# If intelligence maximizes future freedom of action, then are intelligent agents free?

Alex Wissner posits that there is an equation for intelligence:
`F = T ∇ Sτ`
or:

intelligence is a force, F, that acts so as to maximize future freedom of action. It acts to maximize future freedom of action, or keep options open, with some strength T, with the diversity of possible accessible futures, S, up to some future time horizon, tau. In short, intelligence doesn't like to get trapped.

http://www.ted.com/talks/alex_wissner_gross_a_new_equation_for_intelligence

It seems there is a contradiction here:

If the above is true, then intelligent agents maximizing future freedom of action are at the same time perpetually forced to make the same choice. That is, ordering their current choices by the number of future freedoms of action, they must always choose the one with the maximum value. There is no freedom in being forced to make that choice.

So the question is, can we be free and intelligent at the same time?

First, statistical equations just don't work that way. There is no most-random arrangement of the molecules of a gas, otherwise, that optimum would be observed and become a reliable structure. Then we would see it as expected, and not random at all. Maximizing stochastic entropy is never a deterministic process.

Likewise, there is no 'most free future', there are many futures with largely equivalent freedom, so it is possible to maintain paths to an array of maximal points, and virtually impossible to establish a single maximum at any time. Choosing cannot be done by ordering options, there are infinitely many of them, nor does it require the same decision as any other holder of the same options at a given time, as a huge swath of values will be equivalent within the knowable tolerance.

So yeah, someone trying to do this is still free.

Still, this would not be my definition of intelligence, but of prudence. Briggs is not an idiot, and the continuum between convergence (J-ity) and divergence (P-ity) in intellectual preferences is observed and realistic. Deciding one of the two is 'really intelligence' is just silly.

If we just pursue the optimal freedom in the future, we may well all be dead before we start enjoying any actual freedom in the present. I don't want really happy grandchildren at the cost of universal depression right now!

Not that we could not use a lot more 'prudence', by this definition than we as a culture are showing currently...

Notes:

1. yes, we are avoiding '-ness' on purpose...
2. The J and P in Myers-Briggs lingo are Judging and Perceiving, but in a more traditional sense they are Justice and Prudence (crossed by Courage and Temperance), thus the choice of the word.

For me, the term free is not some absolute on/off value, it is not either you have it or you don't. Freedom is about the number of choices you have at a point in time, you may have few choices so less freedom or more choices hence more freedom comparatively.

Now let's say you have 3 choices at a point in time and you have choose one of them to move forward, then what should be procedure to determine what choice to make? There could be many ways to do that, for example: A lazy person may choose the one which would need less effort to carry out, a greedy person may choose the one which has most benefits in terms monitory gains and so on. Once you have made a choice and move forward you are presented again with a set of choices depending on what you choose at earlier point. This new set of choices may be large or small as compared to the previous one. The equation in the question says that the procedure that so called intelligent agent will use to decide the choice will be based on which choice right now will allow me with larger number of choices in future set of choices. Obviously the future should be some sensible threshold otherwise the agent will never be able to make the current choice.

So, yes, you can be free and intelligent at the same time.

You are assuming that there are unique options maximizing "freedom of action". If "freedom of action" is measured by the number of available choices at some future time, as Wissner has it, there may be multiple options with the same number of future choices. Intelligence then reduces the options, but not necessarily down to a single one, so it is compatible with freedom. This is partially intuitive, intelligence is supposed to restrict choices to "smart" ones, but one would think that some measure of approaching some goals would be maximized, rather than the number of choices.

Of course, there is another potential problem. To learn what choices will be available at a future time the intelligence needs some means of predicting the future, which presumably is contingent on its own free choices, and choices of others. How Wissner imagines this is supposed to work, or how the number of choices becomes finite beats me. Perhaps number of choices is just a simplification for modeling purposes, and what is actually maximized is the hypothetical "number of choices" obtained by projecting intelligence's best guesses about the future. In that case there is even less of a problem with being restricted to a single choice.