AI EA - Next level of EA's?

 

Hi all,

Recently i can see a lot of buzz around the Artificial intelligence EA's.
About how they can learn the pair characteristics, and get better all the time.

Anyone can point me to a more detailed resource about that niche? Some articles? Maybe a book?
Is it still involve only MQL4/C/C++ or there is a need to learn new languages, and use special API's?

Oh and the most important of all, what's your overall opinion about this kind of EA's?

 
msbusinessil:

[...] Artificial intelligence EA's. [...] Anyone can point me to a more detailed resource about that niche? Some articles? Maybe a book?

These algorithms are a variation of Neural Network algorithms. There are a few articles about NN in the Articles section: https://www.mql5.com/en/articles/mt4/search/neural%20network.
 
msbusinessil:

Hi all,

Oh and the most important of all, what's your overall opinion about this kind of EA's?
In my opinion, AIEAs at this stage are more of a marketing tool than a real solution to the never-ending challenge of finding signals that work 100% of the time.
 
it's difficult to be objective in this question. i'm glad that it exists. there are a lot of different information on freerobot2010. if someone is interesting, try to surf this site.
 
gordon:
These algorithms are a variation of Neural Network algorithms. There are a few articles about NN in the Articles section: https://www.mql5.com/en/articles/mt4/search/neural%20network.

My son-in-law for his doctorate thesis developed an AI program that could read mammograms and determine from the slope of the gray-tones whether a speck was benign or cancerous.

He had hundreds of specialist opinions on old mammograms with the evidence of subsequent developments to build into his AI program.

When I showed him the work currently under way in AI forex trading, he said "if you read the introduction to my thesis, you will find the same theory, but it's a long way from a solution".

In his opinion, an AI engine to predict the currency market is multiple times more difficult than his challenge of reading mammograms by computer.

 

gordon
:

These algorithms are a variation of Neural Network algorithms. There are a few articles about NN in the Articles section: https://www.mql5.com/en/articles/mt4/search/neural%20network.

Thanks for the great resource!

engcomp:

My son-in-law for his doctorate thesis developed an AI program that could read mammograms and determine from the slope of the gray-tones whether a speck was benign or cancerous.

He had hundreds of specialist opinions on old mammograms with the evidence of subsequent developments to build into his AI program.

When I showed him the work currently under way in AI forex trading, he said "if you read the introduction to my thesis, you will find the same theory, but it's a long way from a solution".

I understand that this is not the ultimate solution. But is it better to not use it at all, then optimizing your code and get the advantages of it?
BTW, can you give us some details on it's biggest advantages / disadvantages?
 
msbusinessil:
BTW, can you give us some details on it's biggest advantages / disadvantages?

The biggest disadvantage in trusting an AI solution is the same as trusting your interpretation of the fundamentals.

In my experience I found that I get attached to my interpretation because I invested so much time in it.

When the market doesn't agree with me and does the opposite, I used to stubbornly insist that I am right. No longer.

So you invest a lot of time in an AI solution. Will you trust it in the face of a contrary market and defy the market?

If not, you might as well stay with your moving average cross with a decent whipsaw filter.

 
engcomp:

My son-in-law [...]

On a similarly anecdotal note, I know a lot of people who did doctoral and post-doctoral work in AI labs, and now work in financial services building automated trading systems. But it's almost entirely because they're used to building computer systems which process large amounts of data, and because they're good mathematicians, not because they're actually using anything which you could reasonably call AI.
 
The AI concept have allot of moral drawbacks. If I'm an employed trader, I'd never cooperate fully with the AI programmer who's trying to put me out of work. If computer AI could figure out the markets better than humans, the financial markets would be doomed. Otherwise, It'll be a race for the better AI, like Chess no-one would make any money as the Game between AI's would end in a draw every time. Just my opinions tho. But I guess that's one way to create a Efficient Market.
 

Markets move and react on human emotion based on fundemental factors. The best AI could do currently is to self optimize based on past history, and the predictive value would be limited at best.

 
I guess nobody really likes AI EA's :)