No, this is not a blog post about being open-minded.
Nor is it a blog post about brain surgery.
It’s a blog post about Open-Sourcing the code of wiglets allowing others to develop artificial intelligence (AI) algorithms.
The fact is, even though I have a graduate degree from MIT, I may not be the best one to write the AI code for wiglets.
Okay, maybe I am the best … BUT, do I have the time? Is there any time left in the day as I try to start a company?
And besides, open-sourcing the AI component of autonomous animated characters is totally reasonable, considering that the primary goal of our technology is to allow for user-generated content: digital goods created by all you people out there in user-land. I want the wonderful world of wiglets to emerge from the populace – not from the board rooms of marketing teams.
Your creativity and interest can be the driving factor for how these critters come into being, and eventually evolve into the muppets of the digital age.
So, how will we make the brain open-source?
The key is to use the four pillars of situated AI:
Think of actuators as the body. Your body acts on the environment (and generally changes the environment in the immediate vicinity of the body). The sensors perceive the environment and inform the brain of what’s going on. The brain then takes it in and decides (or not) what to do.
Here’s the cool part: what happens inside of the brain can be just about anything. When I was at MIT, Marvin Minsky told a bunch of us that the brain is a magnificent hack: there is no single perfect AI algorithm. In fact, there are many many hacks that have been messily munged together over the course of animal evolution to give us the brains we have.
It’s our frontal lobes that create the illusion that we are making clear, rational decisions – that our brains are well-designed.
This is why some of the early AI programmers made the mistake of looking for the perfect AI. It would seem (to them) that there must be a way to engineer that perfect-feeling of clarity that we call consciousness and rational thought.
But it’s just a feeling.
Uh, what’s my point? My point is that we can take this fact of animal intelligence and apply it to the simulation of wiglets. You (the folks who I’d like to be put in charge of building the brains of wiglets) get to use whatever you want to make wiglets do what they do.
Think of it as crowd-sourced AI.
You can use neural nets; you can use finite state machines; you can design a thousand if-then statements to account for every combination of stimuli; you can attach a big pipe to Google and use the power of the internet; you can make it completely random and hallucinogenic.
Let’s start a Cambrian Explosion of Brain Design!
I have finished version 1 of the Brain Interface, which implements the sensors and actuators (the inputs and outputs). If you, or if anyone you know – knows the C++ language and would like to try out our new brain interface, let me know :)