A.I.

I recently read an article on modern A.I. development and integration. This comes at an opportune time, as I continue trying to develop my own A.I. for various systems (i.e. games). One of the hardest pieces of code to implement in any software is teachable A.I. Normally, A.I. is static and responds to any given situation using predefined, hard-coded rules. While it is possible to create a very believable and intricate service or opponent (if it is a game), in the end a user can identify the patterns and rules the A.I. follows in response to given stimuli. With a static A.I. the more time you spend hard coding possible situations, the more believable the response, because it has been more finely tuned to possible given stimuli. There is nothing wrong with this approach, but it obviously has its limitations.

Maybe the ultimate Artificial Intelligence will be built upon familiar Human Intelligence.

Modern A.I. is moving toward a heuristic approach. Heuristics (as defined by Merriam Webster, of or relating to exploratory problem-solving techniques that utilize self-educating techniques (as the evaluation of feedback) to improve performance) takes a learning approach. Instead of hard coding possible inputs and outputs, the A.I. is given a few ground rules to follow and also the ability to form new sub-rules based on input. This allows the A.I. to ‘learn’ on a basic level. I find it interesting that our most advanced A.I. functions the same as a human infant or toddler. Humans have found that in order to create true artificial intelligence, we must start treating our advances as human intelligences.

There are still limitations in modern heuristics, there must be a few basic hard-coded ground rules for the system to work, and often the range of learning is determined by the amount of time a human has spent on the scope of the “learning code.” One step back has taken us from hard-coded response to hard-coded learning, and as a result A.I. has grown in capability significantly, but what will it take for the next leap? Another step back, maybe? What is it about humans that allows us to learn and grow in intelligence? I like to think it has a little something to do with the way our brains are wired. It’s all about connection, synapse in our own brains. How do you allow an A.I. to make its own connections in this way?

While I think about that, share your ideas, and maybe we can put together Skynet or some Cylons.