AIs vs Human Traders

I have one simple question: do you think that in the coming decades Artificial Intelligences will eventually make human traders useless?


Being a compulsive future student (mainly of Ray Kurzweil, he has great stuff), and being aware that in ten years computers will be 1,000 more powerful, and in twenty years they will be one million times more powerful than today, i can't help but think that eventually they'll be so powerful that they could get much much better at trading than humans. If nowadays they're already finding their place in the markets, in the future they may become so much better that they'll make humans eat dust. Just think of chess as an example (in the past, people would laugh if someone said that a computer could eventually beat a grandmaster).

I'm aware that chess and trading requires two different skills, and the big advantage traders still have over computers is that they're so much better at recognizing patterns, but if nowadays there are already people working on programming computers to recognize patterns, in ten or twenty years i can only imagine how good computers will get.



What do you think? I can not form a completely educated opinion on this because while i know the fields of trading and of countless visions of the future, i have no knowledge about programming and its several nuances and hence i have no idea of its possibilities with much more powerful computers. I'd love to hear from a few programmers.

Let's look at the bigger picture...

What ultimately drives the markets? Supply and demand. Look at the forex market for example, what is the driving force behind price movements? Is it the banks? Is it the hedge funds? No. It's your plain vanilla multinational corporation. To illustrate...

Apple makes iPods in China, therefore, it has to pay its workers in Yuan or RMB (RenMinBi - The People's Money). iPods are sold in the US. That is to say, Apple gets dollars for its ipods. Apple cannot pay its Chinese employees in dollars, so they go to the local bank to trade their dollars for yuan. The same applies to other corporate giants such as Microsoft, Walmart, Dell, HP, Exxon Mobil, Goldman Sachs, and so on.

Our role in this market is not really necessary. Traders like us take advantage of specific ineffeciencies in the market. For example, Apple releases this new super product. It's so amazing that over 4 billion people in the world try to buy it. Let's further say that each item is $100. That means that Apple has $400,000,000,000 (400 billion) dollars equivalent in other currencies (euro, pound, rand, ruble, rupee, yen, frank and so on). Apple needs to trade their foreign currencies for dollars. This means that there is a higher demand for dollars. With the increased demand, the price of a dollar relative to other currencies increases, which raises the price of the dollar. Traders like us see the trend and we too buy dollars so we can sell them to other buyers at a higher price.

Now, if we were to replace every human trader in the world with a neural network, or other robot, I doubt there would be much chaos in the market. Each robot may very well end up working together.

So long as there is an inefficiency in the market, there will be a need for a trader. Otherwise, there will not be any room for us or machines as traders to make money.

Good reply, Ronald.

But the question of whether machines will get better than humans at spoting inefficiencies in the market in the coming decades still remains. I believe they will, it's almost impossible, with their future computing power, that they won't. So this has led me to the doom conclusion that if i'm going to trade for a living i better make enough money to retire in the next one or two decades.

You answered your own question in your 3rd paragraph.

Any computer can be programmed to recognize any patterns but it take a special human eye to decipher the true meaning of the pattern.

That's the current reality Jhig, and equilibrium, I do believe that machines will eventually surpass humans. But that doesn't mean human traders will be obsolete. Look at what I currently do, I don't actually place trades myself anymore. I have a small army of EAs which I use to place the trades for me.

You're cracking up Ronald. Take it easy. It's Friday night.

Well then in your opinion the future for human traders is either to retire or to become programmers... good to know that someone with knowledge in programming thinks like this too (well it's actually bad because gives more confirmation to my fears).


Of course, i'd still like to hear many more opinions.

The more you get into AI, the more you realize it's limitations. All AI boils down to some kind of search or regression, with a metric for ranking solutions, the general problems as I see them are...

1) AI cannot define which inputs are relevant, humans have to do this.
2) AI cannot separate rubbish solutions from physically meaningful solutions (e.g. 'curve fits', 'local maxima' etc).

This is not to say it is not useful, it most certainly is (I use it myself), but a panacea it most certainly isn't. My take on the original question is that humans will always have a place, even if it is just designing better AI algorithms.

Another angle from which to approach this question is are you in the 'Strong AI' camp? This point of view is that human brains are nothing more than very complex Turing machines, so it is inevitable that AI will exceed human intelligence at some point. I personally don't subscribe to this point of view based on the arguments laid out in the book 'The Emperor's New Mind' by Roger Penrose, but many others do.

Will AI surpass humans?

I don't really know, market geometry is really complex, i've worked with NNs for years, and i really belive that right NOW with the current hardwares there is not the possibility to create a such complex NN that is able to learn a such complex market geometry.

But, the future can give us mane changes, let's think about the Quantum computing NNs, they can be the revenge of Neural Networks, of course we don't know if they will be powerful enough to learn soo much complex things, maybe yes, maybe no. We need to wait to understand, but right now, the AI is totally bullshit upfront the complexity of the market.


P.S.
There are also right now AI programs that are able to have regular profits from the market, what i was meaning in this post is an AI that can understand at 100% the market (the perfect trader - the holy grail AI), but right now there are tons of profitables AI that are not perfect, but work well.


(No, there is no AI that is good as a human for now, and maybe there will never be)

I'm bored.

Waiting for the P&L to come back from risk.

Never.
The Best that an AI can do is make you aware that a certain set of circumstances have been met.
After that, it's up to you.

Computers are already superior in certain areas of trading where speed matters. High frequency trading would not exist without computers running highly optimized trading algorithms. However this has little to do with AI and much to do with latency...

Point being, computers already has and will continue to create new ways to approach the market, altering the market along the way, but it does not mean that humans does not have a place in this market. Those who understand how to exploit new the technology do have a huge advantage though (as in always).

Even EAs require recognizable traits in price action. To be profitable you have to have volitility to give you price action .. MOMENTUM ..
Even in the currently choppy range bound market we are experiencing now in the majors to be profitable you have to reduce the time frame to see the tradable trends, the daily at this point is nearly untradable. (hmmm is that a word)...... however if you shift to 4 hr or less you will begin to see some very good trend trade opportunities. You will have to adjust your SL and TP to reflect these shorter TFs however.

As you can see the tradable trends become more obviouse as you reduce the TF.

http://img33.imageshack.us/img33/719...encapturex.png

http://img10.imageshack.us/img10/641...ncapture1j.png

http://img196.imageshack.us/img196/4...ncapture2q.png

http://img11.imageshack.us/img11/918...ncapture3q.png

You all seem to be missing the point, here.

Artificial intelligence is not a very powerful computer or a very well programmed computer. The definition of artificial intelligence is a computer that can think for itself. It will be able to learn and, given enough processing power, will be able to extend its knowledge and ability to the maximum extent of its available inputs.

Whether such a computer would achieve anything useful to humanity or even accept commands from a human being is open to debate.

The Future ???
If the Random Walk theory is correct, it makes no difference.
Even after any NN {or AI} examines most recent data (& adjusts for it), It still could be wrong for the next set of Data.
That leaves (i would argue) Nothing But The NOW.

The Now
Wouldn't it be nice if someone came up with, even a reasonable expectancy of, something based on what is Happening At This Moment!
Now that idea would get a lot of attention.


The Past ???
If History does, in fact, repeat itself?
Well; That's a different story & offers credence to any Indicator that relys on Old Data.

Things that make you go...
Hmmm...
{my brain hurts when I attempt to extrapolate these concepts. :nerd:}

Computers with artificial inteligence ....? Hmmm... and here I've been thinking all along that the human brain was already the epitomy of artificial intelligence.

In the greater context of things...anything made by a human is technically 'artificial'

This may be a bit of a stretch but...

Have you noticed how the current crop of "World leaders" are increasingly fond of mouthing the words "New World Order"? If they and their henchmen (the beaurocruds) have their way, it may be entirely plausible that the world will end up with a singular "New world currency". Lately we have heard China make noise about a new reserve currency and the Russians have already "minted" their version. The EURO is in effect an early experiment in such a Socialist currency so we may not be all that far removed from a single world denomination. That will make currency trading obsolete and Artificial Inteligence as applied to currency trading a mute point.

The models used in programming of these AI will be created by humans. Such models act based on predefined probabilities based on the historical data of market movements. Under normal circumstances these system most of the time will take right decisions which will be the most probable outcome. They will define the Risk:Reward ratio better than humans and take unnerved decisions. But in abnormal market movements when the volatility and market moves make unnatural and abrupt moves if the AI model is unable to match the current situation with anything that has occurred previously then based on the most probable outcome it will make wrong assumptions and fail. This will happen because it will not have emotions like fear and greed which could be useful at that time.

Just my opinion..