Google Cloud to Offer Specialist AI Models for Science Research
Google will add SandboxAQ’s so-called large quantitative models to its cloud marketplace, allowing researchers to combine Google’s Gemini language models with AI systems trained on scientific equations and laboratory data. Unlike large language models, which excel at generating text, large quantitative models are trained on numerical data to help solve problems in chemistry, biology and physics
Alphabet Inc.’s Google will start offering specialist artificial intelligence models from software company SandboxAQ through its cloud service, expanding enterprise and research access to technology designed to accelerate drug discovery, materials science and semiconductor manufacturing.
Google will add SandboxAQ’s so-called large quantitative models to its cloud marketplace, allowing researchers to combine Google’s Gemini language models with AI systems trained on scientific equations and laboratory data. Unlike large language models, which excel at generating text, large quantitative models are trained on numerical data to help solve problems in chemistry, biology and physics.
SandboxAQ, an Alphabet spinout, is among a growing crop of companies focused on using AI to try and answer thorny research questions. Others include Isomorphic Labs, founded out of Alphabet’s Google DeepMind, which is using AI for drug discovery. OpenAI has also released a model for life sciences research.
The first SandboxAQ model available through Google Cloud Marketplace, AQCat, is designed to help researchers identify promising catalyst and materials candidates for applications ranging from semiconductor manufacturing to battery development. A second model, AQPotency, will focus on drug discovery by helping researchers identify molecules most likely to bind to disease targets.
“We’re saving years, but we’re also exploring a bigger candidate space, and that means that we can find novel solutions we couldn’t find with any amount of time,” said SandboxAQ’s Chief Executive Officer Jack Hidary.
The Google partnership follows SandboxAQ’s $500 million award through the US government’s CHIPS program to develop AI models for semiconductor manufacturing. Hidary said AQCat is one of the models supporting that effort, which focuses on developing new battery chemistries, catalysts, magnets and methods for breaking down PFAS, often referred to as “forever chemicals.”
SandboxAQ originally started as a quantum tech lab in Alphabet’s X division for “moonshot” technologies, spinning out to become an independent company in 2022. Its backers include former Google CEO Eric Schmidt and Salesforce Inc. CEO Marc Benioff.
Hidary said the company relies on quantum techniques embedded in its AI models. Those techniques currently run on conventional accelerated computing hardware such as GPUs and TPUs, while future quantum computers are expected to expand their capabilities as the technology matures.
Hidary said biopharmaceutical research represents the company’s largest near-term opportunity, with customers using the technology to study cancers including glioblastoma and prostate cancer, as well as Alzheimer’s disease, Parkinson’s disease and cardiovascular conditions.
