Execute a configured synthesis workflow
Execution has been demonstrated in a bounded physical setup; novelty, characterization and generalization need separate assessment.
SCIENCE UNDER AUTOMATION
AI in Materials science: capabilities and evidence
36/57 targets synthesized in 17 days
What has AI demonstrated?
36/57 targets synthesized in 17 days
- Test
- Corrected report: 63% success; four further outcomes inconclusive by X-ray diffraction.
- Comparison
- No matched human-laboratory control extracted.
- System and version
- A-Lab · 2023 paper, corrected January 2026
- Evidence checked
- 2026-10-20
Selected air-stable inorganic targets and one bespoke lab; product presence does not establish high purity.
What does the field require?
Synthesize selected inorganic powder targets
- Select a synthesis route
A-Lab plans within a selected air-stable target domain.
A-Lab · 2023 paper, corrected January 2026 - Execute and recover a synthesis
A bespoke laboratory obtained 36 of 57 targets in the corrected report.
A-Lab · 2023 paper, corrected January 2026 - Identify the resulting material
Manual reanalysis and uncertain phase characterization limit autonomy.
A-Lab · 2023 paper, corrected January 2026
Develop a material for a new application
- Define useful target properties
No representative autonomous application-selection evaluation.
Multi-agent autonomous materials laboratories - Synthesize and characterize across conditions
The evidence set does not cover broad synthesis conditions.
Multi-agent autonomous materials laboratories - Validate device performance and durability
End-use validation is not established.
Multi-agent autonomous materials laboratories
Research tasks
Can AI replace the whole workflow?
The results above measure specific tasks. The field assessment also requires completion of the connected workflow, including the requirements listed here. A field-wide replacement percentage has not been established.
Read the assessment method