Where Is All the A.I.-Driven Scientific Progress? - Hard Fork Recap

Podcast: Hard Fork

Published: 2025-12-26

Duration: 39 minutes

Guests: Sam Rodriques

Summary

AI promises to revolutionize scientific discovery, but its impact remains uneven. While tools like Kosmos can significantly speed up data analysis, claims of AI curing diseases are hindered by the slow pace of clinical trials.

What Happened

Sam Rodriques, founder and CEO of FutureHouse and Edison Scientific, introduces Kosmos, an AI agent designed to perform months' worth of research in just 12 hours. Kosmos orchestrates hundreds of AI agents to achieve specific research goals, costing $200 per prompt due to its intensive computational demands.

Despite the potential for rapid data analysis, Rodriques points out that AI's impact on curing diseases is limited by the slow process of clinical trials. While AI can optimize trial planning, the actual bottleneck in medical progress is the trials themselves.

Generative AI models are emerging as a new capability, allowing for the creation of novel organisms and proteins. This technology has led to significant advancements, such as de novo antibody design, which streamlines the process of developing new antibodies.

Rodriques argues that while AI can uncover new insights from existing data, it cannot single-handedly solve complex problems like Alzheimer's disease. The role of AI is to assist human researchers by preserving serendipity and enhancing discovery where information is already available.

AI tools are slowly changing the landscape for working scientists, though the adoption of new methods is often slow due to established protocols in fields like biology. Coding, a historical bottleneck, is being addressed by AI tools like those developed by OpenAI.

The episode also highlights how the AI landscape is defined by clear benchmarks, such as those used in mathematics, which provide an infinite range of problems to solve. This structured environment contrasts with the slower adoption of AI in other scientific areas.

Key Insights

• Kosmos is an AI agent capable of conducting months' worth of research in just 12 hours by orchestrating hundreds of AI agents, with each prompt costing $200 due to high computational demands.
• Generative AI models are enabling the creation of novel organisms and proteins, leading to advancements such as de novo antibody design, which streamlines the development of new antibodies.
• AI's impact on disease treatment is limited by the slow process of clinical trials, which remain the primary bottleneck in medical progress despite AI's ability to optimize trial planning.
• AI tools are gradually transforming scientific research, with coding bottlenecks being addressed by AI solutions like those from OpenAI, though adoption is slowed by established protocols in fields like biology.