The Skepticism and Optimism of Artificial Intelligence

In the last 30 days I've had many conversations with executives about cognitive, or artificial intelligence (AI) capabilities. I think we've made tremendous progress with our text analytics suite at Kingland over the last 7 years and it's fun to show people what we've done and where we're headed. The surprising thing to me, though, is the amount of skepticism I still hear from these leaders. When it comes to AI, there's almost a "too good to be true" cautiousness in the discussions. I think this is driven from the sheer volume of attention AI is getting as these leaders are asked daily to join the "AI Club". At the same time, though, I hear an incredible amount of optimism. Executives are thinking "maybe AI is one of the key investments I need to make to really make a difference." The efficiency, the automation, the chance to deliver something innovative for their company; all of these breed optimism.

I'd like to explore both sides of this skepticism / optimism coin. Let's look at one of the most popular questions I get:  "So...is that really AI"? 

It's a good question. What really is AI and why are we seeing so much AI attention right now? Understanding the history and context may help a bit. The first progress in AI started in the 1950s and started to pick up steam in the 1980s. In fact, many of the algorithms used by companies today are 10, 20, or 30 years old. At a high level, these AI algorithms are trying to simulate human decision making. The "artificial" comes from using a computer and the "intelligence" comes from understanding a decision. It's that simple. Anywhere you have people making decisions, there MAY be an opportunity for a computer to replicate that decision making process using AI.

A few weeks ago I wrote about how we like to focus our AI efforts on teaching our software to think, reason, and remember. Why? To make a decision. People that make sound decisions are viewed as "intelligent." They use their minds to remember what they've learned previously, to think through all of the different options, and then they search for the right answer. If you talk to computer scientists (which is always fun by the way), they'll explain to you that a great deal of what AI does is simply use a computer and all available data to "search" for potential answers and then "recognize" the best answer. So, if you have hundreds of people working on an operational process, making thousands of decisions daily, they are using intelligence to do their work and do it well. AI looks to automate that decision making. What's the optimal order for a series of transactions in banking or offers in retail? What's the most important paragraph in this 200 page paper? What was the sales charge for that particular product? Humans can reliably use their intelligence to make these decisions because the human brain is an incredible calculating machine - we're wired to "search" and "recognize" effectively for these answers. This is the "Intelligence" in AI. 

The "Artificial" is where the computing comes in. To simulate some aspects of the human brain, many of these AI algorithms have to perform millions, billions, or even more calculations in seconds. Years ago this was difficult because you had to go buy a large computer, or in the early days, have access to a super computer. The cloud has changed all of this. For companies that have invested in DevOps and the cloud, AI algorithms can now run across thousands or millions of compute nodes in seconds, or even milliseconds. This is why we're seeing such an explosion in AI, because we can now more easily and affordably scale. Cloud + AI = simulated human decision making.

Still, the optimism and skepticism exists. There are thousands of algorithms that have been created by brilliant mathematician and computer scientists. AI is exciting, but complicated. I was talking with one of the CDO leaders at a top 5 global bank last week and he said this. "The difference across the AI space is in context. Vendors with AI products and deep subject matter expertise in something like legal entity data make it significantly easier to realize the benefits from AI." This makes great sense. To simulate human decison making, you want to simulate based upon true expertise.

When you're finding yourself somewhat skeptical or overly optimistic, I would advise you to dive into the expertise. If the AI software looks good and demonstrates well, and your potential vendor has deep knowledge and real expertise, I'd be optimistic.

Visit our text analytics suite to learn more about our approach to AI.