Evolving the infinite game

The “Infinite Game” is a game whose purpose is to never end. In other words, a successful game is one that does not end and if it does, you lose. It was brought up in the context of evolution and technology. Furthermore, the concept brings about not just that the game does not end, but that the game increases in diversity, complexity and specialization, just like evolution and technology. In other words, the game’s purpose is to have more of it on any “complexity”-dimension.

Another ingenious idea, by Schmidhuber, is the concept of evolving the rules of the game such that it will be enjoyable. The novelty here is the definition of what is an enjoyable game. According to Schmidhuber’s theory of fun, an enjoyable game is one that is learnable, i.e. that it is hard to learn but feasible to learn. He then evolves the rules of the game such that learnability is increased, as measured by how long it takes for a simulated player (a neural network) to learn the game.

Let us combine the two ideas. Now we want to evolve a game whose purpose is to never end, and yet for it to be enjoyable. This means that the rules must not only evolve but also change with time, otherwise it will not be an “infinite game”. Hence, the evolution will have to take care of comparing the rate of learnability, i.e. how fast can one learn the rules and understand the game in order to advance, and the rate of rules’ change, i.e. how fast do the rules change such that a good player becomes not so good again.

This concept of evolving a game to be both fun and never ending leaves out an obvious concept of games: actual winning. Here, as in Schmidhuber idea, the game can end and can be “won”. But now, the game’s desire, not the player’s, is for it to not end. There is thus a balance between the player’s goal of ending the game and winning and the game’s goal to continue and challenge the player.

If evolved correctly, this can result in a rule for how rules change. In a more complex setting, this rule can be related to the progress of the player, her learnability and her adaptation to the very rule that changes the game. Yep, this is a circular thing that can go into your head, but create super cool games!

Close-loop body

A combination of new emerging fields may enable the opportunity to "close-the-loop" on our own body. The first is "personalized medicine", which promises to tailor each person the specific medication she requires based on a full analysis of her genome, epigenome, proteome and other –omes. The basic tenet of this new and exciting field is that by knowing each person's genetic and protein make-up, we can better design a medicinal treatment for each illness.

The second field is "self-monitoring", which at its extreme results in taking a blood sample each day and completely analyzing its cellular and molecular content. In an ambitious self-experiment, a researcher (forgot his name) did that on himself for several weeks. One finding, for example, was that these measurements indicated a flu he had much earlier than any other symptom he felt. Furthermore, the decrease of prices for such tests, following Moore's law, can results in each person being able to administer such measurements on a daily basis.

Finally, the field of "specialized consumption" (I made this name up) claims that one can drink a single fluid that contains the entire bodily requirements of the human body. No other food sources are needed to survive and thrive. These liquids, while being probably not that tasty and expensive, enable a much tighter control over what goes into our body.

Combining these three fields I suggest an experiment to try and close-the-loop on a human body. By this I mean that a person consumes only the aforementioned liquid, thus fully controlling the input to the body, while at the same time having a complete monitoring of the body functions, via the aforementioned measurements. Closing-the-loop here means changing the contents of the input liquid based on the results of the measurements in an attempt to reach some kind of equilibrium.

Obviously this is not an easy experiment, but it is not dangerous or harmful in any way. The results would also obviously be different for each person, but by closing-the-loop one may attempt to "conform" the body function.

If this experiment works, it means a whole new level of human existence, in the sense that one can have a full, and rather straight-forward control over one's own bodily functions.

Objective news assessment

News come in all forms and shapes. While most filter them according to content and subjective assessment of interest, why not try and introduce some objective measures of each news item? I propose several relative easy and simple to implement criteria of assessment by which news can be categorized and then filtered or segmented.

The first is the number of people it reports on. While this seems like a crude measure of importance, a news item related to one person that got rehabilitated (no matter who that person is) objectively is less important than famine influencing several million people. By tagging each news item on the number of people reported on (or a rough estimate thereof), one can slice news to people-importance.

The second is monetary. The cost of the item reported, either in tax-payer money or cost of reconstruction, cost of change, influence on GDP, etc. This gives a rough estimate on the financial effect of the news item. Mind you that not only financial news should be assessed by this measure, but rather every news item. Thus every natural disaster or car accident, every joyous event or festival should be tagged with a monetary measure, such that the influence of the reported item on finance can be immediately filtered.

The third is the political hierarchical level of the news item. By this I mean how up does the news item reach? Obviously, presidential reports are pretty way up, but senators, mayors and even local news can be rated according to objective measures of hierarchy, e.g. how many people between the reported people involved  and the president. This way one can assess the political influence the news item has.

One can introduce other important objective measures of news items, e.g. technological measures, social measures etc. This way, one can filter, select, evaluate and relate to news items not only based on content, but actual measures of influence. This way news media can be assessed on the importance of their items. Obviously, this is not intended to replace the content filters but rather augment them. It would be nice to see how much prime-time news are actually about important things.