(Bloomberg) — Google DeepMind has released a new version of AlphaFold, a breakthrough tool for predicting protein structures. This puts artificial intelligence software on track to make breakthroughs in biological research and power a business that Google's AI chief says is valuable. Over $100 billion.
AI systems have the potential to revolutionize healthcare and create “tremendous commercial value” for DeepMind spinout company Isomorphic Labs, CEO Demis Hassabis said in an interview with Bloomberg Television. He spoke at “I want to do both with Isomorphic. We want to build a multi-hundred billion dollar business, and it has the potential to be incredibly beneficial to society and humanity. ” was previously built to commercialize DeepMind's AI for drug discovery. DeepMind first released his AlphaFold in 2018, making progress in deciphering protein shapes, a scientific problem often compared to mapping the human genome. Now in its third iteration, AlphaFold can now model a variety of molecular structures, including DNA and RNA, and predict how they will interact.
“To really understand biology, you need to think about the interactions between different biomolecules,” Hassabis says. “And AlphaFold 3 is a big step towards that.”
Hassabis said the update, which was also published in the scientific journal Nature, was “essential for drug discovery” as it provides important insight into the types of compounds used to design and test new drugs, including vaccines. Ta. “We hope to see the first AI-designed drugs in clinical practice, probably within the next few years,” Hassabis added.
AI in healthcare is a hot field. It is based on the belief that software algorithms can be used to find and develop new drugs at significantly lower cost and much faster than the decades it currently takes. The outbreak has drawn pharmaceutical companies, venture capitalists and tech giants like Nvidia into a market estimated to be worth $50 billion. Investors have poured more than $18 billion into “AI-first” biotech companies over the past decade.
But despite more funding and advances in computing, companies working on this technology have yet to show major clinical success. Some biotech companies specializing in AI drug discovery, such as BenevolentAI and Exscientia Plc, have struggled in the public markets.
Isomorphic Labs announced its first two commercial partners in January: Eli Lilly & Co. and Novartis. The partnership is “valued at nearly $3 billion” if successful on various performance milestones, the technology company said.
Alphabet has previously tried to turn its healthcare innovations into new business opportunities with units such as Verily, a “precision health” platform that manages medical data. However, these have so far had limited commercial success.
Hassabis previously said Google would spend more than $100 billion to develop AI weapons. “The benefits of generative AI models, such as in drug discovery, will far outweigh these costs in the long run,” he said in an interview this week. DeepMind is responsible for some of Google's biggest advances in AI research. Since merging with Google's other AI division last year, DeepMind is now leading the company's efforts on Gemini, the company's basic AI model that competes with OpenAI.
To create AlphaFold 3, researchers rebuilt the program “from the ground up” using methods that compensated for the shortcomings of previous versions, such as the lack of training data, Hassabis said. The new version relies on an AI technique used in generative AI called a diffusion model, which allows computers to do things like turn text into hyper-realistic images. In a paper in Nature, DeepMind researchers write that their model predicts certain molecular interactions (proteins and nucleic acids, antibodies and antigens) with higher accuracy than existing computer models. “It's much more generalizable,” Hassabis said of the new model. “And it's much more powerful.”
Along with this update, DeepMind launches AlphaFold Server, a tool that allows outsiders to access most of the program's features. Hassabis said millions of researchers have used AlphaFold to date, although “we have to get a little technical” to effectively roll out the program. “Servers are even simpler,” he said. “If you're a biologist, you don't need to understand models or technology at all.”
DeepMind plans to make the tool available to scientists for “non-commercial use.” However, Hassabis added that Isomorphic Lab's pharmaceutical partners have access to servers and other more advanced tools.
–With assistance from Naomi Kresge.
©2024 Bloomberg LP
