One the things I have had a passion for quite sometime is Artificial Intelligence. However, thinking about the requirements, and the technology that already exists it seem unlikely to be possible.

Recently though, while spending time with my nephews, who are both under 5 years of age, it occurred to me that the technology may already exist. In its most rudimentary form, AI is code learning from a history, its environment, and what others can teach it. It then will need to respond to interaction, be it from speech, text, or visual.

So my idea (and trust me its really just an idea), why could it not be done with things like neural networks in code. Building on the basis of code that rather than searching through an index for answers, through the power of distributed computing, use large caches and runtime code that, from template neurons, generate new code from what it has previously learned. use 'short term' memory for common tasks (speech, perception, learning, ect). Then have rapid storage mechanisms for long 'long term' memory. Essentially design the code the way the mind processes data. I realize this presents many technological hurdles, and there are likely complications, or attempts at this very idea that have failed for any number of reasons.

This is my call to the Open Source community. Anyone have ideas? Consider starting an Open Source Project to make even the smallest step towards AI?

EDIT: Just as I finished authoring this I cam across OpenCog. It looks like it could be a jumping point for me idea above.

Donate today, build a world dominating robot tomorrow