About me

Hi! I’m Ben, a final-year PhD candidate in Computer Science at Brown University advised by George Konidaris and an incoming visiting researcher at MIT’s Comptuational Cognitive Science lab. My research focuses on how language and, more generally, sign systems emerge and acquire meaning. My early work grounded language by mapping it to formal structures in decision-making (RLang); my current work argues that meaning should not be stipulated at all – that we should instead create the conditions for it to emerge on its own. I draw inspiration from linguistics, cognitive science, philosophy of mind/language, and semiotics.

I’m a recipient of the NSF Graduate Research Fellowship and Brown’s Morgan Edwards Fellowship. I was the lead organizer for the Brown Robotics Talk Series, and a founding member of the Brown AI Safety Team.

Research Philosophy

I’ve published a full statement of my research agenda on my Substack, but here’s a short version:

Language was invented by humans to reason about and describe their subjective experience of the world. To ground language, we must render it in terms of perception and action – the machinery of decision-making. My entire research agenda is about how to achieve this rendering. My current work uses multi-agent RL and Bayesian inference to study the conditions under which artificial agents invent human-like languages about their environment, with the long-term goal of simulating the core language acquisition, invention, and understanding faculties that humans have, and eventually applying them to building grounded artificial languages that rival the expressivity of natural languages.

I keep a running record of my academic reading: papers here, books here.

Highlights

• Visual Theory of Mind Enables the Invention of Proto-Writing (CogSci 2025, oral presentation) – a computational model of pictographic signification that results in artificial sign systems resembling ideographic writing systems. Covered by Perplexity and mentioned in WIRED; here’s a short summary on Substack.

  Why did only humans invent graphical systems like writing? 🧠✍️
  
  In our new paper at @cogsci_soc, we explore how agents learn to communicate using a model of pictographic signification similar to human proto-writing. 🧵👇 pic.twitter.com/3NtveUk4hu

  — Benjamin Spiegel (@superspeeg) April 22, 2025

• AGI Is Not Multimodal (The Gradient, 2025) – a position piece arguing against the modality-centric view of general intelligence. Discussed on Hacker News and featured in TLDR AI.

News

• June 2026: I’m living in San Francisco for the summer as an ML Research Intern on the Behavior and Planning Team at Nuro — let’s grab coffee!
• March 2026: I successfully delivered my Thesis Proposal to my committee of George Konidaris, Ellie Pavlick, and Michael Littman 🎉
• AY ‘25-‘26: I’ve been on a talk circuit giving “A Path to Language Understanding: Grounding Language to Markov Decision Processes” – at MIT (CoCoSci, host: Josh Tenenbaum; CompPsyLing, host: Roger Levy), Stanford (Cognitive Tools Lab, host: Judy Fan), UC Berkeley (CHAI, host: Cam Allen), U Edinburgh (Centre for Language Evolution, host: Kenny Smith; School of Informatics, host: Dave Abel), and UMass Amherst (Autonomous Learning Lab, host: Philip Thomas).
• July 2025: I gave my first oral presentation at CogSci 2025.
• June 2025: “AGI Is Not Multimodal” was published at The Gradient.
• April 2025: Visual Theory of Mind Enables the Invention of Proto-Writing was accepted for oral presentation at CogSci 2025.

Personal Interests

I read widely [bookshelf], watch and log films [Letterboxd], play piano and produce music, solve Rubik’s Cubes (3x3 PB: 9.58s, best Ao5: ~12s). I founded and facilitated a Junto, a bi-weekly Salon-style discussion group for grad students at Brown. I also practice mindfulness meditation, enjoy the Making Sense podcast, and was briefly a frequenter of LessWrong.