Its been 10 years or so since I looked at ANNs. The last time I did, training them almost meant showing them every possible combination of inputs. There were some that were a little better than others. Does anyone know if there are newer ANN (or other AI) techniques that are better able to guess a function with smaller sample sets? I've seen deep-ai and convoluted ai, these still require large sample sets / lots of training, right? Not to mention complex set-ups? Along those lines, anything else interesting come up in the last 10 or 15 years? The last cool thing I saw was the novelty-seeking attempt at artificial consciousness, but it was just that first paper. I REALLY want my own artificial consciousness.
‘Artificial Intelligence’ Has Become Meaningless It’s often just a fancy name for a computer program. https://www.theatlantic.com/technology/archive/2017/03/what-is-artificial-intelligence/518547/ It's not clear that consciousness has any impact on better trading in any event?
First off, the author of the article, Ian Bogost, is a professor of interactive computing at the Georgia Institute of Technology. What exactly are your "scientist" credentials, by comparison? Back to talking about the actual article instead of an ad hominem (as well as inaccurate) attack on the author, I think it's spot on. We have a situation where a bunch of non-scientists are calling everything that contains a line of computer code AI when in fact it's no such thing. If we're talking about using AI for trading, we're no where near there yet and there in fact isn't anything to indicate that AI would be beneficial to trading. We do have some pretty big correlation engines working on finding correlated assets, and other similar pattern matching programs. We also have a bunch of high speed arb algos. But nothing that someone who really knows what AI is would call AI.
http://bogost.com/uncategorized/about-me/ Bogost holds a Bachelors degree in Philosophy and Comparative Literature from the University of Southern California, and a Masters and Ph.D. in Comparative Literature from UCLA. He lives in Atlanta.
I'm sorry, I missed your qualifications as a "scientist", can you run those by us all again? I am but a humble electrical engineer, so also I guess not a scientist. But like anyone with a liberal arts education I certainly understand both the concept of the scientific method and the definition of an ad hominem attack. As far as I can tell the author is a professor of interactive computing at the Georgia Institute of Technology, and you're a jackass on the internet. His achievements say enough that it's worth me reading his article and then I'll judge it it on its merits. Since you haven't demonstrated that you're anything other than a jackass on the internet and don't have any points to make at all.....well you're really just destroying value here aren't you?
http://bogost.com/uncategorized/about-me/ Bogost holds a Bachelors degree in Philosophy and Comparative Literature from the University of Southern California, and a Masters and Ph.D. in Comparative Literature from UCLA. He lives in Atlanta.
Yes so do I. These days they are not called ANN but usually go by the name of Reinforcement Learning, Q-Learning, or Deep Q-Network etc. Programming Agents for trading is not so easy however. As to 'small datasets' - it could be as small as a 3X3 tic tac toe grid, or even only feedback parameters around a simple poles balancing movements.
Its almost nothing but curve fitting but once in a blue moon you might luck out . My software just ' does ' it . If there is logic behind a system the optimization can be beneficial but in the main i think this isnt that useful