ESM3 AI Model Simulates 500 Million Years of Evolution: A Breakthrough in Synthetic Biology

Scientists at EvolutionaryScale used their AI model, ESM3, to simulate 500 million years of evolution, creating a new green fluorescent protein (GFP) that would have taken nature 500 million years to develop. This breakthrough could solve pressing environmental issues, like plastic waste, by engineering synthetic proteins.

In a groundbreaking achievement, the AI model ESM3 has successfully simulated 500 million years of evolution, creating a new green fluorescent protein (GFP) that is 58% similar to its closest real-world cousin. This milestone was achieved by EvolutionaryScale, an AI startup founded by former Meta engineers and backed by Amazon and NVIDIA.
ESM3, the third model in the EvolutionaryScale series, was trained on 3.15 billion protein sequences and 771 billion tokens. It fills in gaps in incomplete protein code, designing proteins that could exist based on all potential evolutionary pathways. The new GFP, nicknamed esmGFP, has a genetic sequence that would have taken nature roughly 500 million years to develop.
This achievement is significant because it opens up new possibilities for synthetic biology. The model can be used to engineer proteins that solve pressing environmental issues, such as breaking down plastic waste. By leveraging large language models, scientists can now create biological solutions more efficiently than ever before.
The study, published in the journal Science, highlights the potential of AI in evolutionary biology. It challenges the idea of determinism in evolution, showing that even with selective pressures, evolution can be unpredictable. The team behind ESM3 is excited to share this tool with the scientific community, hoping it will lead to groundbreaking discoveries and solutions.

1. What is ESM3?
   ESM3 is a multimodal generative AI model developed by EvolutionaryScale that simulates 500 million years of evolution to create new proteins.
2. What did ESM3 create?
   ESM3 created a new green fluorescent protein (GFP) that is 58% similar to its closest real-world cousin.

3. How long would it have taken nature to develop this protein?
   It would have taken nature roughly 500 million years to develop this protein.

4. What are the potential applications of ESM3?
   ESM3 can be used to engineer proteins that solve pressing environmental issues, such as breaking down plastic waste.

5. Who developed ESM3?
   ESM3 was developed by EvolutionaryScale, an AI startup founded by former Meta engineers and backed by Amazon and NVIDIA.

The ESM3 AI model’s ability to simulate 500 million years of evolution marks a significant breakthrough in synthetic biology. This achievement not only challenges our understanding of evolutionary determinism but also opens up new avenues for creating biological solutions to environmental problems. The potential applications of ESM3 are vast, and its impact on scientific research and environmental sustainability is promising.