Problems With a Machine View of Intelligence

This article critiques a recent TED Talk by Alex Wissner-Gross entitled, “A New Equation for Intelligence.”

Here is the talk.

The talk is on a ridiculous equation which is not an equation in the traditional sense of the term. You cannot do algebra on this equation without ruining it. The equation is: F = T ∇ Sτ, which is prima facie meaningless because the del symbol is ambiguous. This immediately tells us that the equation was written poorly, and the talk follows the pattern.

Alex’s thesis is that intelligence is something entities do. I’ll give him half of a point for that. At least he delineates intelligence from entity, but intelligence is usually thought of as something entities have, not something they do. Yet he wants to develop a purely scientific definition of intelligence, so he basis it extrospectively; on the basis of things entities do. OK, you get half a point. Continue.

The thing which entities do which we refer to as intelligence, Alex asserts, is to maximize their range of potential actions. At first this sounds stupid. Then on consideration it sounds a little better. Then on a full thought it is absolutely stupid after all.

Prima facie, maximizing the range of potential actions doesn’t sound like intelligence because intelligence is usually thought of helping us deal in the realm of thinking and information. It improves our ability to ascertain, retain, process, recall, and in some cases act upon information.

Maximizing a range of potential actions seems to deal in the realm of the physical. Wouldn’t becoming stronger or more athletic help us maximize our range of potential actions? Yes, it would. So the element labeled “intelligence” by Alex is really something more like “ability.”

What about if we hold physical ability constant? Even so, this is not intelligence. In the words of G. K. Chesterton, “Merely having an open mind is nothing; the object of opening a mind, as of opening the mouth, is to shut it again on something solid.” It is not intelligent to keep all possibilities open, maximizing the range of potential actions, because this would imply never actually choosing one.In the next place, intelligence is teleological. Thinking for the sake of thinking is not intelligent. Thinking on purpose, to accomplish some goal, is intelligent. Accomplishing a mental task more efficiently is more intelligent and less is less so.

There has to be some win rule for anything to be intelligent. Some purpose, some goal to accomplish. Intelligence is also usually considered a good, at least in the sense of efficiency if not also the moral sense. It could be argued that evil people can be intelligent as well, so let’s stick to the efficiency sense. I would call moral intelligence something more like wisdom.

Keeping all options open is not a good, even in an efficiency sense, and it is not a win rule. If it is a win rule, the entity is not discovering the win rule. It is following the orders of being given the win rule, which is not a complete picture of intelligence. It is a robotic process like we would expect of a computer. Computation efficiency is perhaps a kind of intelligence, but certainly not the full thing.

Other points:

  1. This definition exempts information from explanation, though it also hypocritically grounds the explanation in information theory.
  2. This definition does not account for identity, which is associated with intelligence. Minds are intelligent and possess identity.
  3. This definition fails to recognize heterogeneity of intelligence. Most people who study intelligence find that there are multiple kinds.
  •  
  •  
  •  
  •  

Leave a Comment