ChatGPT users have plenty of options to choose from, especially if they subscribe to one of the premium tiers OpenAI offers. GPT-4o is the base model supporting multimodal input and Advanced Voice Mode. A version of it built around Tasks gives you access to a limited AI agent that can do specific things for you.
Then there are the reasoning models, from the existing o1 to the upcoming o3. Finally, there’s the more sophisticated Operator AI agent ChatGPT model that’s supposed to launch soon, and the GPT-5 upgrade coming at some point down the road
But chances are that you’re not familiar with GPT-4b micro unless you’re involved in Retro Biosciences’ longevity work. The startup, which Sam Altman is an investor in, is studying ways to add years to your life. The AI model OpenAI created for the company is supposed to help with that, and it looks like GPT-4b works as intended.
Retro wants to extend the human lifespan by 10 years, and Sam Altman invested $180 million of his own money to help fund the venture. According to MIT Technology Review, Retro is working on the Yamanaka factors, or a set of proteins that can turn a human skin cell into a young stem cell.
Stem cells can then morph into any type of cell in the human body, which can lead to new therapies for certain conditions. They could also help create artificial organs or replace cells where they might be needed.
The report notes that human scientists will see slow progress without the help of AI. Reprogramming cells might take weeks. Fewer than 1% of cells treated in a lab will make the entire rejuvenation journey.
That’s where GPT-4b micro comes in, the model OpenAI crafted for the startup. What the AI should help with is the creation of protein, which is a key step in the startup’s research. The small model was trained to suggest ways to reengineer the protein factors so they can increase their function.
The Retro researchers used the AI’s suggestions to change two of the Yamanaka factors to be 50 times more effective. The studies that demonstrate this massive progress aren’t out yet, but the preliminary results seem promising.
GPT-4b is a micro model compared to the other ChatGPT options as it was only trained on protein data. The AI saw examples of protein sequences from various species and information about how proteins interact with each other.
The Retro scientists used to suggest redesigns of the Yamanaka proteins, and the AI delivered options that could be then tested in the real world.
It’s too early to tell whether GPT-4b will be responsible for adding years to the human lifespan. But it could set the stage for bigger innovations that might come down the line. I wouldn’t be surprised if more advanced AI models will be used in the future to help researchers create increasingly more efficient Yamanaka factors that can reliably transform human skin cells into stem cells.
It might be years until Retro comes out with commercial products that can extend longevity. And when it does, access to such health tech might not come cheap, but we’ll cross that bridge when we get there. This is all speculation, of course.
Also, it’s unclear how GPT-4b makes its guesses, which is also intriguing and something OpenAI will probably study. But, at the very least, GPT-4b proves that OpenAI might be sitting on an AI goldmine that most ChatGPT users won’t get to interact with. The company could create similar ChatGPT models for the enterprise to assist with very specific tasks.
As exciting as this might all be, the obvious controversy surrounds this AI-related project. Altman has every reason to want Retro’s tech to work.
OpenAI and Retro do not have a financial contact. No money changed hands in the deal. That wouldn’t be a problem if Altman weren’t such a prominent investor in the latter. OpenAI told MIT that Altman was not involved in the work and that it never makes decisions based on Altman’s other investments.
