The word “intelligent” can mean many different things to different people in different contexts. I will describe some of these meanings, assign less ambiguous terms to them, and show some contexts which they make sense. Hopefully, by the end of each section, you will be able to pick out other contexts where people are talking about this particular kind of intelligence.
Say you are learning about how to solve a difficult math problem. You think, and think, and can’t seem to come up with anything. Finally, you ask someone what the answer is, and they give you the answer. A common reaction is to say: “wow, that is a really brilliant solution”. You don’t know who came up with it or how, you don’t know if a machine or an alien generated the answer, but you still think it is brilliant, because it achieved the goal better than your though patterns when you tried to solve it.
Commonly people call old style AI apparently intelligent. We might say that Siri is smart, because it can make jokes, map a route between our apartment and the grocery store, and many other useful tasks we would have to think about carefully. Again, this type of intelligence is seen through our own lens, with our own goals and thoughts in mind. There is an inherent anthropomorphizing that usually occurs when people think about this kind of intelligence (although it is not strictly necessary, only the goal centric idea is necessary).
Perhaps a better phrase for this view of intelligence is “objective intelligence”, since it is dependent only in the context of the goal you are trying to achieve, and is independent from the mechanics of the system which produces the solution.
This type of intelligence is interesting to me because it allows us to talk about the intelligence of natural selection. While natural selection has no agent, and no consciousness, it does result in organisms getting better at reproducing in their environment. We can reason about and appreciate these adaptations, and think about how we would engineer organisms to survive better in this environment, and be pleasantly surprised, and in awe of the intelligence of natural selection when it results in better features than our engineered organism.
There are many other dynamic systems I want to talk about as intelligent, like cultures and political systems. This way of reasoning allows to reason about in what ways, and to what degree these systems are better or worse than other systems (such as our own intelligence), without falsely attributing an agent that is doing the thinking. Calling this intelligence, rather than some other term, is important to my overall purpose because I believe that much of our own intelligence, which we attribute to ourselves, is rather the product of other dynamic systems. I believe that on our own, we would be much stupider by our own judgement. I believe that machine intelligences will also benefit from being part of these non-agent intelligences. This has serious implications for what AI will end up looking like. Perhaps we will end up seeing a diversity of machine intelligences that talk to each other, rather than a singularity.
More philosophically minded people should be interested in this idea because it allows you to talk about the conscious brain, the thing doing the choosing and picking, as intelligent, even though the part of the brain it occupies is small, and its computational power is very poor when isolated from subconscious part of the brain. The idea would be that it brings intelligent results that would not happen without it, even it has to rely on its connections to unconscious elements of the brain. Since the results are intelligent, then it is intelligent. By themselves neither the agent nor their tools are intelligent, only the overall system is.
Think back to the first example that introduces apparent intelligence. You think and think, and can’t come up with an answer, but then the answer is offered, and it is intelligent in a way. Now, say that someone explains an intuitive and simple method of solving the problem. The natural reaction to this explanation is to think that the solution is not brilliant. Instead, perhaps you think that the method to find the solution is brilliant, or that you were just being stupid in not finding this method.
Reduced intelligence is like the solution in this example. With no context, it appears intelligent, but when the method is described, and you can execute that method yourself, it no longer seems intelligent at all, but rather mundane and mechanical. So its intelligence is reduced by a more general algorithm.
To me, this reaction is a deeply pragmatic and natural reaction, rather than illogical revisionism. An intelligence reduction means we can do the work of the intelligent act ourselves. Since we can do it ourselves, it is simply less impressive, less valuable, and we have little reason to acknowledge intelligence.
One great example of reduced intelligence is Siri. When programmers look at Siri, they are unimpressed. They think that Siri is stupid. And often, non-technical people also start thinking that Siri is stupid after learning more about it. The reason seems to be that people start understanding the limits of Siri’s knowledge and understanding, and realize that their own knowledge and understanding far out strip it. They start realizing that they could easily do the work of Siri themselves, if they just had access to a search engine and a keyboard.
Why does our mind change? Why do we feel this way in so many contexts?
Now, what if you kept on reducing intelligent behavior until it could be reduced no more? Why would am I even asking this question? Well, perhaps this corresponds to another objective type of intelligence. Perhaps a true measure of intelligence is how
What would this even look like? Well, in reality, it is impossible in generality, a result from computability theory.