Machine Learning and AI: The Beginnings of Skynet?

One of the more interesting movies of our time was about a computer system that became self-aware and took over the planet (queue “The Terminator” theme music: dudda da da daa-dum).

The convergence of ever-increasing computing power combined with sophisticated computer programming (which was used to teach a computer to program itself) bore out the ultimate cloud system. In the movie, it was an elastic computing platform that could control everything from IoT (internet-of-things) devices like drones/UAVs (unmanned aerial vehicles) to advanced computational platforms for academia focused on solving complex engineering challenges; from hospital treatment systems prescribing cures for complex diseases to factory systems building the very tools that humans used every day. Sound familiar??

This set of utopian solutions set up the makings of a fantastic sci-fi classic. However, as life often imitates our dreams, we are beginning to see these exact solutions realized with the Cloud. You know it as Machine Learning…and we’re hearing a lot about it these days.

But is this “real” AI (artificial intelligence) or just more smoke and mirrors from Cloud providers?

Let’s start with a definition of machine learning…then we’ll talk about where it is today and its potential across all industry sectors as well as the public sector.

What is Machine Learning?

Machine Learning Machine learning is a relatively old term (in computer science, anyway) that is often used to mean just about anything that spikes a person’s emotional response to a computer thinking like a human. Modern day computer science would define machine learning as having a computer with the ability to learn or operate without being explicitly programmed. There is a very important subtle difference there.

For a computer to think like a human, the system would need to have a problem that it wants to solve (curiosity), then the cognition to ask a question, analyze to determine an answer, make a judgment about the validity of that answer, take action, and repeat the cycle. This cycle is often associated with self-awareness – the curiosity aspect that living organisms have (through either voluntary cognition or involuntary genetic response to stimuli). Computers, at least at the time of this article, lack this curiosity aspect.

However, computers are rapidly approaching the computational power to drive the cycle of discovery much faster than humans (and that can be scary!) – and that is machine learning. Simply put, if we can give a computer a model by which it can analyze data, and make a deterministic evaluation over possible answers; the system can take action on those answers, and learn from the results as to the fit of the answer against the data. If you can do this over a sufficiently large data set, so that the statistics (yea, statistics, along with linear algebra and differential equations) bears out an answer that becomes relevant against a data model, you have trained a model against your data without writing any code. Computers are fantastic at applying patterns in this way, and now have the power to drive these calculations at immense speed over internet scale amounts of data. Humans, alas, cannot keep up the same pace!

Real life Machine Learning example

Ok…that sounds complicated, so let’s work through an example.

Let’s say you buy your favorite stock. Your financial advisor or trading partner now has two explicit sets of data to work with – one set about you (name, age, income, wealth portfolio, purchased amount, etc.) and one set about the stock (performance, cost, trading characteristics, demographics that bought it, etc.). Most financial advisors and trading partners will then try to match you (based on your personal attributes) up with other stocks with similar attributers to that of which you just purchased. Sometimes this results in good decisions (gains), sometimes in bad ones (losses); either way, a human is reading reports and interpreting the data to make a decision to recommend a course of action. If they are good (and let’s hope that they are!), they are also learning from their bad actions and adjusting their decision model. Still with me here?

For a computer, it’s not that much different. A human will load an evaluation model (there are incredibly complex models out there in data science that I won’t begin to get into here) into a system like Azure’s Machine Learning in the Cortana Data Platform. This model is then trained against thousands or millions of people that purchased stocks as well as all the stocks that have been traded for days/weeks/months/years.

I say “trained” because Cortana uses the model to calculate deterministic answers to the various sets of properties it has access to, and determines which ones really made a difference in selection. In short: Cortana has now learned what people like you and I purchased that made the types of gains we wanted, and can suggest additional stocks that are tailored to us based on our own individuality. All the while, it is continuing to train the model against new data that is coming in, and discarding old answers that are no longer statistically relevant. It feels like artificial intelligence, but it is really math – done on such a scale that no human could possibly do those calculations real time – almost like magic.

Are the computers going to take over?

What makes the computer so great at machine learning is obviously the Cloud scale computing power. The financial advisor or trading partner in the example above “knows you” and has access to implicit data that a computer doesn’t. In essence, bring on the emotions. Stocks go up and down based on human emotion and reactions to events that a computer cannot predict – like an election or a natural disaster. The system also lacks the ability (at least today) to see the trends outside of the explicit model that was setup or ask the follow up questions about “why.”

Until we can build systems that can apply a machine learning algorithm on top of the results of another machine learning algorithm, and build a neural network of computation, the system will not be able to evaluate and predict all behaviors – for now. Given a sufficient amount of time, with enough computational power, and the right type of recursive evaluation models, one day a computer may be able to develop curiosity and start to ask its own questions – but we are far from that today.

For now, however, we can take advantage of machine learning to make ourselves better at our existing jobs. We are sitting on top of petabytes of data; the history of everything that we have done over the decades. Each industry has burning questions that can be answered with the right machine-learning models:

  • Financial advisors can suggest better wealth investment vehicles based on more accurate performance data.
  • Rather than relying on impact studies—which focus on what has already happened (and that’s too late), state and local governments can build impact “what-ifs” to help them prepare for, prevent, and be proactive regarding an impending or possible change to the economy. For example, how will the community be impacted if a large business closes?
  • The Federal government also benefits. For example, the VA can use predictive analytics to reach out to people returning from combat in an effort to prevent or mitigate the impact of PTSD, rather than identifying and beginning treatment after problems have already surfaced.
  • Insurance underwriters can augment their already powerful data models with the compute power of the Cloud to refine their risk tolerance.
  • E-commerce sites can better suggest other products for you to buy based on your purchase history.
  • Bottoms up drug evaluation results can be better modeled against effective treatment for acute illnesses.

The ability to accelerate our decision cycles to address our own curiosity is available to us right now—through the Cloud.

Do you want to go beyond business intelligence and analytics? To make decisions based on what’s coming rather than what has already happened? AKA can help, offering expert Cloud Services that will help you develop and execute your Cloud strategy, including native cloud development that can bring Machine Learning into that strategy.

Ready to get started? Talk to the Cloud experts at AKA about how you can start capitalizing now on the power of Azure and predictive insights.

By | 2018-03-30T14:26:09+00:00 July 26th, 2017|Machine Learning/AI|0 Comments
Alternative Text

Contributor: Greg Inks

With two decades specializing in Microsoft and Azure platforms, Greg leads AKA's Cloud practice. He is a Cloud evangelist offering deep expertise in Cloud architectures and adoption strategy. Greg has developed subject-matter expertise and wide-ranging business acumen by working with some of the largest, most successful technology providers and client companies on the planet.

Leave A Comment