AI’s third wave: Less science fiction, more focus on customer problems

The futurist Roy Amara is best remembered for an adage he coined, which became known as Amara’s Law: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

The intermodal shipping container is as good an example as any. The first international standard for containers was established in 1933, and they were aggressively promoted by Malcom McLean, in particular, starting in 1955. But containers didn’t start to really revolutionize shipping until the 1970s and didn’t play a pivotal role in helping create a truly globalized economy until even later.

What I saw about artificial intelligence at a recent CloudExpo conference and other events helped confirm my belief that technologists and users are starting to put earlier overestimates behind them while settling in for the long haul.

As CloudExpo conference chair Roger Strukhoff put it, this is the third wave of AI, and “this one is more practical. In the past, AI was trying to boil the ocean. Today, AI is very focused – much more focused on specific problems.” To Strukhoff’s comments, I’d add that many seasoned technologists are stepping back and reassessing breathless hype around difficult and multi-faceted problems such as self-driving cars.

Progress will certainly continue and we’ll see more autonomy in limited domains. But urban robo-taxis are probably many years out.

If that sounds like pessimism, though, it’s not!

AI gets practical

What’s happening with AI today is exciting in part because it involves practical solutions that address complexity, the need to handle more and more data, and demanding customers. McKinsey says, “75 percent of online customers expect help now.”

Strukhoff highlighted how companies are struggling with operational experience (OX). “Service agents are overworked with repetitive work,” he said, adding, “IT/Cloud/SREs involve very hard-to-triage issues.” Part of the problem is that cloud computing and DevOps revolutions notwithstanding, “a lot of IT is still very manual, people-driven, and lacks automation.”

In response to these problems, Strukhoff identified three strategic CIO priorities: business growth through digital transformation, offering a great experience to employees and customers, and replacing legacy with AI and cloud innovations.

What does this look like with respect to AI? Strukhoff quoted a Gartner stat to give a flavor: “By 2022, 40 percent of large enterprises will combine AI & ML to automate and replace ITOPS and ITSM processes and tasks, up from 5 percent today.”

[ Read also: AI vs. machine learning: What’s the difference? ]

Will we ever get those flying cars?

But perhaps you’re thinking that all sounds rather mundane. Or you’re thinking about the rather limited abilities of the chatbots or virtual assistants that use machine learning to react to your queries. Where’s the transformative future on offer?

Some of it you can glimpse if you squint and extrapolate. Amazon’s Echo quickly gets confused if you veer far from the fairly limited set of commands it understands. But it’s not really too much of a stretch to imagine future generations of today’s devices that could handle most routine tasks involving interactions with other machines.

For example, booking travel arrangements for a trip requires knowledge of preferences, context, access to data sources, and a variety of things that make it complicated for a bot (or an inexperienced person). But there’s nothing fundamental about such a task that would lead us to think it might well remain in the realm of science fiction indefinitely. Fully autonomous vehicles are probably similar.

These advances will take time. Probably longer than our intuitions think they “should.” But they’ll happen.

AI is much more than machine learning

Complicating our future-gazing is that much of what we call AI is the fairly narrow slice of machine learning that’s delivered such remarkable advances in the past few years. There are swaths of study in linguistics, cognitive science, and elsewhere that have been slower to advance than those largely driven by computational horsepower. It may turn out that some of those related fields may be important for solving certain types of AI problems.

To return to Roy Amara, though, linear extrapolation isn’t really the point. What we’re really building today is a foundation for the future. The power is in the connection to data, the refinement of algorithms, and the development of new techniques that will create possibilities that are hard to imagine today.

[ Get real-world lessons learned from CIOs in the new HBR Analytic Services report, An Executive's Guide to Real-World AI. ]