Clarifying the uses of artificial intelligence in the enterprise

Artificial intelligence. It’s dominating headlines with the promise of self-driving cars and virtual assistants becoming more real every day. But despite all the talk around AI, no one seems to really understand what it is or how companies can use it.

Is AI the computer that competed on Jeopardy? Or Johnny 5 from the movie Short Circuit? Will machines really take our jobs? As data volumes surge and analytic engines become more mature, has technology finally caught up with the hype?

Artificial intelligence does in fact encompass elements of all these things, but there’s been increasing market confusion around what it is and how businesses can successfully use it.

The efforts of OpenAI and industry development around AI are exciting, but it’s important to accelerate understanding of the topic and terminology. With this in mind, here’s a primer on artificial intelligence and what it means for business.

Artificial intelligence

Many people think of AI as the blending of humans and machines. They’re not far off, but AI is an incredibly broad term — more of an umbrella term, really — that simply means making computers act intelligently. It is one of the major fields of study in computer science and encompasses subfields such as robotics, machine learning, expert systems, general intelligence and natural language processing.

Apple’s Siri, Google’s self-driving cars and Facebook’s image recognition software are standard examples of AI. But it’s much broader than that. AI also powers product pricing on Amazon, movie recommendations on Netflix, predictive maintenance for machinery and fraud detection for your credit card. While these applications are all powered very differently and achieve different goals, they all roll up into the umbrella term of artificial intelligence.

Intelligence has never been sexier.

From a business perspective, companies wouldn’t simply “buy” an AI solution. Rather, they would likely leverage one or more of the subfields of AI and buy software packages like R, Python, SAS, and MATLAB for statistical analysis. But new technology is pushing beyond traditional statistics, and machines are acting more intelligently than ever — they’re not just doing the analysis, machines are now finding patterns in data and figuring out how systems “work”… often without any human intervention.

Let me stop here for a quick, yet important, PSA — neither artificial intelligence nor machines will replace all of our jobs. This is perhaps the biggest misconception about AI. Everything under the AI umbrella — including machine intelligence and machine learning — is data-driven, but requires human expertise to apply answers and discoveries to solve problems.

AI will allow people to do new and interesting things that have never been done before. That’s the fun part, so let’s take a look.

Machine learning

Early on, AI was focused on expert systems: establishing if-then rules to mimic human knowledge and decision-making. Expert systems fell out of favor because they didn’t leverage data, didn’t learn from data and didn’t scale. They were entirely limited by the programming and cerebral capacity of the people who created them. Today, machine learning has replaced expert systems.

Machine learning refers to the statistical arm of AI. Most of the people or companies leveraging “AI” are referring to machine learning, not doomsday robots. The focus of machine learning is on programming algorithms to learn from data, complete tasks and make predictions with an emphasis on high statistical accuracy. It is not used for the discovery or interpretation of data — this is important to know, and something we’ll cover more in-depth later.

Machine learning algorithms can be developed, run and tuned with libraries from software packages like SAS, R and Python. Data scientists and statistical analysts typically work in these programming languages, ultimately applying the resulting algorithms to enterprise applications like sales forecasting, email spam filtering and determining where the next hurricane is likely to develop.

Because of machine learning’s widespread applications, there exists a plethora of tools that empower its implementation. Yet the “rules” and predictions uncovered by machine learning algorithms are still unable to solve many business problems on their own. Without data scientists, businesses have trouble interpreting the algorithms and developing an understanding of the why.

Has technology finally caught up with the hype?

If we’re a Fortune 500 retailer, do we care exclusively about predicting sales, or are we equally concerned with why that number? What are all the variables and relationships at play, and what can we change to improve that outcome? How much of our revenue can be chalked up to weather patterns or marketing spend? Are we stocking the correct amounts of each product, or should we optimize for impending tastes and trends?

If we’re an engine manufacturer, is it useful to merely predict that the assembly line will produce a faulty component 1 percent of the time, or is it more helpful to be able to understand all the knobs and levers that contribute to the failure so we can actually change the future and our processes?

Welcome to the next phase of AI.

Machine intelligence

Machine intelligence is the newest subfield of AI, focused on learning and interpretation of data. It’s a natural progression of machine learning, but takes it a step further.

One of the shortcomings of machine learning is that machines are learning but not conveying to us what they’re finding from the data. To make data valuable, we need to be able to understand it and explain it; only then can we connect the dots and apply it back to the business.

This is machine intelligence — the interpretation and understanding of data — and why it’s so important.

IBM had one of the first attempts at machine intelligence. Watson, IBM’s Jeopardy-aceing robot, uses natural language processing to interact with and process data, by converting speech to a scalable search through the World Wide Web. Watson can listen to anyone, interpret the information then search its database for a response to a question that somewhere, at some time, has already been answered. Where Watson falls short is its inability to infer new ideas or answers. Similarly, other services like Microsoft and Amazon provide platforms for running machine learning algorithms, but do not facilitate interpretation of results.

What if we could take this a step further and have machines discover answers that are not yet known? Machine intelligence has the ability to not just uncover answers, but to understand and interpret what it has learned.

Machine intelligence is the next exciting progression of artificial intelligence. Whereas machine learning will accurately predict that your electric bill will increase next month, machine intelligence will accurately predict that your electric bill will increase next month — and tell you why: your travel schedule will be light, the weather will be hotter than usual and your air conditioner has been deteriorating. Machine intelligence teaches back to the human the reasons why things happen or will happen, arming users with the ability to make quick and justified changes in strategy. Machine intelligence, not “big data,” offers the “actionable answers” businesses need.

Unfortunately, some people have been exploiting the ambiguity around artificial intelligence. AI, in and of itself, is a relatively basic, high-level idea that machines can be programmed to act “intelligently.” Machine learning dives into the idea that machines can actually learn without explicitly being programmed. Finally, machine intelligence exhibits the ability to not only learn from data, but to actually clearly articulate answers and discover why. That is, machine intelligence is the first instance of machines teaching the human and relaying brand new discoveries automatically.

This is an exciting time, and we’re only on the cusp of what’s to come in AI. Machine learning has been the panacea for decades. Machine intelligence is the next step in the progression, allowing people (and companies) to understand why things happen and what to change to generate their desired outcome. Intelligence has never been sexier.