Machine Learning is the new C++

I can't help but think this is the new dead end that AI and Neural Nets was in the 1990s. C'mon, who is really using this with a real money allocation to the strategy.

http://www.automatedtrader.net/headlines/153852/gdtrm15-machine-learning-is-the-new-c++

 

Pretty cool article. thnx 4 sharing.

A lot of the problems hat NNets had in the 1990's were related to the capacity of hardware back then as well as several problems with the design of the nets which caused signals to dissipate in deep nets... Now a days we are able to create deep nets which can perform tasks which were unthinkable back in the 20th century...
I haven't heard of any NNet making it huge in the market, but I expect such a net to be kept quiet if it already exists.

 

Outside of trading its used a lot. Kaggle.com

 

umm Renaissance - Jim Simmons.

 

That's one. A very secretive firm at that. For all we know they could be trend followers maintaining a buzz with their clientele. Maybe somebody writes a book about it 20 years from now, like the way Statistical Dispersion strategies have cluttered the internet with power point files. Last time it was highly profitable, in the 2000s?

You can buy stat_arb-in-a-box from ivolatility, btw. Not to say, I don't use the framework for punting stocks. Machine learning, however, is the top of the mountain in graduate curriculum. Cliff notes, pls.

 

As someone that has worked in the AI (NASA) and financial field for the last 20+ years I can tell you it's unlikely that even in 100 years machines will be advanced enough to possess the intelligence of an infant human. What does that say?

 

As my little niece would say: " Hee hee, you're retarded!"

But seriously, it means be careful of the hype machine in every aspect of life.

 

first, they need to even define "intelligence".

 

I disagree... the worst AI algorithm is more useful than the smartest of human infant . I also come from an aerospace background, and that is where my AI experience orignates from (Adaptive Control and Estimation).

Artificial Intelligence isn't really intelligent at all...the name causes people great confusion and gets them all worked up about machines taking over the world. In reality, AI algos are just effecient pattern recognition and optimization algorithms. There is no actual intelligence involved. The designer still needs to tell the algorithm what parts of the world to look at (input data), and how to judge itself (cost function), but the algorithm is much better than humans at developing high dimensional correlations.