Simons Institute report offers TCS a 12-point AI playbook
TL;DR
- The Simons Institute released 'AI and TCS: The Next Six Months' on September 28, distilling 134 survey responses into 12 near-term recommendations that achieved broad consensus.
- Proposals include a standardized AI methodology section in every paper, capping author submissions to conferences at five, and covering LLM expenses from research grants.
- PhD programs should build in training and evaluation components performed without AI assistance, and hiring committees should evaluate candidates on no more than three works.
The Simons Institute for the Theory of Computing has published a report distilling 134 survey responses and two days of in-person discussion into twelve near-term recommendations for how theoretical computer scientists should adapt to advances in AI.
The working group met September 9-10, 2026 at the Simons Institute and was organized by Tony Feng, Zeph Landau, Amit Sahai and Nikhil Srivastava. The report, titled "AI and TCS: The Next Six Months," was released September 28. Participants included Cynthia Dwork, Tim Roughgarden, Moses Charikar, Jelani Nelson, David Woodruff and Shayan Oveis Gharan.
The recommendations bite into everyday research practice. Papers should carry a standardized methodology section describing any AI use, and trace the lineage of AI-generated ideas. Conferences should cap author submissions at five, post to arXiv alongside submission, and experiment with video talks. Hiring committees should evaluate candidates on no more than three works. PhD programs should include training and evaluation components performed without AI assistance. Research grants should cover LLM expenses.
On exposition, the report is blunt: a paper "may be rejected with a one-sentence review...solely because of exposition that is sufficiently unclear."
None of this binds anything yet. Individual conferences, departments and funding agencies would each have to adopt the recommendations for them to take effect. Terry Tao noted the release on his blog this week, and two researchers we track have already shared the report.
Shared on Bluesky by 2 AI experts
-
Earlier this month, we hosted a working group to think about how the TCS community should respond to AI progress. Their report, released on Monday, highlights 12 near-term actions that received broad support among the pa…
View on Bluesky →
Originally reported by simons.berkeley.edu
Read the original article →Original headline: AI + TCS Working Group