Graduate Student Room 32-392 Computer Science and Artificial Intelligence Laboratory (CSAIL) 32 Vassar Street Massachusetts Institute of Technology Cambridge, MA, USA 02139

Phone: (617) 253-8581 (arrange in advance) E-mail: jakebeal@mit.edu CV (last updated 3/25/2008)

I am a postdoctoral associate at MIT CSAIL (formerly the AI lab).

I work with Project MAC and the Genesis Group. My main research interests are building human-like intelligence via collaborative communication and amorphous computing---the control of extremely large and unreliable spatially distributed systems. Recently, I have come to realize that both these pursuits are part of a larger theme, which I am currently calling "engineered emergence." I also dabble in a number of other things on the side that attract my interest or appear as obstacles in my path.

I'm married to Abi Harper, who is a massage therapist. Other than that, I don't plan to post my non-academic life on this web page.

Research Publications & Talks

Publications

• Analyzing Composability in a Sparse Encoding Model of Memorization and Association, Jacob Beal and Thomas F. Knight, Jr, IEEE 7th International Conference on Development and Learning (ICDL 2008), August 2008.
• Learning Composable Signals for a Cognitive Substrate, Jacob Beal, Cognitive Science Conference, July 2008.
• Shared Focus of Attention for Heterogeneous Agents, Jacob Beal, Short Paper, 7th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2008), May 2008.
• Learning by Learning to Communicate, Jacob Beal, PhD Thesis, August 2007.
  Also listed as Tech Report MIT-CSAIL-TR-2007-042. If you have difficulty with the PDF, try it in postscript
• Developmental Cost for Models of Intelligence, Jacob Beal, AAAI 2007 Workshop on Evaluating Architectures for Intelligence, July 2007.
  Also listed as Tech Report MIT-CSAIL-TR-2007-026
• CogSci to AI: It's the Brainware, Stupid!, Jacob Beal and Gerald Jay Sussman, AAAI 2006 Spring Symposium "Between a Rock and a Hard Place: Cognitive Science Principles Meet AI-Hard Problems", Stanford, March 2006.
• Leveraging Language into Learning, PhD Proposal, February 2004.
  Previously partially developed in AI Memo 2003-007
• An Algorithm for Bootstrapping Communications, Jacob Beal, International Conference on Complex Systems (ICCS), June 2002.
  A longer version appears as AI Memo 2001-016, August 2001.
• Generating Communications Systems Through Shared Context, Jacob Beal, Master's Thesis, AI Tech Report 2002-002, January 2002.

Talks

• Learning by Learning to Communicate, PhD Thesis Defense, MIT CSAIL, August, 2007. Material after page 43 is supplementary slides that were not used in the talk, but could be brought up to help respond to questions.
• Developmentally Inspired Cognitive Architectures, presented at AAAI 2007 Workshop on Evaluating Architectures for Intelligence, July 2007.
• CogSci to AI: It's the Brainware, Stupid!, presented at AAAI 2006 Spring Symposium "Between a Rock and a Hard Place: Cognitive Science Principles Meet AI-Hard Problems", Stanford, March 2006.
• Integration by Coincidence: Status and Speculation, presented at MIT Biologically Inspired Cognitive Architectures (BICA) workshop, January 2006.
• Implementing Valiant's Neuro-Logic, internal presentation, December 2005.
• Leveraging Language into Learning, presented at AAAI 2005 Doctoral Consortium, Pittsburgh, July 2005.
• Generating Communication Systems Through Shared Context, RQE presentation, Spring 2004.

Publications

• Autonomy in Spatial Computing, Jonathan Bachrach and Jacob Beal, Third Workshop on Hot Topics in Autonomic Computing, June 2008.
• Fast Self-Healing Gradients, Jacob Beal, Jonathan Bachrach, Dan Vickery, and Mark Tobenkin, ACM Symposium on Applied Computing 2008, March 2008.
  Received Best Paper Award (Artificial Intelligence & Agents Theme); Also available as MIT CSAIL Tech Report 2007-050
• Constraint and Restoring Force, Jacob Beal, Jonathan Bachrach, and Mark Tobenkin, MIT CSAIL Tech Report 2007-044, August 2007.
• Continuous Space-Time Semantics Allow Adaptive Program Execution, Jonathan Bachrach, Jacob Beal, and Takeshi Fujiwara. IEEE SASO 2007, July 2007.
  Also available as MIT CSAIL Tech Report 2007-038
• Amorphous Computing, Hal Abelson, Jacob Beal, and Gerald Jay Sussman, draft article for "Encyclopedia of Complexity and System Science" to be published by Springer-Verlag.
  MIT CSAIL Tech Report 2007-030, June 2007
• Building Spatial Computers, Jonathan Bachrach and Jacob Beal, MIT CSAIL Tech Report 2007-017, March 2007.
• Programming a Sensor Network as an Amorphous Medium, Jonathan Bachrach and Jacob Beal, extended abstract for poster at DCOSS 2006, June 2006.
  Our poster at DCOSS.
• Infrastructure for Engineered Emergence on Sensor/Actuator Networks, Jacob Beal and Jonathan Bachrach, IEEE Intelligent Systems, (Vol. 21, No. 2) pp. 10-19, March/April 2006.
  Preprint version of the article and sidebar.
• Amorphous Medium Language, Jacob Beal, Large-Scale Multi-Agent Systems Workshop at AAMAS 2005, July 2005.
• Biologically-Inspired Robust Spatial Programming, Jacob Beal and Gerald Sussman, MIT AI Memo 2005-001, January 2005.
• Programming an Amorphous Computational Medium, Jacob Beal, in Unconventional Programming Paradigms International Workshop, September 2004. Updated version in LNCS Vol. 3566, August 2005.
• RamboNodes for the Metropolitan Ad Hoc Network, Jacob Beal and Seth Gilbert, MIT AI Memo 2003-027.
  A tightened version was published in Workshop on Dependability in Wireless Ad Hoc Networks and Sensor Networks, part of the International Conference on Dependable Systems and Networks, June 2004.
• Near-Optimal Distributed Failure Circumscription, Jacob Beal, AI Memo 2003-017, August 2003.
• A Robust Amorphous Hierarchy from Persistent Nodes, Jacob Beal, AI Memo 2003-012, May 2003.
• Persistent Nodes for Reliable Memory in Geographically Local Networks, Jacob Beal, AI Memo 2003-011, April 2003.
• Leaderless Distributed Hierarchy Formation, Jacob Beal, AI Memo 2002-021, December 2002.
  This didn't work as well as I wanted, and led me to start working with Persistent Nodes for hierarchy formation instead.
• Amorphous Infrastructure for Language Implementation, Ryan Newton and Jacob Beal, MIT CSAIL Tech Report 2006-015, December 2002.
  Written as a project in an Amorphous Computing seminar, this work was the origin of the Persistent Node algorithm. We made an official recording of it in 2006 to make referencing easier.

Talks The reader is advised that most spatial computing talks also involve a live demo that is usually at least indicated in the PDF slides.

• Spatial Computing and the Challenge of Engineered Emergence, presented at Harvard CRCS Privacy and Security Lunch Seminar, April 2008.
  A shorter, related talk, Autonomy in Spatial Computing was presented at the Third Workshop on Hot Topics in Autonomic Computing, June 2008.
• Fast Self-Healing Gradients, presented at ACM SAC 2008, March 2008.
  Associated paper won the Best Paper Award (Artificial Intelligence & Agents Theme).
• Programming Cell Aggregates, presented to the MIT Synthetic Biology Lunch, January 2008.
• Continuous Time Programming, presented at IEEE SASO 2007, July 2007.
• A Brief History of Amorphous Computing, internal presentation, April 2007.
  An informal talk explaining how my work fits into the larger context of amorphous computing. Take the categories with a grain of salt.
• Programming Spatial Computers, internal presentation, September 2006.
  A 5-minute summary of what the Amorphous Medium approach is all about, for new students at CSAIL.
• Programming Manifolds, presented at Dagstuhl Seminar 06361, "Computing Media and Languages for Space-Oriented Computation," Germany, September 2006.
  An expansion and update on the continuous semantics of Proto.
• Continuous Semantics of Proto, internal presentation, January 2006.
• Amorphous Medium Language, presented at the Large-Scale Multi-Agent Systems (LSMAS) Workshop in AAMAS-05, Netherlands, July 2005.
  Contains a good overview of AML abstraction layers, execution model, and language concepts.
• Amorphous Computing's Programming Languages, internal presentation, April 2005.
• Programming an Amorphous Medium, invited talk at Unconventional Programming Paradigms workshop, France, September 2004.
• Near-Optimal Distributed Failure Circumscription, presented at PDCS 2003, Los Angeles, November 2003.

Publications

• Sidestepping Impossibility: Combat Consensus in the Assassins' Guild., Jacob Beal, MIT CSAIL Student Workshop 2006, September 2006.
  Followup to the first Assassins' Guild paper, this one examines carefully how combat consensus gets around an impossibility result.
• What the Assassins' Guild Taught Me About Distributed Computing, Jacob Beal, International Conference on Complex Systems (ICCS) 2006, June 2006. (Also MIT CSAIL Technical Report MIT-CSAIL-TR-2006-038)
  The deeper I look at how live action roleplaying runs distributed simulations, the more surprises I find. I think there's something here for computer scientists to pay attention to.

Talks

• Principles for Engineered Emergence, Jacob Beal, Unconventional Computation: Quo Vadis?, March 2007. (Published as a CSAIL Work Product).
  The first talk explicitly about engineered emergence. A shorter but more recent version was presented at ICCS in November, 2007
• Sidestepping Impossibility: Combat Consensus in the Assassins' Guild., Jacob Beal, MIT CSAIL Student Workshop 2006, September 2006.
  See note above on the associated publication.
• What the Assassins' Guild Taught Me About Distributed Computing, Jacob Beal, International Conference on Complex Systems (ICCS) 2006, June 2006.
  See note above on the associated publication.

Publications

• Cognitive Security for Personal Devices, Rachel Greenstadt and Jacob Beal, MIT CSAIL Tech Report 2008-016, March 2008.
  This is a position paper on attaining security by having a device imprint on a broad-spectrum collection of passively available biometrics.
• Learning From Snapshot Examples, Jacob Beal, MIT AI Memo 2005-012, April 2005.
  The ideas in this memo need more contextualizion w.r.t. other machine learning work. As it stands, it documents an abstraction barrier and learning heuristic I needed and was unable to find elsewhere. Obsoleted by the IIES mechanism in my PhD thesis.
• Shrinking the Leap of Faith, Jacob Beal and Tim Shepard, unpublished draft, March 2005.
  This sketches some ideas about exploiting the locality of wireless connections to make it easier to establish trust with distant servers on the Internet.
• Deamplification of DoS Attacks via Puzzles, Jacob Beal and Tim Shepard, unpublished draft, October 2004.
  Tim was trying to figure out whether the puzzles in the IETF's proposed Host Identity Protocol were actually a useful defense against an attacker with a large network of hijacked computers. We found out that it is.
• Predictive Modelling for Fisheries Management in the Colombian Amazon, Jacob Beal and Sara Bennett, International Conference on Complex Systems (ICCS) 2004, June 2004.
  This was an application of modelling to a real-world problem that has petered out because the project it was to support lost funding.
• The Development of a Small Autonomous Helicopter Robot for Search and Rescue in Hostile Environments, Jacob Beal, Carl Blaurock, Keith Bonawitz, Kyrilian Dyer, Paul Elliott, Paul Eremenko, Eric Feron, Emilio Frazzoli, Benjamin Ingram, Michael Lester, Manway Liu, Stefan Marti, Joshua Napoli, Kailas Narendran, and Scott Rasmussen, Proceedings of the AUVSI Annual Symposium, July, 1999.
  When I was an undergrad, I was on the MIT Aerial Robotics Team, whose purpose was to compete in an unmanned aerial vehicles contest. I recently discovered online the paper we sent one year to the conference associated with the contest. I have no idea which portions I wrote or who even went to the conference.

Talks

• Learning from Snapshot Examples, presented in CSAIL Student Seminar, MIT, April 2005.
  See note above on the associated publication.
• Analyzing Failures as Noise Presented with Seth Gilbert at LIDS Student Conference 2004, MIT, January 2004.
  I don't know any technique for analyzing the failure response of a network where nodes are continually being added and removed. Seth & I kicked around ideas for a while and brought them to a student conference looking for others who might have ideas. We still haven't figured anything out.

• A Distributed Building Evacuation System, Dany M. Qumsiyeh, M.Eng. Thesis, June 2008.

These are written informally, impolitely, and imprecisely. Don't take them seriously unless they happen to inspire you...

• Human-Like vs. Complementary Intelligence
  I sometimes think that AI is incoherent because we mean different things when we say "intelligence". This rant tries to pry the definitions apart and figure out what human-like intelligence might mean. (October, 2004)
• How do we know when we're winning?
  Thoughts on AI research methodology presented during the Genesis Group's internal workshop in 2003.