Guest blog (SandboxAQ): after ChatGPT, what’s the next evolution of AI?
For nearly two years, generative AI and large language models (LLMs) have grabbed headlines, captivated users, and driven numerous innovations. However, the next generation of AI – Large Quantitative Models (LQMs) – is already emerging and its impact on business and society will be even greater than that of LLMs.
LLMs excel at creating content or deriving insights from textual or visual data, so their impact is focused mainly around our digital world. LQMs, on the other hand, impact our physical world, leveraging physics-based data to generate new insights and develop new materials for use in biopharma, chemicals, automotive, aerospace and other sectors. They can also be used in banking and insurance, leveraging financial and other data to optimize investment portfolios and manage risk exposure. LLMs are not well-suited for these tasks and are better used for marketing, customer service, summarizing text, drafting documents, and similar activities.
Powered by cutting-edge GPUs, LQMs have achieved unprecedented advances in computational chemistry calculations, which are used to accelerate both drug discovery and materials science. Clinical and technological breakthroughs that were seemingly impossible 24 months ago are now transforming industries and pushing the boundaries of what is possible with AI. In order to extract the maximum benefit from AI, government agencies and enterprises now realize they need to expand beyond the capabilities and limitations of LLMs and explore LQMs.
Overcoming the AI data deficit in drug discovery
According to Epoch AI, LLM models could run out of training data by 2028. While this poses a significant challenge for LLM innovation and scalability, LQMs are not similarly affected. In fact, they can access a near infinite supply of synthetic data generated by running Quantitative AI simulations on vast amounts of chemistry data. The insights gleaned from these physics-based simulations help predict how molecules interact and perform in various scenarios.
This capability opens new frontiers in drug discovery, particularly for challenging medical conditions for which large amounts of data does not exist. For example, researchers at the UCSF Institute for Neurodegenerative Diseases have achieved unprecedented breakthroughs in Alzheimer’s research using LQMs. After other advanced computational tools failed to produce results, LQMs generated a 50x to 100x increase in hit rates (i.e., positive interactions between drug compounds and their biological targets) and identified many promising new compounds that otherwise would have taken decades to develop.
Using LQMs and powerful computers to simulate a drug’s potential efficacy, toxicity, solubility and other traits, biopharma companies can identify and focus R&D efforts on only the most promising compounds. This approach can significantly reduce the 10-15 years and billions of dollars it takes to bring new drugs to market, and improve the 88% failure rate they experience before or during clinical trials.
Creating better, stronger, safer products
Beyond biopharma, LQMs are having a similar impact in other sectors such as chemicals, energy, manufacturing, defense and more. The same innovative technologies and approaches can be used to lower the time, cost and risk of creating better, safer and more eco-friendly products and materials.
For example, leading battery technology company NOVONIX is using LQMs and Quantitative AI simulations to identify the most promising materials and chemistries for the next generation of Lithium-ion batteries, and reduce testing times by 20x – from years to months. The U.S. Army is also using LQMs to develop better batteries for EVs, unmanned vehicles, and portable power solutions, but also to discover novel metal alloys that will protect armored vehicles better while making them lighter and more fuel efficient to operate and transport.
More than just accelerating product and materials development, LQMs are fundamentally changing how organizations approach R&D and unlocking exciting new possibilities. While incredibly powerful and innovative, LLMs merely scratched the surface of AI’s true potential. The next evolution of AI – LQMs combined with Quantitative AI simulations – is driving a new wave of innovation that will profoundly impact business, governments, society, and our physical world.
Authors
