Key Moments
Stanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Applications, AI in Life Sciences
Stanford OnlineWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
AI is poised to revolutionize drug discovery by drastically shortening development timelines from 10-15 years to just a few. While the technology shows incredible promise, the true impact hinges on integrating AI with wet lab experiments and streamlining complex clinical trial processes.
Key Insights
The average drug development timeline is 10-15 years, with AI aiming to compress this to under 5 years, and potentially even fewer.
The drug development process has 5-10 significant bottlenecks, not just one, requiring AI to address multiple stages from target identification to clinical trials.
Currently, only about 30 new drug targets enter clinical trials annually, while thousands are available, highlighting a massive unmet need that AI could address.
Chai Discovery focuses on molecular generation, aiming for 'zero-shot' drug discovery, meaning drugs ready for patients directly from computer design.
Anthropic's vision is to train Claude to accelerate the entire R&D process, from basic research to clinical development and manufacturing, aiming for order-of-magnitude acceleration.
The convergence of large language models, other foundation models, and massive biology data generation is the key driver behind the current AI revolution in life sciences.
The ambitious goal: drug discovery in years, not decades
The development of a new drug typically spans 10 to 15 years from initial idea to FDA approval. This lengthy process is attributed to numerous bottlenecks, not just isolated to drug design or clinical trials. Josh, co-founder of Chai Discovery, and Eric from Anthropic, shared their insights on how Artificial Intelligence, particularly large language models (LLMs), is poised to drastically shorten this timeline. Their companies are building tools and platforms to accelerate every stage of research and development (R&D), with the ultimate aim of achieving 'zero-shot' drug discovery – designing molecules that are ready for patients directly from computational models. This ambitious goal could reduce the overall development time to under five years, or potentially even less, by tackling the core challenges in target identification, molecular design, and clinical validation.
Bottlenecks in drug development: a complex landscape
Drug development is a multi-stage process with significant challenges. As Eric Agrawal explained, the journey begins with identifying a disease and selecting a drug target, which is a molecule in the body that the drug will interact with. Currently, the industry faces 'target crowding,' with only about 30 net new targets pursued annually out of a potential thousands, limiting progress. Once a target is selected, the critical preclinical phase involves designing a drug molecule that can effectively bind to the target, be safe for patients, and have appropriate pharmacokinetic properties. Historically, this preclinical phase takes around four years. This is followed by clinical trials, typically divided into Phase 1 (safety), Phase 2 (initial efficacy), and Phase 3 (large-scale validation), taking another six to nine years. Simultaneously, manufacturing processes need to be established and validated. AI is seen as a powerful tool to accelerate multiple points within this complex pipeline.
Chai Discovery's mission: computer-aided design for molecules
Chai Discovery is focused on building a computer-aided design (CAD) suite specifically for molecules, with the long-term vision of moving much of the lab work to the computer. Their hypothesis is that one day, researchers will be able to design drug molecules, such as antibodies (which constitute about half of approved drugs), directly from a computer with zero trial and error. While most biotech companies aim to become pharmaceutical manufacturers, Chai's contrarian bet is to partner deeply with the ecosystem, providing their design tools to existing pharma companies. They believe this approach will enable more drugs to be designed using their platform. The company has made significant progress and is already working with major pharmaceutical companies. Their focus on molecular generation is seen as apex for other advancements in the field.
Anthropic's platform approach to R&D acceleration
Anthropic aims to accelerate the entire end-to-end R&D process in life sciences using their Claude models. This encompasses basic research, drug development, clinical trials, regulatory processes, and manufacturing. They are training Claude to achieve human expert performance in scientific fields like bioinformatics and chemistry, as well as in clinical and strategic aspects of drug development, such as picking targets and managing R&D programs. Beyond model capabilities, Anthropic emphasizes making these tools accessible through optimized products for life science professionals, including an upcoming 'Claude code for bio' interface designed for scientists to visualize and manipulate molecules. Their goal is an order-of-magnitude acceleration across the life sciences world, focusing initially on basic research and therapeutics development. They are even running a wet lab to test and dogfood their own AI tools in areas like metagenomic discovery.
Why now? The confluence of AI and biological data
The current surge in AI's application to life sciences is driven by several converging trends. The dramatic improvement and scalability of large language models and other foundation models are paramount. Simultaneously, there's been a massive increase in data generation capabilities in biology, with advancements in techniques like single-cell sequencing and proteomics. This data fuels the training of more sophisticated AI models. Furthermore, there's a growing willingness within the industry to experiment with AI, partly due to geopolitical pressures, particularly from China's rapid advancements in drug discovery. This confluence provides a unique window of opportunity for AI to revolutionize the field.
Value accrual and future opportunities
Historically, value in life sciences has primarily accrued to companies selling drugs, rather than those providing tools. However, both Chai and Anthropic believe this dynamic is shifting. As AI tools become more powerful and demonstrably increase success probabilities and reduce costs and timelines, more value is expected to flow to the toolmakers. The accessibility of these advanced tools is also expected to democratize drug development, potentially leading to more startups and innovation. Future opportunities lie in areas like designing non-antibody modalities, scaling the discovery of high-quality drug targets, developing more sophisticated medicines, and connecting AI directly to wet lab instruments for automated experimentation. There is also significant potential in optimizing clinical trials through AI-driven patient recruitment and administration, and in developing novel therapeutic approaches for areas like lean muscle mass and sleep disorders.
Unsolved problems and the path forward
Despite the rapid progress, significant challenges remain. One key area is expanding AI's capability to design drug modalities beyond antibodies, such as small molecules and emerging options like molecular glues and genetic medicines. This involves transforming these areas into engineering disciplines, similar to how antibody design has evolved. Another critical challenge is scaling the discovery of high-quality new drug targets, moving beyond the current 'crowding' around a limited set of targets. While LLMs excel at tasks with shorter feedback loops, applying them to the long, complex feedback cycles of drug development, especially clinical trials, requires continued innovation. The future vision involves AI not only designing molecules but also directly interfacing with lab instruments, enabling autonomous experimentation and a new era of 'programmable chemistry'.
Mentioned in This Episode
●Software & Apps
●Companies
●Organizations
●Drugs & Medications
●People Referenced
Common Questions
Shai Discovery is building a computer-aided design suite for molecules, aiming to make drug discovery an engineering problem. They partner with pharmaceutical companies to help design drugs, with the long-term goal of enabling zero-shot design of molecules ready for patients.
Topics
Mentioned in this video
An AI safety and research company developing large language models, including Claude, for applications in life sciences.
An AI research and deployment company. Josh (Shai Discovery) was an early employee, and both guests have connections to the company.
A technology company where Josh previously worked before co-founding Shai Discovery.
A partner of Anthropic's Claude for life sciences, indicating integration with lab workflows.
A company whose models are used by large pharma companies like AstraZeneca and Pfizer for drug design.
A pharmaceutical company that uses Recursion's models for drug design.
A pharmaceutical company that uses Recursion's models for drug design.
A molecular diagnostics company previously run by Eric Abrams.
A company that makes it easy to produce materials and execute wet lab experiments at scale, favored by AI for integration.
A company that makes it easy to produce materials and execute wet lab experiments at scale, favored by AI for integration.
A quantitative hedge fund mentioned for its Medallion Fund, used as an analogy for how Shai Discovery employees might be compensated.
A company that makes it easy to produce materials and execute wet lab experiments at scale, favored by AI for integration.
An AI model developed by Anthropic, discussed as a tool for accelerating R&D in life sciences, from basic research to drug development and operations.
A partner of Anthropic's Claude for life sciences, indicating integration with lab workflows.
A protein language model developed by Josh's team at OpenAI, which is highly cited in life sciences and drug discovery.
Mentioned as the early version of LLMs that Eric Abrams saw, which sparked his interest in applying AI to life sciences R&D.
An upcoming product from Anthropic designed for life science professionals, an interface for working scientists to visualize and interact with biological data.
More from Stanford Online
View all 88 summaries
35 minStanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Economics of Generative AI
48 minStanford CS547 HCI Seminar | Spring 2026 | Toward Ontological Multiplicity in AI and Computing
40 minStanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Infrasctructure, Enterprise AI, SaaS
48 minStanford CS547 HCI Seminar | Spring 2026 | Just-in-Time Objectives for Specialized AI Interactions
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free