Artificial intelligence — and “deep learning,” a specific form of artificial intelligence — is changing the face of investing.

This is not something that will happen in the future. It is already happening. The changes are widespread and they will only intensify in the coming years.

This could be seen as good or bad news, depending on how you look at it. Artificial intelligence, or AI, is a powerful form of technology that is brand new to most people. That makes it frightening as well as exciting.

There is also a lot of hype surrounding AI, including overblown predictions of what artificial intelligence will do (and what science will be able to achieve). Sometimes it’s hard to separate the reality from the fantasy — and some have a vested interest in drumming up fear and worry.

The good news is that AI, through a technique known as “deep learning,” will be an incredible resource for individual investors. The widespread availability of low-cost computing power and increasingly powerful software programs means that the benefits of AI will ultimately flow to Main Street, and not just Wall Street or Silicon Valley.

And yet, when you hear about artificial intelligence in the news these days, it is often attached to a big prediction or a slightly ominous sounding breakthrough. For example, a London-based artificial intelligence company called DeepMind — which is owned by Alphabet, the parent company of Google — recently announced an “intuition” breakthrough in one of its game-playing AI programs.

AlphaZero, a descendant of DeepMind’s AlphaGo, is an AI machine that shows human-like creativity in games like chess and shogi (Japanese chess).

Human grandmasters, like the world-famous Garry Kasparov, have confirmed the surprising degree of “creativity” in AlphaZero’s style of play, with a willingness to take risks and make bold, unconventional moves that feel more human than machine-like.

The gameplay, however, is just a means of testing capabilities and generating publicity. The goal of DeepMind is not to design an ever-more-impressive roster of game-playing machines, but to tackle tricky and lucrative real-world problems, like the development of pharmaceutical drugs.

In the pharma world, for example, predicting the three-dimensional structure of proteins — also known as “protein folding” — is an important area where DeepMind hopes to apply deep learning techniques.

Protein structures are at the core of many life-saving drugs. If you count up all the seconds that have ticked by since the universe began, there are more potential protein-folding combinations than that very large number.

That makes the design of new protein-based drugs very challenging, and an area where an AI program — like a pharmaceutical version of AlphaZero — could accelerate the process by orders of magnitude.

It’s important, though, to clarify what is not going to happen.

AlphaZero is not close to “awareness” or “consciousness” or anything resembling human brain activity. And in fact, this particular area of AI is widely hyped and overblown.

Broadly speaking there are two types of artificial intelligence: narrow and general.

“Narrow AI” is focused on a very specific task. It is far more like a software program than anything else. The much more ambitious “general AI” is the notion of computer code achieving something like human consciousness.

Narrow AI is already here. You see it in all kinds of places, and likely make use of it multiple times without realizing it on any given day.

Narrow AI does things like auto-complete the text you type into your smartphone, translate a foreign language web page, or give suggestions for restaurants or coffee shops based on your GPS location. It does very specific things, typically by parsing large amounts of data in real-time.

General AI only exists in Hollywood movies. It is the computer that is supposedly smarter than a million humans, or the army of terminator droids trying to wipe out humanity.

A handful of experts think General AI could arrive in a decade. But those are extreme outliers. A far higher number of experts think General AI — a sort of computer-based consciousness, or a program smart enough to have awareness — could take 50 to 100 years, or in fact may never arrive at all.

In spite of the fact that smart speakers like Alexa can mimic a conversation, conscious AI is nowhere close to being achieved, and may not even be possible.

The current AI techniques being deployed aren’t in the same ballpark as General AI. They aren’t even the same sport. We may actually be closer to reaching Mars, or even colonizing Mars, than we are to any kind of substantial General AI breakthrough. That’s how big the Narrow vs. General gap is.

So, even though DeepMind as a company talks about its AI having “intuition,” we shouldn’t confuse that with any kind of march toward human consciousness. It is possible for an AI program to show what appears to be creativity, and to be useful in all kinds of powerful ways, without being conscious at all.

This is what “deep learning,” a specific type of artificial intelligence, is all about — helping human researchers (and investors) become more powerful in various ways.

Deep learning relies on “neural networks.” Because of this, the claim is that deep learning, as a technique, mimics the structure of the human brain. This is far from true.

It sounds sexy to suggest that AI mimics the human brain, because it implies that, if you go along this path far enough, you get consciousness.

The reality is far more basic, but still fascinating. It is possible to exploit the power of “neural networks” without trying to copy the human brain at all, except in a super-abstract way, and that is what deep learning does.

The “neural” part means storing information as a network of nodes, with recognition capabilities — the program’s version of awareness — distributed across multiple nodes, rather than residing in any one place like a text file.

To understand how deep learning works, imagine you come from a far-off land where there are no housecats. You have never seen a housecat before, or a cat of any kind.

Traveling to the United States, your host wants to teach you what a cat is. But instead of describing a housecat, or introducing you to a live one, they show you pictures of housecats.

After a while, in order to test your knowledge, your host starts showing you thousands of pictures, some of them with housecats and many of them without.

You make guesses as to which picture contains a housecat and which doesn’t. With each guess you get feedback — “correct” or “incorrect” — and over time your guesses improve.

Eventually, sticking at this for long enough, you have a pretty good sense of what a housecat looks like, thanks to huge volumes of trial and error.

Deep learning as an AI technique essentially does the same thing.

A computer program is taught to identify a pattern — like, say, the shape of a cat in a picture.

The program tries to “guess” at the pattern, over and over, getting feedback each time. After hundreds of thousands or even millions of guesses, the program has a pretty good sense of what a cat looks like.

This methodology requires huge volumes of data for the program to train itself. That is why giant tech companies like Google, Amazon and Alibaba have such an edge in these areas — nobody else has access to oceans of data like they do.

But again, this brute-force means of recognizing patterns within patterns is nowhere near human consciousness. It is a million miles away from it.

And yet this deep learning technique is extremely valuable, because AI-powered software can:

  • Detect data patterns that are very subtle and complex
  • Sift through huge mountains of data (find needles in a haystack)
  • Get better at pattern-spotting through testing and feedback
  • Identify useful or valuable patterns instantly and in real-time

And that, in turn, explains why deep learning is the future of investing.

In the hands of regular investors, artificial intelligence tools can scan vast quantities of data to identify useful and important patterns via deep learning techniques. Investors can then use those patterns to make better investment choices.

A key point here is that, from an investing perspective at least, the human being does not get replaced. Human behavior still plays a key role. Human decision-making and human emotion are still big factors.

But software with AI-like capabilities, enabled by powerful deep learning algorithms, can serve as an investor’s eyes and ears.

The software can scan thousands of stocks in real time — something a human can’t do. The software can also look for subtle patterns in the investor’s own behavior and investment record, and make possible suggestions for improvement. These capabilities, and more, make the investor more capable and powerful — and potentially more successful.

Those ideas just scratch the surface. The point is that, while artificial intelligence is a big, complex topic that is both exciting and a little frightening, the dawn of AI — via deep learning — is opening up a world of new possibilities for individual investors.

And this AI investment revolution, so to speak, is truly democratic because the barriers to entry are low and falling. With each passing month, if not each passing day, chips get cheaper and investment software gets more powerful.

That means you won’t have to be a Silicon Valley tech titan or a rich hedge fund mogul to utilize AI-powered software. We can be certain about this because our mission and vision, as a software company, is to empower individual investors.

It’s our bread and butter, and we can’t wait to show you what’s in store for 2019.

Richard Smith, PhD
CEO & Founder, TradeStops