The cognitive sciences became a hobby for me to read and think about. Artificial Intelligence and understanding of the construction and function of the human mind if paramount in order to create more intelligent machines.

As a programmer, I see two major stumbling blocks for AI:

- Stack-based machines: These machines go deeper and are complex because they need to remember the state where they come from. My intuition says that stack-less machines that always move forward with their state are key to doing this efficiently and reducing the complexity. These stack-less implementations pop up here and there, for example Python. Functional programming (Haskell) is not exactly the same thing, as these use function closures and still have some kind of “stack” available to them. The first CPU that has a totally different architecture should be much better suited.
- Associative search: Not “associative” arrays. When we talk about something or construe our ideas, we don’t load our CPU with an exhaustive search on some index to see if it’s there. It just comes up as a direct association with a particular idea. Computers don’t do that. They need an exhaustive search on some array/index on a particular topic which costs CPU. Processing signals very quickly therefore is very difficult.

I’m reading some books from Steven Pinker about these topics. It shows very nicely that there are different methods or reasons for learning and that not everything we take for granted as “experience” is necessarily learned (but can be embedded in DNA or otherwise is developed very quickly after birth). A machine as a lifeless piece of silicon has thus a long way to go before it can be judged “intelligent”.