Intelligence without representation by Brooks, R. A. (1991)

Artificial Intelligence
(Brooks 1991)


What is intelligence

Intelligence cannot be thought of as a collection of building blocks that may one fall into place to form a coherent whole.

The authors argue for another approach to build artificially intelligent systems:

  • Build the systems incrementally, with complete systems each step of the way to ensure that the pieces and their interfaces are valid.
  • Build intelligent systems at each step of the way that should be let loose in the real world with real sensing and action.

We have been following this approach and have built a series of autonomous mobile robots. We have reached an unexpected conclusion (C) and have a rather radical hypothesis (H).

  • (C) When we examine very simple level intelligence we find that explicit representations and models of the world simply get in the way. It turns out to be better to use the world as its own model.
  • (H) Representation is the wrong unit of abstraction in building the bulkiest parts of intelligent systems.

Creating intelligent systems

The timeline of natural evolution seems to suggest that problem solving, language, expert knowledge, etc. are relatively simple to obtain once the core of the evolutionary machinery gets going and the essence of being in the environment is available. Natural evolution has spent most of its time developing the ability to move, interact and sense its environment, indicating it is probably the hardest part of the problem.

By concentrating on the end goal (human-level intelligence) and its most exceptional features (high-level intelligence) we may miss the most essential parts to building an AI system, which are also the hardest to develop. Moreover, we must forget about this single example of intelligence. This is to adopt a more bottom-up approach to recreating the processes that have led to it rather than replicating it.

This quote resonates with the open-endedness assumption that it is necessary to have no objective to reach hard goals.

We do claim however, that there need be no explicit representation of either the world or the intentions of the system to generate intelligent behaviors for a Creature. Without such explicit representations, and when viewed locally, the interactions may indeed seem chaotic and without purpose.


This is an interesting paper which I believe is still relevant today. Brooks' conception of intelligence is in my opinion very interesting. He applies it to robots because this is probably the only approach if you consider this path to AI.

However, it might be that the complexity needed to have the above interactions happen doesn’t have to be exogenous. A complex system, generating complex behavior on its own could be the source of chaotic interactions our AI system is based on (and could even be a part of)


Brooks, Rodney A. 1991. “Intelligence Without Representation.” In Proceedings of the International Joint Conference on Artificial Intelligence, 1:12.

← Back to Notes