Demis Hassabis On What AI Will Do Next

2026-05-25 · Source: Two Minute Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

Demis Hassabis, CEO of DeepMind, discusses the evolving role of AI, highlighting Gemini's impact on health, including life-saving advice. He personally uses LLMs for brainstorming and research summarization, considering them valuable sparring partners. Hassabis envisions a "second order Nobel" from AlphaFold's 3 million users and introduces "Co-Scientist," a specialized Gemini variant designed as a research assistant for hypothesis generation and data analysis, which has already improved computer science algorithms like matrix multiplication. He projects AI could cure all diseases within 10-20 years by developing a platform of "half dozen to a dozen AlphaFold level models" for drug discovery, extending beyond protein structure prediction to clinical trials. DeepMind is also partnering with EVE Online to use the game as a "sandbox" for testing AI agents in complex economic and narrative environments. Hassabis also details efforts towards "closed-loop automated discovery," including building an automated material science lab in London to test 200,000 new material designs.

Key takeaway

For AI Scientists and R&D Directors strategizing future initiatives, recognize that AI is rapidly transitioning from assistive tools to co-creative partners in scientific discovery. Your focus should shift towards developing integrated AI platforms capable of hypothesis generation, complex data analysis, and even accelerating clinical trials. Prioritize investments in automated lab environments and "AlphaFold-level models" to achieve exponential breakthroughs in areas like drug development and material science, potentially curing diseases within 10-20 years.

Key insights

AI is evolving into a co-creative partner, poised to accelerate scientific discovery and revolutionize human health within decades.

Principles

• AI excels as a collaborative brainstorming partner.
• Drug discovery can achieve exponential progress via AI.
• Games offer safe, complex environments for AI testing.

Method

Co-Scientist, a specialized Gemini, aids scientists by generating hypotheses, analyzing data, and summarizing literature, functioning as a research assistant.

In practice

• Employ LLMs for creative brainstorming and research summarization.
• Investigate AI tools for medical scan pre-evaluation.
• Utilize AI to enhance computer science algorithms.

Topics

• AI Drug Discovery
• AlphaFold
• Co-Scientist
• Automated Scientific Discovery
• AI in Gaming
• Large Language Models

Best for: AI Scientist, Research Scientist, Director of AI/ML

Related on AIssential

• He Raised $70M to Cure Every Disease With AI — AI agents, particularly specialized models, significantly accelerate scientific discovery by removing human talent bottlenecks and enabling high-throughput reasoning.
• Uniting biological toolkits for a new approach to ALS — AI-driven tools can accelerate interdisciplinary scientific collaboration, bridging expertise gaps to generate novel research pathways.
• Why We're Not Close to AI Drugs Yet | Isomorphic Labs — Drug design comprises multiple distinct ML problems, not a single one, despite AlphaFold 2's protein structure prediction success.
• NVIDIA & Eli Lilly: The AI Revolution in Drug Discovery | Jensen Huang & David Ricks — Co-designing computing stacks for specific problems accelerates applications far beyond Moore's Law.
• 🗞️ Techie uses ChatGPT and AlphaFold to build DIY mRNA cancer vaccine, saves dog — AI tools empower individuals without specialized training to conduct complex scientific research and develop novel solutions.
• Teams of AI agents boost speed of research — AI agent teams can rapidly generate plausible scientific hypotheses and identify drug candidates, significantly accelerating early-stage research.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Two Minute Papers.