Artificial Intelligence

I know how to create artificial intelligence. Real artificial intelligence. I conceptualized it in a way that has never been done before. Now I need some talented anon /prog/rammers to help me.

First question : where would you start if you had to reinvent it all?

>>30
I don't think I claimed that Numenta's model is the only way or that it's the most viable solution around, however their model is nothing more than a(n) (couple of) implementations of a given theorethical model of the neocortical column. The actual model may not be perfect, but it does seem to be very close to what other people doing research on the neocortical column have been proposing. It would be validate (or disprove, so we can move forward) these hypotheses about the function of the neocortical column, for example by the means of simulation.

What I'm claiming is that a good model of the neocortical column and a reasonable hierarchical network of them could lead to mammalian-like and quite possibly human-like general artificial intelligence which would be practical to implement and train at a lower cost and in real time (of course, it's more complex than that, there's the problem of sensors and feedback). In no way I'm claiming that this approach is the only one - it's one of many, however I do strongly believe it is worth pursuing. There are of course many other approaches to general AI, and this particular one has its own shortcomings.

>>26
No, I actually believe the problem is simpler than we think.
I tried to work out in my mind how a lot of the usual mental tasks we do can arise out of this one specific model, and to some degree it seemed to work.
In practice this model has been proven to be good enough to be used to identify specific objects in a (moving/animated) scene, identify sounds and other recurring spatial/temporal patterns as well as form "correlations" between inputs. It has been shown to be able to classify objects and reconstruct images, not unlike we do in our brain through the V1<>V2<>V4><>IT circuit. It does show reasonable results at the small scale, however the model itself is just a simplification of a theorethical model of how our brain might work. I'd really like to see how it behaves at the large scale (increase sensory inputs, implement a way for feedback (for example, motor control)) in a real-world environment. Of course, for it to work in a real-world environment, it has to be faster than realtime, which can't be achieved without a hardware implementation. I should also mention that I'm not proposing here to implement a huge neural-net or anything of that sort - that might be useless and you won't be able to actually glean useful data from it. These models are actually debuggable to some extent, although since they're mostly made to learn unattended (but it's not the only way you can train them), actually making sense of some of the data can take effort (however, much less than analyzing even tiny neural nets. You can think of the difference between analyzing a random analog circuit and analyzing a human-designed digital circuit. The analog one is highly unpredictable and may require fairly advanced math to make sense of, while the human designed digital one can be made sense of if you understand the basic principles involved.)

Of course, I may be wrong and I may be overestimating the possibilities of some models, and they might not perform nearly as well at the large scale as they do at the medium scale (for example, processing moving picture/video data) or the small scale (tiny image recognition, pattern detection, ...), however it is the next step I would try if I had the time and resources.