Premium

OpenAI and longevity startup train GPT-4b model to extend human life

OpenAI and Retro Biosciences have trained a new AI model that can engineer proteins.

The model has been trained on examples of protein sequences across species. (Express Image/FreePik)

OpenAI is backed by Microsoft. (Express Image/FreePik)

OpenAI has reportedly developed an AI model to enhance stem cell production efficiency. According to MIT Technology Review, The Sam Altman-led AI powerhouse has trained a model called GPT-4b, in collaboration with longevity startup Retro Biosciences.

Up until now, AI’s contribution in science was mostly known to be Google DeepMind’s AlphaFold, a protein-folding programme that won the Nobel Prize for its creators last year. However, with its latest feat, OpenAI too has entered the fray.

OpenAI’s latest is a model for engineering proteins. The company claims to have developed an LLM that can visualise proteins capable of turning regular cells into stem cells. Interestingly, the model has surpassed humans at the task. While Google’s AlphaFold predicts the shape of proteins, OpenAI’s first model is custom-built for biological research.

This is OpenAI’s first model based on biological data and this is being seen as a step towards ascertaining if AI can make a difference in terms of scientific discovery. This would also mean a significant leap towards Artificial General Intelligence (AGI). Earlier this month, Altman claimed that his company was confident that his company knows how to build AGIs. The CEO had said that superintelligent tools could massively accelerate scientific discovery and innovation.

Also Read | Microsoft shows how Gen AI can aid material discovery with MatterGen AI model

GPT-4b has been trained on protein sequences from various species and data about how proteins interact. While this is a massive dataset, it is relatively smaller when compared to OpenAI’s flagship chatbots. GPT-4b can be called a small language model that has been designed for specialised tasks with a specific dataset.

According to the MIT Technology Review report, the protein engineering project began a year ago after Retro Biosciences approached OpenAI about the collaboration. Sam Altman was already linked to the company as he had personally funded it with $180 million. Retro has been working towards extending normal human lifespan by 10 years.

GPT-4b micro has been developed to re-engineer the protein factors to increase their function. According to the company, the researchers used suggestions by the model to change two of the Yamanaka factors to be more than 50 times as effective. The Yamanaka factors are four special genes that can turn a regular adult cell back into a stem cell-like state, where they can grow into almost any type of cell in the body. Simply put, it is like resetting a cell to its factory settings. The GPT-4b micro mode has been developed by a team of researchers led by OpenAI’s John Hallman and Rico Meini from Retro.

Tags:
artificial intelligence