From Meta to Mastery: EvolutionaryScale Raises $142 Million to Revolutionize Protein Generation with AI
In a notable shift following Meta’s strategic downsizing in August 2023, which saw over 20,000 layoffs as part of Mark Zuckerberg’s “year of efficiency,” the disbanding of a key AI biology research team has led to a groundbreaking new venture. Alexander Rives, former head of the dissolved “AI protein team” at Meta, swiftly pivoted from the setback to spearhead EvolutionaryScale, a startup focused on utilizing large language models (LLMs) for innovative protein generation.
A New Chapter in AI and Biology EvolutionaryScale, which Rives established with a core team from Meta, is on a mission to make biology programmable. Their approach focuses on developing AI that can generate new protein structures, potentially revolutionizing fields from medicine to environmental science. For instance, their technology aims to produce proteins that could develop new drugs or break down environmental pollutants like plastic.
Impressive Seed Funding and Ambitious Goals The startup recently announced a staggering $142 million in seed funding, led by tech luminaries Nat Friedman, Daniel Gross, and backed by Lux Capital, Amazon Web Services (AWS), NVentures (Nvidia’s venture capital arm), and various angel investors. This financial backing underscores the industry’s confidence in Rives and his team’s vision to redefine biological research.
Innovative AI-Driven Protein Modeling With its new generative model, ESM3, EvolutionaryScale is setting benchmarks in the field. Trained on nearly 4 billion proteins, ESM3 can analyze DNA sequences, protein structures, and functional attributes simultaneously. This model has already demonstrated its prowess by designing a new fluorescent protein, a feat that under natural conditions would take millennia of evolution to occur.
Training and Computational Challenges Building such a sophisticated model requires immense computational resources and expertise in both biology and AI. The substantial funding reflects the necessary investment to support these demanding computational needs, signaling the high stakes and high potential of EvolutionaryScale’s technology.
Competitive Landscape and Industry Impact EvolutionaryScale is not alone in its pursuit. Companies like InstaDeep and Profluent are also exploring the use of LLMs for genomics and protein design, while Google DeepMind’s AlphaFold has made significant strides in predicting protein structures. Yet, the ambition and scale of EvolutionaryScale’s vision set it apart, with plans to offer both open-access and commercial versions of their models to accelerate scientific research and pharmaceutical development.
Meta’s Legacy and Future Prospects Despite the dissolution of their team at Meta, Rives and his colleagues are grateful for the incubation period at the tech giant, which laid the groundwork for their current endeavors. With a team described by investors as a “dream team” of AI and biology, EvolutionaryScale is poised to lead a new era of bioengineering innovation.
As EvolutionaryScale advances, it embodies the potential of AI to transform not just digital landscapes but the very building blocks of life, promising a future where the possibilities of biotechnology are as limitless as the imagination of those programming it.