noted that OpenAI made it clear that CEO Altman was not directly involved in this research, and the company’s decision-making has nothing to do with his personal investment.
Although the research results still need further verification, the emergence of GPT-4b micro undoubtedly brings new hope to anti-aging research, and its potential application value is worth looking forward to. In the future, as the technology matures and more research results are released, this technology is expected to bring revolutionary changes to human health.