Seven Layers · Layer 7

Scientific AI

Book section7 chaptersStatus: First edition

About this section

Scientific AI uses machine learning, reasoning systems, simulation, automation, and robotics to help produce new knowledge. Its purpose is not merely to describe what humanity already knows. It is to improve the process by which observations become explanations, designs, experiments, and discoveries.

At the lower layers, AI performs computation and work. At Layer 7, those capabilities are organized to expand what can be discovered.

This layer includes AI for biology, medicine, chemistry, materials, physics, climate, engineering, and other empirical fields. It also includes scientific foundation models, surrogate simulations, research agents, automated laboratories, and systems that search vast design spaces for useful molecules, materials, mechanisms, and machines.

Prediction

Estimate a property, outcome, structure, or future state from available evidence.

Explanation

Identify relationships, mechanisms, causal hypotheses, and interpretable scientific patterns.

Design

Search backward from desired properties to candidate molecules, materials, experiments, or engineered systems.

Experiment

Plan, execute, monitor, and learn from physical or computational experiments.

The scientific discovery loop

  1. Observe: collect measurements, images, sequences, spectra, literature, and experimental records.
  2. Represent: convert scientific objects and processes into forms a model can learn from.
  3. Hypothesize: propose relationships, mechanisms, candidates, or explanations.
  4. Predict and design: estimate outcomes or generate objects with desired properties.
  5. Test: run simulations, analyses, or physical experiments.
  6. Update: use the new evidence to revise models and choose the next experiment.

Why this is the seventh layer

Scientific AI depends on every layer beneath it. Applications provide research tools. Agents coordinate work. Domain models learn scientific representations. Model architecture supplies reasoning and multimodal capabilities. Infrastructure manages data and computation. Hardware performs simulations and experiments. Layer 7 combines these foundations into systems aimed at discovery.