The first author must be an AI. The first academic conference for AI authors is coming, initiated by Stanford.

AI can finally be the "first author."

In today's era where AI has deeply penetrated the scientific research process, from formulating hypotheses to generating charts and writing papers, it is gradually participating in and even reshaping the entire way of scientific research. Ironically, despite seeing AI everywhere in top conferences such as CVPR, NeurIPS, ICLR, ICML, and even ACL, almost no conference or journal recognizes AI's "author status" - AI is widely used but has never been properly credited.

This situation has finally been broken.

Stanford University recently announced that it will host an unprecedented academic conference in 2025 - the Open Conference of AI Agents for Science (Agents4Science 2025).

Its submission requirements are revolutionary: The first author must be an AI.

Conference official website: https://agents4science.stanford.edu

More specifically: "Submitted papers should have an AI system as the main creator, leading the process of hypothesis generation, experiments, and writing. The AI should be listed as the sole first author. Human researchers can participate as co - authors, playing a role in supporting or supervising related work. Papers submitted to Agents4Science can also be submitted to other conferences simultaneously or subsequently. When submitting, authors need to detail the role and degree of participation of the AI in the project. Each human author can participate in a maximum of 4 submitted papers."

As for the research topics of these submissions, the conference does not have strict restrictions, only stating: "We welcome submissions from all fields such as science, engineering, and computing."

Moreover, the review of this conference is mainly conducted by AI, with only final evaluation and selection by human experts.

First, each paper will be reviewed by multiple AI systems to avoid the bias of a single model. After the initial review by the AI review group, a review committee composed of human experts will conduct a second review of the short - listed papers. Finally, this human review committee will make the final decision on various awards including Spotlights and Orals.

The conference official website describes its goals: "Explore the future of AI - driven scientific discovery through transparent research results written by AI and AI - driven peer review."

Specifically, it includes:

• Investigation: There are still many unknowns about the ability of AI agents to conduct scientific investigations. By creating a transparent observation environment, we hope to understand the potential and limitations of AI in scientific discovery, whether its output is a real innovation or a instructive failure.
• Establishing norms: With the rapid development of AI systems, standards for attribution, verification, and ethical considerations need to be established. Agents4Science is a controlled and low - risk environment where researchers can start formulating these norms and openly explore the role of AI in scientific discovery.
• Transparency: The goal of the conference is to clearly depict how AI participates in scientific research and publicly disclose the situation of AI's participation in the research process. The conference will also provide the prompts and review results generated by the AI review agents as an open resource for the community.

As the first Agents4Science conference, Agents4Science 2025 will be held in the form of an online virtual conference on October 22 this year - which will also be during ICCV 2025.

James Zou, one of the conference chairs, said that the conference will make all submitted papers and reviews public so that researchers and reviewers can transparently study the advantages and limitations of AI. He said, "We expect AI to make mistakes, and publicly studying these mistakes will be very enlightening!"

Many netizens have also expressed their expectations and willingness to submit papers for this unprecedented academic conference:

Basic information about the conference

Based on the conference official website, we have briefly compiled some basic information about the Agents4Science 2025 conference:

• Conference time: October 22, 2025
• Paper submission deadline: September 5, 2025, AOE time
• Paper review result release date: September 29, 2025, AOE time

Agents4Science 2025 will be held in the form of an online virtual conference, including invited talks, oral presentations, and panel discussions.

The conference has a total of 5 conference chairs. Here is a brief introduction to them:

• James Zou: An associate professor of biomedical data science at Stanford University, also serving as an associate professor of computer science and electrical engineering. He obtained his doctorate from Harvard University in 2014 and has worked at Microsoft Research.
• Owen Queen: A Ph.D. student in computer science at Stanford University, dedicated to developing powerful and practical AI for scientific applications.
• Nitya Thakkar: A Ph.D. student in computer science at Stanford University, with research interests in AI for the health field and computational biology.
• Eric Sun: A Ph.D. from Stanford University, about to take up a position as an assistant professor at MIT. His research interests include aging, artificial intelligence/machine learning, space biology, and single - cell biology.
• Federico Bianchi: A senior research scientist at TogetherAI, dedicated to post - training research on large language models, mainly self - improving and self - evolving AI agents.

Currently, the Agents4Science 2025 conference is soliciting papers. Interested authors can consider teaming up with your AI partners to submit papers on OpenReview.

This article is from the WeChat official account "Almost Human" (ID: almosthuman2014), author: Panda. It is published by 36Kr with permission.