While some thought replaying the tape of history was impossible, evolutionary artificial intelligence has defied the impossible to create a new-ancient protein.
The world recently witnessed an unprecedented achievement in the fields of evolutionary biology and evolutionary artificial intelligence, as a team of researchers managed to develop a large language model capable of replaying 500 million years of evolutionary history in a very short period. This scientific leap opens up new horizons for understanding how life took shape on our planet, paving the way for wide-ranging applications in genetic engineering and solving pressing environmental problems.
In recent years, incredible progress has emerged in the field of large language models powered by artificial intelligence algorithms. These models have found numerous applications in natural language processing, text classification, and other functions. However, this new use in studying biological evolution and generating novel proteins could mark a revolutionary turning point that enhances our capabilities to create vital solutions serving the environment and health.
Read also: Artificial intelligence and biology: a new approach to tackling mass insect extinction
The “EvolutionaryScale 3” large model
“EvolutionaryScale 3” is an advanced model in the field of evolutionary artificial intelligence, developed by a startup of the same name founded by former Meta engineers and backed by major companies such as Amazon and Nvidia. This model is capable of generating novel proteins by studying billions of protein sequences and data related to their structures and functions. It goes beyond simply generating proteins, extending to modeling the evolutionary path that led to their emergence.
Researchers fed the model billions of different protein sequences along with information on how they changed over time. Utilizing the immense computational power of this system, scientists simulated 500 million years of evolutionary history. According to recently published studies, the protein produced by this simulation is 58 percent similar to its closest contemporary natural protein, representing a vast temporal journey by evolutionary standards.
Promising environmental and medical applications
The green fluorescent protein (commonly known as GFP) is of critical importance, serving as an essential tool in numerous molecular biology research applications. EvolutionaryScale 3 demonstrated its ability to produce a new version of it, pointing to the potential for bioengineering to create proteins that perform more specialized functions. As the developing company opens access to the model for the scientific community, horizons expand for its use in creating proteins capable of solving complex environmental problems, such as helping degrade plastic waste or purifying water.
Expectations suggest that artificial intelligence in biology could help discover new drugs and develop more targeted, precise therapeutic methods. By analyzing protein structures and predicting subtle changes that make a drug more effective, models like EvolutionaryScale 3 can accelerate drug development processes and reduce the time and costs required to reach innovative medical solutions.
Read also: From Microsoft to Nvidia: AI agents are coming in 2025
The path of evolution: between probability and determinism
One of the prominent insights capturing scientists’ interest is the feasibility of rewinding the tape of evolution from the start. Evolutionary biologists have long wondered: “Would living organisms evolve in the same way if we turned the clock back?” Some believe in a determinism where selective pressures steer the path toward close outcomes, while others view that chance and simple contingencies could lead to entirely different results. By comparing the AI-generated protein with its natural counterparts, the team estimated a timeline indicating that this protein took roughly 500 million years of natural evolution to reach this stage, suggesting at least a partial determinism in this case.
Future prospects
What this project delivers is not limited to producing a new protein, but points to a future where researchers can use artificial intelligence techniques to forge biological solutions with a massive positive impact. With continued research and development, new generations of this model will become more accurate and adjustable, allowing the creation of advanced molecules capable of helping curb environmental problems and tackle global health challenges.
As we look forward to the expanding applications of artificial intelligence, the hope remains to use these ethical and scientific technologies for the public good. Redrawing the map of evolution may just be the beginning of a long developmental journey empowering us to face major challenges confronting our planet, whether in environmental preservation or advancing human health.
The achievement represented by the EvolutionaryScale 3 model provides a glimpse into a bright future for synthetic biology, where science and technology can collaborate to formulate radical solutions. With openness toward the scientific community, the door swings wide open to dozens of potential applications that could fundamentally transform our approaches to handling environmental and medical issues alike. This simulation spanning half a billion years reminds us of the human mind’s capability to harness science for the benefit of life and the planet.
