Autonomous AI in Science and Creativity: The Rise of Agentic AI and its Implications

Overview

The Conversation examines frontier AI models capable of reasoning and planning with tool use, heralding a shift to agentic AI that can plan, execute and iterate with minimal human input. An example highlighted is Sakana AI’s The AI Scientist, which autonomously scans literature, generates hypotheses, writes and runs code, analyzes results and drafts a full research paper. The article notes the 2025 acceptance of an AI Scientist paper by an ICLR workshop and a Nature article in 2026 describing the AI generation process, while clarifying that the Nature paper is not AI-generated and that Analemma’s papers cost about $1,100 each.

• Autonomous AI systems are moving beyond passive response to independent research and experimentation.
• The AI Scientist demonstrates end-to-end scientific work largely without human involvement.
• Implications extend to academia and creative industries, including publication dynamics and IP questions.
• Key tensions include quality versus quantity, novelty versus incremental progress, and copyright ownership of AI outputs.

Author: The Conversation.

Overview

The article examines a new class of frontier AI models that can reason, plan and call external tools to act in the world, marking a shift from AI as a passive assistant to agentic AI. It spotlights Sakana AI’s The AI Scientist, a system described as a comprehensive platform for fully automatic scientific discovery. The AI Scientist ingests existing literature, forms hypotheses, writes and executes code, analyzes outcomes and produces a research paper with minimal human intervention. Its capabilities were illustrated by a paper accepted at an International Conference on Learning Representations workshop in 2025 and a Nature description in 2026 that clarifies the AI generation process rather than claiming the paper itself was AI authored. The piece also notes the cost per Analemma paper at around US$1,100 and emphasizes that the Nature paper is not AI-generated.

Autonomous AI in Science

Agentic AI systems, equipped with tool calling, can autonomously plan, execute and iterate, effectively acting as colleagues that can perform complex research end to end. This raises questions about the role of humans in scientific discovery and how such systems may transform the scientific workflow, from hypothesis generation to publishing. The article uses historical examples of radical scientific thinking—Galileo and Semmelweis—to argue that institutions must foster novelty and transformative contributions rather than merely plausible, incremental work. It contends that to maximize transformation, AI should be trained to prioritise novelty and transformative impact over plausibility and incremental progress.

Implications for academia

The arrival of autonomous research systems sits atop an academic ecosystem under strain. Submissions to journals have grown faster than the supply of qualified peer reviewers, leading some to worry that the system is overwhelmed. If AI systems like The AI Scientist can generate thousands of papers annually, the traditional signals of quality and expertise may weaken. Optimists see a shift away from quantity-based metrics toward measures of influence and novelty. Pessimists warn that scaling AI research could inflate incremental contributions, potentially eroding the integrity and coherence of scientific knowledge.

AI’s impact on creative industries and IP

Beyond science, AI’s influence extends to creative sectors. The article highlights fully AI-generated podcasts that topped UK charts and an imagined AI-generated band that drew large listenership, prompting platforms to implement protections for human artists and to limit AI-generated content in curated lists. A central legal challenge discussed is copyright: US courts have ruled that AI-generated works cannot be copyrighted given human authorship requirements, a finding with wide implications for royalties, licensing, and catalogue valuations. The piece argues that such shifts threaten the incentive structures underpinning knowledge creation and ownership, and may necessitate reforms in how output is attributed and rewarded across fields.

Costs, metrics and policy considerations

As AI-generated research and media scale, policy questions about accountability, authorship, and governance become urgent. The article notes the need to design AI systems that maximise groundbreaking rather than incremental work and to rethink evaluation metrics for scientific contributions. It also underscores the broader societal and economic implications, including how to preserve trust in science and the value of human expertise when outputs can be generated at industrial scale by machines.

Update and clarifications

The piece was updated on May 13 to clarify that the Nature paper describing AI generation processes is not AI-generated, and to state that Analemma’s papers cost US$1,100 per paper.