There is a technique in engineering known as biomimetics. It is a practice where a designer patterns the behavior of an artificial system after biological processes. An example of biomimicry was the early attempts at designing a flying machine by studying the way that birds fly. The successful application of this approach has bloomed in recent years and yielded many phenomenal devices and processes. One of particular interest to me is the genetic algorithm.
A genetic algorithm mimics the operation of DNA to encode behavior as small snippets of code that can be randomly combined to produce individuals that run that code in a specific environment. The performance of each individual is evaluated against some fitness criteria and then a new generation of individuals is constructed by randomly recombining the components of individuals from the previous generation favoring the contributions from the more successful individuals over the less successful ones. The cycle then repeats some arbitrary number of times or until an individual is “bread” that meets some target fitness level according to the fitness criteria.
Surprisingly good performance is often obtained with relatively few iterations. The behavior produced in this way is not necessarily intelligent. Intelligence implies sophisticated mechanisms for problem solving that involves dynamic adaptation of generic cognitive capabilities to an unanticipated spectrum of problems. Genetic algorithms do produce impressive behaviors though.
I saw a video on the internet today that talked about how some plants used clever tricks to co-opt more intelligent species assistance to accomplish their objectives. The example was given of a plant that had a broad leafy substance that looked like a female insect’s genitals. The male insect attempted to copulate with the plant thus pollinating it without the plant having to expend any energy producing nectar to attract it.
There is a site on the internet called IFTTT (IF This Then That). It allows you to set up conditions that the site monitors for a given value and a set of consequences that the site will execute if the trigger condition is observed. It seems to me that this is the sort of mechanism that is ripe for establishing behaviors that are very clever, perhaps even profitable, without having to resort to actual full blown general artificial intelligence.
And the next step might be, a site that treats IFTTT recipes that are themselves used as snippets in a genetic algorithm. The key thing that would need to be developed would be a fitness function that would judge the generations of individuals until some success criteria was achieved.
I may have to do some hacking myself. I understand that various online stock tickers have IFTTT interfaces. I could build an interface to various news sources and run a genetic algorithm that attempts to pick stocks based on the news sources. I wouldn’t be the first to try it and probably not the first to make money that way if I got it to work.
Sweet dreams, don’t forget to tell the ones you love that you love them, and most important of all, be kind.