people as stochastic calculators? | The Coffeehouse

jason818_253.33
Yippi skippy

lets say you want to know what the stock market is going to do next month, or if its smart or not to get into the buisness of healthcare in the year 2012, whats the best process for creating dynamic programming or building a quantum computer?

Can you send a person/person’s on a random walk using reinforcement learning then take the data and run a Markov process and use that information to decipher a course of action based on their collective desision making?

Currently we don’t have a computer that rival the human intelligence but we do have computers that crunch data and numbers extremely effectively. What if there was a way to merge the two by taking the actions and ideas of people collected and create data sets, then take those data sets and combine them all to come up with an average of some kind that predicts a specific out come. Is this kind of thing already being done?

Bass
www.spreadfirefox.com/5years/

Humans also crunch numbers very effectively, they just do it subcoinsiously. You are a Calculus pro even before you were born!

jason818_253.33
Yippi skippy

In reply to Bass

#Jul 10th @ 6:13 PM

Yes, I agree with you. Wink

Bass
www.spreadfirefox.com/5years/

In reply to jason818_253.33

#Jul 10th @ 6:14 PM

So humans have an easy time crunching numbers too, but intelligence is much harder to describe then some simple equation. Artifical intelligence (AI) is an area of active research in computer science, and always has been. We don't have general AI (AI as you probably would expect it), yet.

But a lot of the research that was for general AI ended up spawning tons of other things completely unrelated, like operating system theory that still exists today.

magicalclick

Stock? In USA, there is no reason, only randomness. I can build the AI just using Random(bool). I am able to sense stock market crash as a whole, but, the indiviual stock is just luck. There is totally no intelligence behind it. For example, you can buy KMart Chapter 11 stock or CitiBank $1 stock and make hack of greens. And you can buy GM which goes bankrupt. For me, it is all about luck and gutt.

Ion Todirel
buuu

>What if there was a way to merge the two by taking the actions and ideas of people collected and create data sets, then take those data sets and combine them all to come up with an average of some kind that predicts a specific out come. Is this kind of thing already being done?

sounds like an AI agent to me, now let me see, we had those since... ever? thought smarter than the old ones

jason818_253.33
Yippi skippy

In reply to Ion Todirel

#Jul 10th @ 7:17 PM

Precisely, a human is a complex IA intelegent agent. I’m suggesting that an IA agent be run threw a complex maze that simulates decision making regarding any one particular topic. But the maze would be an open ended maze meaning the agent would have an unrestricted course of action to get to its destination using a random walk. Then take that data from the Brownian motion/random walk and its out come to see if there isn’t some kind of helpful information garnered from the agent and its ability to get to its end destination. I’m no mathematician, but I think it would be interesting to hear a mathematicians in put on the subject.

jason818_253.33
Yippi skippy

I would have asked these quesions in the Microsoft Research Forum but i didnt see a relavent discussion. So i thought i would try the C9 Coffee House.

Dodo
I'm your creativity creator™ :)

The human brain (average person) uses multi-node processing. A node can contain the electrical value of 0, + or - (that's three possible values). Also, a node influences any surrounding nodes. Values change at an exchange rate of aroung 80Hz. Now, take 80HZ * neural nodes * 3/2 and you get a theoretical performance measure of the human brain. The only problem with computers is, that a logical processing unit isn't able to serve as data storage or able to adapt it's performance path by moving 'wires' or transistors elsewhere.

It's gray matter you're toying with here... if you'd ever figure out a proper self optimizing time-memory tradeoff method you won't be far away from the 'human computer'.

jason818_253.33
Yippi skippy

I found this paper yesterday. I am an intelligent agent not un-like the mars rover. Ok a little more complex living in a seemingly infinitely complex environment in multiple dimensions with lots of noise. I’m making decisions based on a partially observable Markov decision process. There are so many unknowns in making a decision. Its hard some times knowing there is so many unknowns. It can be challenging to get to a specific location when you don’t know of its location or were about to begin with. I guess this is why we make up belief or a policy based off of past experiences to help us when faced with a decision that is new to us. Maybe that’s why we have religion. A belief and policy to help us when we hit a road block. One of the biggest questions facing humanity is how to survive. But what happens when policy isn’t sufficient enough to get us to the next location, in this case survival, then what do we do? Is there an algorithm that comes up with a new and better policy? Any way, that is maybe to big of a question to ask ‘how to survive’. Let’s take the decision of where to eat lunch each day. Do you make a new policy each day to where you will eat or do you eat at the same place day after day to avoid having to wade threw the unpredictable?

ManipUni
Proving QQ for 5 years!

In reply to magicalclick

#Jul 10th @ 6:32 PM

Stocks aren't random, even a little tiny bit.

The information set is just so large that nobody can predict it any better than for example predicting the future. In fact if something could accurately predict the stockmarket then I'd love to see how that same mechanism would do at predicting the future of the universe.

But the good thing about stocks is that even knowing 1/1^1000000000th of the information can make you money. The more you know the more money you make.

jason818_253.33
Yippi skippy

In reply to ManipUni

#Jul 14th @ 2:03 PM

I think your right Manipuni. It is my opinion that Stocks are not random. But we use statistics to help us predict the out come of a very complex situation. Can some one elaborate on the topic of reinforcement learning and what that has to do with all this? Smiley

ManipUni
Proving QQ for 5 years!

In reply to jason818_253.33

#Jul 14th @ 2:29 PM

"It learns from it's mistakes." More or less.

The problem with reinforced learning is that you cannot predict what lessons it will take away. I think AI might be good but only after we're able to get so much information that no human could do the job, and we just aren't at that stage yet.