RFP Overview    |    FAQ    |    Examples    |    Application Form    |    Timeline

Bookmark and Share

Intelligence in the Physical World
An International Request for Proposals

Theme and Example Questions

This appendix gives background information about the program theme and a grid of example questions, any of which could form the basis of a suitable proposal.

The theme for this RFP is Intelligence in the Physical World. It is intended as a companion to our previous program on Agency in the Physical World. The program will support investigations into the many connections between basic physical laws and the capabilities of intelligent agents.

Agents in the physical sense are systems capable of causing physical changes aligned with goals, reasons, drives, or needs. A basic question is, how do agents do what they do? We identify seven capabilities linked to intelligence: (1) Remembering by acquiring and storing information in internal states; (2) Computing to process the remembered information (3) Learning patterns and correlations in and between the external and internal world; (4) Predicting the evolution of systems under different assumptions; (5) Optimizing by comparing potential actions against goals, reasons, drives, or needs; (6) Acting upon the world so as to increase the probability of desired results, and (7) Experiencing subjectively some or all of the preceding processes.

Of each capability, we propose three questions. How are the world and its laws constituted so as to make it possible? How are agents able to create consonance with the laws and structure of physics so as to make it effective? And, what are the ultimate limits of it as dictated by fundamental physics?

The table below gives examples for each combination of capability and question.

Capability What makes it possible? What makes it effective? What are the limits?
Remembering Why does our universe admit a local description including relatively isolated stable subsystems? Could it have been otherwise and if so, which physical laws make the difference? How isolated can a system be while remaining useful as a memory device? What are efficient error-correcting methods in a noisy environment? How much information can possibly be stored in a given spatial or spacetime region? What is the thermodynamic cost of information acquisition or bit erasure?
Computing What is the smallest change to our physical laws that would destroy their ability to support universal computation? If the physical world allows classical and quantum computation, are there other interesting types that are allowed? Do any “natural” quantum computers exist? What is the thermodynamic cost of computing? How fully can these be circumvented using reversible and quantum computation?
Learning What properties of our universe enable small subsystems of it to effectively model the world around them? Are learning systems maximally effective “on the edge of chaos”? Is this generally true of intelligent systems? Is there a thermodynamic cost to learning? How close are living systems to paying just this price?
Predicting Why are there stable mathematical physical laws? Why can they be constituted as initial-value problems? How and why can we often ignore fine- grained details? What systems in general admit coarse-grained and collective variables that can be effectively predicted? What is the thermodynamic cost of prediction? Are there fundamental limits to agent effectiveness based on fundamental indeterminism in physics?
Optimizing How can intelligent systems infer “rules” governing a system they interact with that are different from its fundamental rules yet useful? What shortcuts do humans or other biologically-evolved systems use to come to quick “intuitive” answers to problems with a huge search space? What problems of practical importance cannot be short-cut? Does physical law prevent efficient solution to NP- hard problems? Why?
Acting If physics is deterministic, how can meaningful action/decisions occur? What accounts for the role of “free choice” in quantum theory and foundations? How can tiny causes (e.g. in brain cells) have huge “desired” effects without being drowned in chaotic amplification of many other small effects? How much control can an agent in principle have to create desired effects without “side effects?” What are the ultimate bounds on what can be accomplished by an intelligent agent limited only by the laws of physics?
Experiencing Do the laws of physics guarantee that any system exhibiting general intelligence is also conscious in the sense of having subjective experience? Can subjective awareness be measured or probed via information-theoretic tools? Can consciousness theories be experimentally tested, and if so, can we rule out any currently popular theories? Is consciousness necessarily unified or could it “pay attention” to many things at once? Is the amount of information apprehended by consciousness small by necessity, or unbounded, or something in-between?
Please enter your e-mail address:

Note: Joining the FQXi mailing list does not give you a login account or constitute membership in the organization.