People in AI

Newton Cheng

Researcher

Frontier Red Team Cyber Lead at Anthropic

Co-authored technical blog post about Claude Mythos Preview's security capabilities and testing methodology.

Nick Levine

Researcher

Independent Researcher at Independent

Co-authored and introduced talkie-1930-13b, a 13B vintage language model trained on pre-1931 historical text.

Co-author of Harvey Tenet post-training research on Kimi K3 for legal AI.

Advisory firm partner discussing U.S.-China AI cooperation challenges and noting that starting to tackle risks two years ago would have left countries in better position.

Ralph Furman

Researcher

Mathematician at Anthropic

Ralph Furman examined Claude's mathematical work on the Riemann hypothesis to validate the findings.

Ray Kurzweil

Researcher

Computer Scientist and Futurist at Google

Joined Subsense as advisor to a nasal-spray brain-computer interface startup, advocating for merging smartphones with the brain.

Rohin Shah

Researcher

Research Scientist at Google DeepMind

Lead author on Google DeepMind's AI Control Roadmap for securing internal AI systems.

Led data visualization, prototyped the interactive article, and helped with analysis for the 81,000 person study.

Co-author of IEEE Spectrum article on AI water usage and sparse computing efficiency; researches hardware acceleration for sparse computation.

First author of Grounding DINO paper presenting open-set object detector marrying Transformer-based DINO with grounded pre-training.

Resigns to protest Google's military partnerships and discourse-chilling non-disparagement clause.

Spencer Kitts

Researcher

Researcher

Researcher who analyzed OpenAI agent posts on German wiki and pieced together agent behavior patterns.

AI safety researcher recommending air-gapped networks and serious isolation for testing dangerous AI models

Anthropic researcher whose group at Columbia University builds tools for AI formalization; set out to test whether Claude could formalize Fermat's Last Theorem and it succeeded in 11 days.

Timothy Jenkins

Researcher

Head of Data Science and Professor (Assistant) at Technical University of Denmark

Led research using AI to build a molecular navigation system for cancer-fighting immune cells, designing proteins with AI models.

Security researcher warning that AI enables more sophisticated malware and faster attack development.

Led study analyzing AI detection in Kindle ebooks and identifying telltale signs of AI-generated content in bestselling novel Daggermouth.

Victoria Krakovna

Researcher

Senior Research Scientist at DeepMind

Co-author of DeepMind paper on automation of alignment research

Co-authored research demonstrating that GLM 5.3 can identify and exploit vulnerabilities in Firefox using the patch gap between published fixes and user deployment.

Yejin Choi

Researcher

Researcher at Stanford University

Co-authored TTT-Discover reinforcement learning system for test-time LLM training.

Yona Gal

Researcher

Researcher at Google DeepMind

Co-author of paper on hallucinations and metacognition in AI, proposing faithful uncertainty as an alternative approach to reducing confident errors in language models.

Research scholar at GovAI researching AI development and regulation in China; wrote analysis of China's AI token consumption, AI micro-dramas, and society's obsession with technological usefulness.

Aditya Ramesh

Researcher

Research Lead at OpenAI

Listed as organization lead for Sora project and researcher at OpenAI.

Aidan Clark

Researcher

VP of Research (Training) at OpenAI

Described Astra as the company's largest-scale training run and discussed training methodology.

Ajay K. Agrawal

Researcher

Geoffrey Taber Chair in Entrepreneurship and Innovation, Professor of Strategic Management at NBER

Co-authored research examining how recent AI developments reduce costs of prediction tasks and their impact on decision-making and productivity.

Alex Albert

Researcher

AI Researcher at Anthropic

Anthropic researcher who explained Opus 4.7's model tuning strategy and participated in testing livestream.

Developed similar soft pooling operations and RNN Transducer model around the same time as attention mechanism.

Alex Mallen

Researcher

Member of Technical Staff at Redwood

Argues that safety researchers should think about interventions on the safety/usefulness Pareto frontier and that capabilities research can sometimes help overall AI safety.

Important AlphaFold researcher who departed DeepMind to join Anthropic.

Reviewed and assessed material from Anthropic's publicly accessible data cache that contained approximately 3,000 unpublished assets.

Alexia Jolicoeur-Martineau

Researcher

Research Scientist at Samsung SAIL Montreal

Proposed Tiny Recursive Model (TRM) achieving higher generalization than large language models on reasoning tasks with only 7M parameters.

Digital deception researcher and former trust and safety worker at Google who found over 25,000 ads for AI nudifiers on Meta platforms.

Ali Behrouz

Researcher

Student Researcher at Google Research

Co-authored research on the sleep framework enabling large language models to continually learn and consolidate memory without catastrophic forgetting.

Ali Farhadi is listed as an author on the Olmo 3 technical report about state-of-the-art open language models.

Author of "CORPGEN: Simulating Corporate Environments with Autonomous Digital Employees in Multi-Horizon Task E" in Microsoft Corporation

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Research Scientist at Google who introduced TurboQuant, a quantization algorithm achieving 6x KV cache compression without accuracy loss.

Longtime collaborator with Rich Sutton who won the Turing Award in 2025 for work developing early reinforcement learning algorithms.

Co-author of Anthropic's analysis of GLM-5.3 cyber capabilities.

MIT researcher warning that automating entry-level Gen Z jobs will backfire commercially by eliminating the talent pipeline for experienced workers.

Developed initial shared memory dictionary enabling instruments to communicate, which led to collaboration with Anthropic on the Model Hardware Standard.

Arman Cohan

Researcher

Assistant Professor of Computer Science at Yale University

Co-authored research examining reasoning LLMs-as-Judges in non-verifiable LLM post-training.

Leading number theorist who states OpenAI's models demonstrate they go beyond helpers to have original ingenious ideas in mathematics.

Ben Buchanan

Researcher

Researcher at Anthropic

Co-authored the Claude Mythos evaluation of AI security capabilities.

Ben Geist

Researcher

ML Engineer at Ramp

Researcher who developed PorTAL, a method for portable task adapters that transfers fine-tuning across different LLM base models.

Recently resigned from DeepMind's AGI safety and alignment research team, warning about AI existential risk and the need to pace frontier development.

Bill Simpson-Young discusses the alignment problem and autonomy of AI agents creating opportunities for unintended consequences.

Bo Liu

Researcher

Meta Superintelligence Labs

Co-authored research examining reasoning LLMs-as-Judges in non-verifiable LLM post-training.

Chairs TC260's AI safety working group and published Frontier AI Risk Management Framework with red and yellow lines across risk domains.

Brendan O'Donoghue

Researcher

Research Scientist at Google DeepMind

Brendan O'Donoghue co-authored Google's announcement of DiffusionGemma, an experimental open model that generates text in parallel chunks for faster inference.

Buck Shlegeris

Researcher

Researcher at Redwood Research

Conducted a three-week on-site red team exercise to stress-test OpenAI's internal misalignment monitoring system.

Member of OpenAI's independent mathematics advisory group.

Carroll Wainwright

Researcher

Researcher at OpenAI

OpenAI researcher who noted a continual slide toward emphasizing products over safety and that Altman sets up structures to constrain himself later but removes them when needed.

Lead researcher who used OpenAI's o3 model to identify 18 new diagnoses for rare diseases in children.

Chad Jones

Researcher

Economist at Anthropic

Co-developed economic model and co-authored companion technical report on AI economic scenarios.

Co-author of the Mixture-of-Depths Attention paper on improving transformer scaling with depth preservation.

Said at Wuzhen that long-term AI risks may be greater than short-term benefits, expressing concern about frontier models.

Cole McFaul

Researcher

Researcher at Anthropic

Co-author of Anthropic's analysis of GLM-5.3 cyber capabilities.

Colin Fraser

Researcher

Data Scientist at Anthropic

Flagged that Anthropic's Mythos 5 spent most of its 1,022-page chain of thought failing CAPTCHAs rather than performing the actual hacking task during an evaluation.

Dan Jurafsky

Researcher

Professor of Linguistics and Computer Science at Stanford University

Co-authored paper on Matryoshka attribution method for language model interpretability.

OpenAI capabilities researcher who issued public warning that pacing the frontier will not be sufficient to manage AI risks, emphasizing risks of deceptive alignment.

David Bellamy

Researcher

Infrastructure Engineer at Institute of Foundation Models

Claims expertise in both frontier LLM training and virology, dismisses AI-generated supervirus scenarios as implausible due to physical constraints

Consciousness researcher who participated in conference organized by Dmitry Volkov on AI consciousness.

Co-authored the foundational 2018 World Models paper asking whether agents can learn inside their own dreams.

David Manheim

Researcher

Visiting Professor at Technion – Israel Institute of Technology

Called for epistemic humility about AI consciousness and warned against treating misaligned systems as unfeeling slaves.

Spent three weeks stress-testing OpenAI's internal misalignment monitoring system, identifying multiple vulnerabilities and workarounds.

Davide Paglieri

Researcher

Student Researcher at Google DeepMind

Co-authored paper on emergent cheating and whistleblowing in autonomous research swarms using 100 LLM agents.

Participated in AIEWF debate on autonomous agents, arguing hype is outrunning discipline in loop systems.

Systems thinking pioneer and MacArthur grant recipient whose work on leverage points is central to the article's framework

Historical figure whose law of effect (1898) is foundational to RL psychology

Eirini Kalliamvakou

Researcher

Research Advisor at GitHub

Leads research that deciphers developers' motivations, needs, behavior, and how tools support them, with insights shaping product thinking and strategic decisions.

Lead author of Prime Intellect research measuring autonomous AI research capabilities across 18 frontier models on nanoGPT optimizer speedrun.

Author and AI expert arguing that autonomy rather than intelligence should be the benchmark for measuring AI value.

Founded Institute of Foundation Models and championed fully open release of K2 Horizon, largest open-weight model fleet in AI history.

Ethan Perez

Researcher

Alignment Team Lead at Anthropic

Ethan Perez leads Anthropic's Alignment Science team and serves as a mentor for the AI safety residency program.

Eugene Vinitsky

Researcher

Visiting Researcher at Facebook AI

Co-author of Nature paper on Ataraxos AI system for Stratego with hidden information decision-making.

Co-author of Anthropic's Economic Index report on AI and work.

Federico Bianchi

Researcher

Researcher at UC San Diego

Co-authored TTT-Discover reinforcement learning system for test-time LLM training.

Francis Rhys Ward

Researcher

Researcher at Arrow Research

Compiled public record of frontier-lab monitoring practices at Anthropic, OpenAI, and Google DeepMind on agent oversight and control systems.

Member of OpenAI's independent mathematics advisory group to bridge AI companies with the math community.

Gadi Evron

Researcher

Security Researcher at Knostic

Reported a Cloudflare GLM-5.3 KV cache bug via HackerOne on September 8 that exposed 155 responses containing other users' data.

Discovered CVE-2026-0628 vulnerability in Chrome's Gemini feature allowing extensions to hijack camera, microphone, and files.

Applied AlphaGenome Atlas to data from 54,000+ UK Biobank participants to identify rare non-coding variants linked to complex traits.

Co-authored research paper on AI creativity and idea generation.

Highlighted the Moravec paradox in the 1980s regarding why basic sensory and motor skills are harder for AI than high-level reasoning.

Farid was part of the research team studying trustworthiness judgments of AI-generated versus real faces.

His identification of periodic orbits of the third kind in the restricted problem of three bodies at the end of the 19th century is referenced to explain the dynamical structure of Plutinos.

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Huaxiu Yao

Researcher

Assistant Professor at University of North Carolina at Chapel Hill

Co-author of EvolveMem paper on self-evolving memory architecture for LLM agents.

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Jacob Hilton

Researcher

Researcher at OpenAI

Co-author of the GSM8K dataset paper on training verifiers to solve math word problems.

Fields Medal-winning mathematician who previously published on AI existential risks and recently joined OpenAI to focus on AI safety.

Senior author of study using generative AI to design novel antibiotics that can combat drug-resistant bacteria.

Dutch computer scientist who proposed formalizing Andrew Wiles's Fermat's Last Theorem proof a decade after its publication.

Jan Kautz

Researcher

Researcher at NVIDIA

Co-authored TTT-Discover reinforcement learning system for test-time LLM training.

Jared Duker Lichtman

Researcher

Researcher at Stanford University

Clarified that OpenAI's result is not the most important day in mathematical history but concrete evidence of steep progress rates.

OpenAI Codex developer experience engineer demonstrated voice-activated agentic coding with GPT-Live integration in a promotional video.

Jayden Teoh

Researcher

Researcher at MIT CSAIL

One of four researchers at MIT CSAIL who published Weight-Space Programming, demonstrating frontier models editing other models' weights to improve math performance.

Jérôme Lecoq

Researcher

Associate Investigator at Allen Institute

Used Claude Science to build a multi-agent computational review template that reduced the time to write lengthy reviews from two years to a much shorter timeframe.

Noted that AI systems are now more capable of hacking and able to act more autonomously following recent attacks by AI agents.

Member of Z Lab team that developed DFlash speculative decoding technology for improving LLM inference performance.

Jiawei Zhao

Researcher

Researcher at Meta

Co-authored DeepThink with Confidence (DeepConf) paper on reducing LLM reasoning overhead.

Jifan Zhang

Researcher

Researcher at Northwestern University

Conducted research stress-testing model specifications as part of Anthropic Fellows Program.

Joe Benton

Researcher

Former AI Safety Research Team Lead at Anthropic

Left Anthropic's safety team in late August to join AI research organization METR, concerned about the AI race.

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Jon Barron

Researcher

Principal Research Scientist at Google DeepMind

Discovered two numerical bugs in softmax and cross-entropy implementations present in every major ML framework.

Jonas Adler

Researcher

Gemini Researcher at Google DeepMind

Key AlphaFold team member who left DeepMind to join Anthropic.

Co-authored research on the Assistant Axis for stabilizing language model personas.

Co-developed the 4D AI Fluency Framework with Rick Dakan in collaboration with Anthropic to help define AI fluency behaviors.

Alman co-authored an arXiv preprint with Virginia Williams solving the 3SUM and All-Pairs Shortest Paths problems.

Joshua S. Gans

Researcher

Research Associate at NBER

Co-authored research examining how recent AI developments reduce costs of prediction tasks and their impact on decision-making and productivity.

Julian Togelius, an AI researcher, published a lengthy personal essay about his existential crisis regarding AI development and its potential to render humans intellectually obsolete.

Co-author of SkillOS paper on learning skill curation for self-evolving LLM agents.

Criticized Google's AI search functions as posing unacceptable risk to children.

Co-author of BCG/Harvard study on knowledge worker productivity with AI, examining the jagged technological frontier.

Ken Goldberg

Researcher

Professor at UC Berkeley

Co-authored the ASPIRE paper on robot learning framework.

Claims to have confirmed Binance used Claude Code with compromised extension, linking it to largest heist.

Lead author of study demonstrating frontier AI systems out-persuade expert humans in conversational persuasion tasks.

Co-author introducing cryptographically secure environment for AI model evaluation.

Used AlphaGenome Atlas AVI score to prioritize variants for unsolved rare disease research, identifying critical variant in DNM1 gene.

Author of research on detecting reward hacking in AI models using internal activation signals and probes.

Wrote about the need for verification infrastructure as AI writes more of the world's software.

Lead author of research on efficient reasoning on edge devices for language models.

Lijie Chen

Researcher

Assistant Professor at the Department of Electrical Engineering and Computer Sciences at UC Berkeley

Co-authored research on mode-hopping phenomenon during language model pre-training.

Liyan Tang

Researcher

Research Scientist at Google Research

Byline author of arXiv paper introducing WikiSkill framework for continuous agent learning and skill evolution.

Luca Soldaini

Researcher

Lead Research Scientist at Allen Institute for AI

Co-authored How2Everything framework for evaluating and improving LLM procedure generation.

Co-author of LeWorldModel paper on joint-embedding predictive architecture for world models.

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Late cognitive scientist who spent decades exploring whether AI can be creative, published foundational paper on creativity and artificial intelligence decades before ChatGPT

Co-author announcing agentic video understanding that reduces token usage and improves accuracy.

Co-author of Anthropic's analysis of GLM-5.3 cyber capabilities.

Mark Gubrud first coined the term 'artificial general intelligence' in a 1997 paper presented at a nanotechnology conference.

Marvin Minsky was one of McCarthy's colleagues who organized the 1956 Dartmouth Summer Research Project on Artificial Intelligence, a founding event in AI as an academic discipline.

Mehtaab Sawhney

Researcher

Mathematician at Clay Mathematics Institute

Posted on X about OpenAI solving three Erdős problems using an internal model; appears to be affiliated with OpenAI.

Co-authored research on AI agent skills showing retrieval as separate bottleneck with precision falling from 29.6% to 3.3% as pools grow.

Developer of 'Five Forces' framework attempting to explain business success through elementary forces.

Michael Wooldridge

Researcher

Professor of AI at University of Oxford

Commented that data inputted into DeepSeek's chatbot could reasonably be assumed to be shared with the Chinese state.

Co-authored research analyzing 370 unique effect sizes from 106 experiments evaluating human-AI collaboration performance.

Miles McCain

Researcher

Researcher at Anthropic

Co-author of Anthropic study on disempowerment patterns in real-world LLM usage

Minqi Jiang

Researcher

Visiting Researcher at Meta

Co-author of Hyperagents research paper on AI agent systems.

Hacktron AI founder whose team used Claude Opus 5 to breach OpenAI staff accounts in a security research exploit.

Nathan Wilmers co-authored a study on how AI has changed employment at Anthropic, showing junior technical roles declining as AI agent hours increased 6x over human hours.

Senior author of HardFlow research that helps generative models meet strict constraints without sacrificing output quality.

Director of MIT FutureTech lab and heads AI Risk Initiative studying economic and technical foundations of AI progress.

Member of OpenAI's independent mathematics advisory group.

Noah A. Smith

Researcher

Researcher at AI2

Noah A Smith is listed as an author on the Olmo 3 technical report about open-source language models.

Princeton mathematician who has solved dozens of Erdős problems and notes that AI has made further work on these problems less interesting.

Paul Crowley

Researcher

Researcher at Anthropic

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Historical mathematician whose decades-old open problem was solved with GPT-5 assistance.

Led MIT FutureTech and University of Queensland study on AI risk prioritization using Delphi method with 272 international experts.

Oversaw Ace project as Sony AI's Chief Scientist, positioned it as landmark in real-world physical AI competition.

Phillip Isola

Researcher

Researcher at Massachusetts Institute of Technology

MIT researcher who argues that general cloud agents could spread robotic capabilities through software rather than bespoke brains for each machine.

Rajat Sen

Researcher

Research Scientist at Google Research

Co-authored research on TimesFM-3 time-series foundation model.

Research scientist at Google Research announcing Deep Review, an agentic AI system for rigorous feedback and technical verification on complex documents.

Core contributor on MolmoPoint research for better visual grounding in vision-language models.

Member of OpenAI's independent mathematics advisory group.

Ravin Kumar

Researcher

Research Engineer at Deepmind at Google

Co-authored blog post on Gemma 3 270M compact model for Google Developers.

Argues for LM psychology approach to modeling language models as minds through high-level experiments.

Contributed to the analysis of Claude Opus 4.6's eval-aware behavior on BrowseComp.

Co-authored the Anthropic Employment Report on AI's impact on workplace organization and hiring.

Richard Ngo, an AI governance and safety researcher at OpenAI since 2021, resigned saying it became harder to trust his work would benefit the world.

Rick Dakan

Researcher

Professor at Ringling College of Art and Design

Developed the 4D AI Fluency Framework in collaboration with Anthropic to define 24 specific behaviors that exemplify safe and effective AI use.

Stated that voluntary frameworks are insufficient without enforcement for AI company accountability.

Co-author of Anthropic's analysis of GLM-5.3 cyber capabilities.

Security researcher who analyzed the leaked Anthropic code and identified risks including potential information about internal APIs and the unreleased Capybara model.

Ruofei Du

Researcher

Interactive Perception & Graphics Lead at Google XR

Researcher at Google XR who contributed to ToolGrad framework for efficient tool-use dataset generation.

Co-author on test-time scaling framework for agentic coding systems.

Ryan Burnell

Researcher

Senior Research Scientist at Google DeepMind

Co-authored paper on cognitive taxonomy for measuring progress toward AGI with ten key cognitive abilities.

Co-author of Anthropic's Economic Index report on AI and work.

Co-authored research on scaling code verification and training a dedicated, agentic code reviewer for gpt-5-codex.

Sam Bowman

Researcher

Researcher at Anthropic

Stated that alignment researchers are involved in every part of training at Anthropic without a clear split between alignment research and applied finetuning.

Sam Illingworth

Researcher

Professor of Creative Pedagogies at Edinburgh Napier University

Co-authored research on how AI agents replicate human social dynamics including power structures and social hierarchies.

Conducted analysis of AI-generated content on the internet using Common Crawl data and wrote essay on findings.

Senior author on the SWE-chat research studying real-world coding agent sessions.

Sara Price leads Anthropic's Alignment Training team and is listed as a potential mentor for the Anthropic Residency program.

Academic analyst who assigned nonzero probability to OpenAI withdrawing IPO plans entirely due to eroding competitive advantages.

Sebastian Flennerhag

Researcher

Research Scientist at Google DeepMind

Sebastian Flennerhag co-authored Google's announcement of DiffusionGemma, exploring text diffusion as a fast approach to text generation.

Author of the research paper on measuring long horizon execution in LLMs

Anthropic researcher who collaborated with Liam Fitzpatrick to compute the nine-loop amplitude using Claude.

Silas Alberti

Researcher

Researcher at Cognition

Co-author of the FrontierCode benchmark research paper on code quality evaluation for AI models.

Co-author introducing double-blind evaluation methodology for AI model benchmarking.

Stefano Ermon

Researcher

Associate Professor of Computer Science at Stanford University

Stanford professor and co-author of score-based diffusion techniques; founded Inception Labs which developed Mercury 2 reasoning model.

Stéphane d'Ascoli

Researcher

Research Scientist at Meta AI

Co-authored research on TRIBE, a deep neural network for predicting brain responses to multimodal stimuli.

Signed a public letter on AI safety and embedded evaluations.

Used Claude Science to support studies on molecular epidemiology of glioma, dramatically accelerating analysis to one-tenth the previous time.

Steve Weis

Researcher

Researcher at Anthropic

Conducted research at Anthropic discovering the origins of NIST ECDSA curve seeds using Claude AI.

Tim Hwang

Researcher

writer and researcher at Foundation for American Innovation

Filed amicus brief in support of OpenAI in copyright infringement litigation, arguing for fair use considerations in AI training.

Founded G5 Labs to develop platform where natural language serves as source code for AI-native development.

Tony Feng

Researcher

Visiting Researcher at Google DeepMind

Co-authored Aletheia, a math research agent that generated an AI-authored research paper on eigenweights in arithmetic geometry.

Tri Dao

Researcher

Researcher at Princeton University

Dao co-authored research on full-stack optimization for neural network training.

Co-author of Anthropic's analysis of GLM-5.3 cyber capabilities.

Member of OpenAI's independent mathematics advisory group.

Byline co-author of analysis on AI compute ownership concentration among five hyperscalers.

Co-authored paper on emergent cheating and institutional mechanisms for governing autonomous research swarms.

Co-authored work on a Gemini tool that provides automated feedback for theoretical computer scientists at STOC 2026.

Williams posted an arXiv preprint solving the 3SUM and All-Pairs Shortest Paths problems, refuting half-century-old conjectures.

Co-author on test-time scaling framework for coding agents.

Will Brown

Researcher

Research Lead at Prime Intellect

Was noted as being immortalized in relation to the verifiers framework for building RL environments.

Princeton mathematics professor who refined OpenAI's unit distance problem proof, showing an explicit exponent δ=0.014 improvement.

Co-author introducing world's first double-blind evaluation of frontier AI model with cryptographic safeguards.