Skip to: Site menu | Main content

Prediction
The purpose of Intelligence is Prediction. Evolution of the ability to predict agents and phenomena in the environment improved survival rates and created a strong evolutionary pressure to develop better and longer term predictions. This is the reason Intelligence evolved.

On the Origin of Intelligence




These paragraphs are discussing a hypothesis of how Intelligence may have evolved but you should know that my knowledge of Evolutionary Biology is limited.

The earliest brains might have been nothing more than simple nerve clusters that coordinated the movement of legs while walking in multi-legged agents, such as arthropods. Walkers keep some of their legs on the ground at all times which means you only lift one or a few of them at any one time. You will also want to avoid stepping on your own feet. Such coordination requires control from a central point.

When using a "running gait" all legs will be off the ground at the same time for some part of the running cycle. When doing that you can no longer rely on feedback from a leg to tell you that the leg is well positioned on the ground, that the ground under the leg will carry its share of the body mass, and that it won't slip. You need to be able to predict the spot where the leg will land, and the impact time (with something like millisecond precision) so that you can start preparing the appropriate muscles for the landing impact and for the next step. This kind of prediction would also improve your speed in a walking gait, which means there are continuous rewards for small improvements. This means development of this ability through Evolution is Biologically Plausible.

In order for a feature to be Biologically Plausible there must exist a way that feature could evolve. This requires a starting configuration (such as the existence of walking animals) and a gradient of advantage that provides an evolutionary pressure to develop the feature. A continuous gradient where small changes yield small improvements is much preferred over steeper saltations where a major change provides a major improvement with nothing in between.

Vision evolved before running. Running requires vision so that you can run without hitting obstacles. Vision helps you build a model of your environment, often referred to as a "World Model". If you can remember features of the world from one occasion to the next you can remember safe and dangerous places to go, safe and treacherous spots to plant your feet, etc. In essence, you are capable of predicting your environment in order to move faster.

There is a strong evolutionary pressure to get better and better at predicting your environment. You would benefit from developing models of other agents so that you can predict how predators or prey will act, and you would want to predict other members of your tribe such as potential mates and rivals. The single skill of prediction, even if it often fails, yields a big advantage in how well you survive and how likely you are to breed. The better you can predict the near future compared to your immediate rivals, the more offspring you will have.



Jeff Hawkins also believes Intelligence is defined by Prediction. You may enjoy this excellent video. A small detail to note is that that he seems to indicate only higher levels of Intelligence use Prediction but I believe Prediction is a fundamental low level operation for ALL Intelligence.

To summarize, the purpose of Intelligence is Prediction, and the above argument shows that the capability to make predictions is a Biologically Plausible feature.

Cascading and nesting predictions

Evolution rarely throws anything away. I believe Prediction, not Logic, is still the most important low level primitive in human brains.

William Calvin has discussed in several books how prediction of limb movement might be a reason we evolved brains large enough to handle language. I believe we evolved discovery of Semantics as a way to widen and generalize the patterns for events that trigger predictions, i.e. as a way to generalize predictions without having to create Logic-based models; such a development would predate the development of language, would be a reason to enlarge the brain, and would provide the capability to handle semantics that would later enable the development of language.

Friedrich Hayek observed that when sensory input events reach the brain they have all been converted to the same kind of nerve signals that the brain uses everywhere else. So there is no real difference between sensory input and other processing in the brain. Predictions in the brain are events like any other.

It is easy to imagine a layered brain architecture where sensory input arrives "at the bottom layer". Predictions of these inputs, or lower level predictions in general, could be predicted by higher level predictions in several layers. I believe that this kind of nesting of predictions is the origin of higher level semantics, which would allow an agent to generalize the event that triggers the prediction. The development of Semantics as a nesting of predictions is Biologically Plausible.

Longer term predictions are possible by cascading predictions. Interaction with higher-level predictions that maintain the context (the semantics) could be used to limit the combinatorial explosion that results from cascading.