Guest Blog Post by Innovation Keynote Speaker Kian Gohar

It seems like everywhere we look, Artificial Intelligence (AI) is poised to be a major disruptive force in our lives: it’ll help us search for information, it’ll power self-driving cars to transport us, it’ll help us diagnose diseases earlier, and it may even automate some jobs to the dustbin of history.

But what does AI really mean and how can it be harnessed for business innovation?

The truth is, AI is no one single thing. It’s a generic term used to describe a suite of software algorithms that manipulate data to accomplish human tasks like reasoning, learning, problem solving and planning.

In fact, there are many different strands of AI and it’s helpful to become familiar with the different approaches so you can better understand how to apply them to your business. AI can help us with simple decision-making, it can provide us with predictive analytics around big data, and it can also solve more complicated problems using pattern recognition.

Predictive AI

The most common type of AI is what is referred to as “weak” AI because it is focused on a narrow skill-set, rather than a “strong” AI focused on general intelligence. Narrow AI uses rules-based structures to help augment human decision-making. For example, you may be familiar with digital services that provide recommendations for the next movie to watch (Netflix), next song to listen to (Pandora), or next book to buy (Amazon). These recommendation engines plug in user-generated data (for example, online clicks or likes) into an AI algorithm, which then suggests something similar you may like based on your past experiences.

As we’ve generated exponential reams of data, AI can help identify hidden correlations and insights in big data using predictive analytics. Let me give an example. A couple years ago, a study at Stanford University used machine learning to scan millions of high-resolution satellite images to predict poverty in the African country of Uganda. The AI software was fed nighttime and daytime images of the same villages, and was taught to correlate lack of evening electricity with higher likelihood for poverty. The prediction level wasn’t perfect, but it was much faster and cheaper than undertaking a real demographic survey with experts going from village to village asking about poverty.

So if you have a large training set of information, you can use similar machine learning algorithms to train an AI to provide you with quick predictive analytics about your customers, products or market.

Machine Learning

Speaking of machine learning, what does this buzz term really mean? Machine Learning is an enabling technology that improves performance on a specific task by training a computer to explore and learn from different patterns. It can help us ask better questions with existing data; it can lead us to ask new questions we didn’t anticipate; and it can analyze new types of data, including audio or video. Let me give you an example.

A European electricity startup called OneWatt uses AI to predict when machines will break down by making sound “visible.” It does so by attaching non-invasive sensors to mechanical motors and collecting over 2 terabytes of acoustic data containing the sound of 16,000 faulty motors.  It then feeds this data to a machine learning algorithm, which has “learned” to recognize the sound of 8 leading motor faults, and can now accurately predict physical failure before it actually happens!

Eventually, machine learning will be built into anything that generates data, whether it’s physical or virtual.  Machine learning doesn’t have to match human expertise. We’re not automating experts. Rather, we’re asking it to “listen to all the customer service phone calls and find the angry ones”. Or, “read all the emails and find the anxious ones”, or “look at a hundred thousand photos and find the coolest people.” If you’ve got a large dataset of information in your business, you can use similar machine learning algorithms to identify outliers and address them to improve business performance.

Deep Learning

Taking AI beyond narrow predictive analytics, we can tackle more complicated human functions like cognition, reasoning and language, by employing deep learning neural networks that recognize complex patterns. Deep learning is a subset of machine learning, and uses layers of neural networks designed to mimic the pattern recognition function of the human brain.

The applications for deep learning are far and wide, from computer vision and facial recognition, to natural language processing, fraud detection, drug discovery, self-driving cars, and even voice-activated AI assistants like Apple’s Siri and Amazon’s Alexa that help you accomplish complex tasks like automating your home office, shopping online, or searching for information.

Beyond Super Intelligence?

We might fear the science fiction vision of a super intelligent AI that will dominate humanity, but the truth is, even the best AI algorithms can be unreliable when faced with situations that differ, even to a small degree, from what they have been trained on. Instead of fearing Skynet, we should really fear that we already trust AI algorithms too much, even though we aren’t fully aware of their limitations.

Consequently, every company should develop an AI strategy to make sure machine intelligence serves human goals. This means developing an AI code of ethics and an AI audit trail showing how coding decisions were made, and also providing a mechanism for remediation if the AI makes incorrect decisions and causes client mistakes.

In a future where business success will require humans to work side-by-side with intelligent machines, it’s imperative that every business leader understands the strengths and limitations of different AI approaches, and knows how to integrate AI tools into the company’s workflow. Your competitive advantage is at stake!

Kian Gohar is Founder & CEO of Geolab, an innovation advisory firm that helps companies develop digital transformation strategies. He is an expert on advanced technologies like AI and speaks regularly on innovation and moonshot thinking.