Future of Research

Frontier Acceleration: Where AI Is Speeding Discovery

The fields where AI is producing measurable acceleration of frontier discovery — and where progress remains stubbornly hard.

Risograph illustration of AI-augmented research environment — Plate / AI is becoming a junior collaborator in most active research fields.

Executive summary

AI is producing measurable acceleration in fields with abundant data, well-defined problems, and tight feedback loops — protein structure, materials, parts of mathematics, and computational biology. Acceleration is much weaker in fields with sparse data, contested problem framing, or slow feedback.

Key concepts

Data abundance
Problem framing
Feedback loops
Where acceleration is real
Where it is not

Where acceleration is clearest

Structural biology (AlphaFold), materials discovery, parts of mathematics, computational biology, and weather forecasting. All share abundant data and well-defined problems.

Where acceleration is limited

Fields with sparse data, contested measurement, or slow feedback loops — much of social science, clinical research at scale, and long-horizon physics — see slower gains.

Implications for funders

Funders are increasingly favouring projects that combine domain expertise with AI capability. Pure-AI projects without domain depth fare worse.

Implications for scientists

Choose problems where AI can compound your effort. Build the AI fluency to use it well.

Key takeaways

01Acceleration concentrates in data-rich, well-defined fields.
02Sparse-data and contested fields see slower gains.
03Funders favour AI-plus-domain combinations.
04Individual scientists benefit most from AI fluency in their own domain.

Frequently asked questions

Will AGI 'solve science'?

Some bounded scientific problems, plausibly. Many problems are about framing, measurement, and meaning, which AI does not solve on its own.

What field should I work in?

One you care about. AGI does not change the importance of working on problems you find meaningful.