I am afraid it is the only way to go in the future. The problem is that AI has to compete with the most complex and fast CPU we faced so far – human brain. As a result every AI algorithm is an attempt to replicate intelligence provided by our brains. Since this is not feasible (so far) researcher are trying to use simpler ways of getting intelligence. A good example is a swarm intelligence. Here complex behavior arises from numerous constellations of simple animats’ reactions to environment. I would not be surprised that this is the way of achieving the final goal – AI intelligence comparable to human.

There was a time when people believed that the Earth is flat, a machine heavier than air cannot fly (up 8)...
... any argument against is only an argument. Any failed attempt is not a proof that AI could be build from small "bricks" (simple mechanisms) but only one example of unsuccessful attempt.

As for human brain… we are competing with other people. The day when we will have another opponent would be a turning point to assign new goals  

I think it sounds great- lifelong vacation for everyone!

Here's one human-brain modeling project that I have been following with interest:
http://www.numenta.com/

And here's an excellent summary presentation by the project's co-founder, Jeff Hawkins:
http://www.ted.com/index.php/talks/jeff_hawkins_on_how_brain_science_will_change...

I have to correct myself on FIFO optionality (after some hint from the video). Let's call it Fuzzy FIFO (FFIFO). Here it is:
Learning (rule in): First, the system (your brain or trading system, pick your ponny) creates very general idea of a new rule. After several encounters with the same rule the idea gets more and more details. It brunches.
Forgetting (rule out): First, the system erases small details (try to remember events from you childhood), then it slowly removes brunches... until there is only initial general idea.

Now the most interesting part: In reality the system does not erase anything. It removes only hash table. All the knowledge is stored. The moment when old pattern re-emerged it brings old memories back...

The funny thing is that in AI there is a system where you can add FFIFO in no time... do not ask me, you have to come to it yourself  

As one increases the number of degrees of freedom, those "FFIFO" tricks become handy (thanks for the new term, Coolcat!). They help models generalise better and avoid the bad curve fitting. There are numerous ways to add those.

I will try to implement FFIFO and see what it would give. It should take several days.

Thank you.  

First, it is offtopic. Later we will move this part of discussion to "OFF-TOPIC" section of this forum.

Second, for our English speaking members and readers: This is very interesting article written in Russian slang (I have to confess that after I have read it my Russian would never be the same, as well as my way of thinking... ).

Third, maybe we should create special section: HUMOR for Algo. I have a tailored joke: He is a nice guy, but he is doing Algo