A conversation with Kevin Scott: What's next in AI

2026-03-17T03:00:05.000Z·★ 100·2 min read
Microsoft CTO Kevin Scott on how large language models and generative AI are transforming the knowledge economy — from GitHub Copilot to protein folding and beyond.

Microsoft's CTO Kevin Scott shares his outlook on how large language models and generative AI will transform work, creativity, and scientific discovery — from GitHub Copilot to protein folding.

The Most Impressive AI Advances of 2022

Heading into 2022, everyone in AI anticipated impressive breakthroughs. But looking back, the magnitude of innovation was genuinely mind-blowing — light years beyond what seemed possible even a few years prior, driven almost entirely by rapid advancement in large AI models.

Three standout developments:

  1. GitHub Copilot — A large language model system that turns natural language prompts into code, dramatically boosting developer productivity and opening coding to a much broader audience.
  1. Generative image models (DALL·E 2) — Making visual creation accessible to ordinary people, giving them a visual vocabulary they didn't have before.
  1. Protein folding breakthroughs — Work with David Baker's lab at the University of Washington and the Institute for Protein Design using RoseTTAFold to advance structural biology with AI.

What's Next: 2023 and Beyond

Scott predicted that 2023 would be the most exciting year the AI community has ever had. The key thesis: the entire knowledge economy would see transformation as AI assists with repetitive aspects of intellectual labor.

From designing new molecules for medicine to manufacturing recipes from 3D models to writing and editing, AI would make work more pleasant and fulfilling across virtually every domain.

The Foundation: Cloud Infrastructure and Responsible AI

Underpinning these advances are critical investments in cloud infrastructure and a strong responsible AI approach. As models become platforms that Microsoft scales for customers, the combination of compute power and ethical frameworks becomes essential.

Key Takeaways


Source: Microsoft Blog

↗ Original source
← Previous: How our open-source AI model SpeciesNet is helping to promote wildlife conservationNext: [P] Built confidence scoring for autoresearch because keeps that don't reproduce are worse than discards →
Comments0