A Survey of AI Scientists
Summary¶
A systematic survey of the "AI Scientist" paradigm — systems architected to emulate the complete scientific workflow from hypothesis generation to publishable findings — synthesising dozens of seminal works from 2022 to late 2025. The authors organise the field through a unified six-stage methodological framework: Literature Review, Idea Generation, Experimental Preparation, Experimental Execution, Scientific Writing, and Paper Generation, and identify a three-phase evolutionary trajectory from Foundational Modules through Closed-Loop Integration to Scalability, Impact, and Collaboration.
Contribution¶
A unified taxonomy and developmental periodisation of AI-Scientist systems, complete with a public project repository, plus a forward-looking agenda on robustness, generalisability, and ethical governance. The framework allows otherwise heterogeneous systems (Sakana AI Scientist, Agent Laboratory, Kosmos, Zochi etc.) to be compared along the same stages.
Method¶
Survey paper; literature mapping across 2022–2025 organised by the six-stage workflow, plus historical/evolutionary analysis.
Relevance to RISE¶
Compact survey of AI-scientist systems. Sister reference to this catalog's project-list view; useful for the bibliographic completeness of the AI-scientist line specifically.
Critique / open questions¶
A survey by its nature foregrounds claims by system builders; independent measurement of "end-to-end" capability across the six stages is acknowledged as an open challenge but not resolved. The three-phase periodisation is convenient but post-hoc and could mask competing architectural lineages.
Key quotes¶
"A new scientific paradigm, the AI Scientist, has coalesced at the intersection of artificial intelligence and epistemology, promising a fundamental shift from AI-assisted analysis to end-to-end autonomous discovery."
"This survey provides a systematic and comprehensive synthesis of this emerging domain by introducing a unified, six-stage methodological framework that deconstructs the scientific process into: Literature Review, Idea Generation, Experimental Preparation, Experimental Execution, Scientific Writing, and Paper Generation."