Intelligence vs Cognition
The definition of intelligence is very vague and controversial. Often it’s defined in the context of human or at least animal intelligence through the list of properties, something like “ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience”. Such attempts fail quickly when you try to find the lower border of intelligence. How about plant intelligence, or virus intelligence, or even swarm intelligence?
As a result of it, there are attempts to define intelligence more generally through its adaptive properties, something like “Goal-directed adaptive behavior” (Sternberg, Salter 1982) or “an agent’s ability to achieve goals in a wide range of environments” (Legg, Hutter 2007). However, it doesn’t solve the problem with defining the lower border of intelligence is not solved, it only shifts it further. A thermostat is definitely intelligent. As well as many other devices with a feedback loop. And what about just any object, which, as we know since Newton’s time, has adaptive behavior (with some space for discussion of the goal part)? Also, as it was already discussed in Don’t mix Agency with Intelligence, adaptivity is only indirectly related to intelligence.
At the same time, we have a much more suitable term, which is defined specifically for mental processes: cognition. Therefore the correspondent domain of science has name cognitive science. Viruses and thermostats don’t have cognition, plants and swarms arguably either. That’s good, but what is cognition beside the formal definitions as “the mental action or process of acquiring knowledge and understanding through thought, experience, and the senses” (Oxford dictionary) or using lists again like “intellectual functions and processes such as attention, the formation of knowledge, memory and working memory, judgment and evaluation, reasoning and computation, problem solving and decision making, comprehension and production of language” (Wikipedia)?
Perhaps the most general way to do it would be something like this: cognition is the ability to represent input, and correspondingly, available through the input world, in a way that allows useful manipulation with it. This ability leads to the spontaneous processing of any given input and building a model of the world as a result.
The internal dynamic of such processing is defined by the presence of input’s structure which can be processed. While there is something to process, it’s going. When it converged to something stable the process is completed until the change of the input. The inference is not just a set of classes, but a holistic representation of the input based on the current model of the world, which can be updated with every new inference.
In real life, there are quick and long parts of processing because of the necessity to support basic function with the given speed and limited resources, but it’s rather implementation details.
So, instead of Synthetic Intelligence, the term Synthetic Cognition would be much more accurate, and occasionally you can encounter this term, but SI as a term has its own history, and it has to be appreciated.
As a final note, the term Artificial Intelligence is completely viable, as it also has unclear borders, and the domain is much broader than SI, and includes, for example, hard-coded rules-based solutions.
