Ongoing Evolution of AI

The holidays are a time to reflect on the past year and think about what lies ahead in the new year. In 2018, AI continued to get a lot of attention in the public eye. There was the usual hype, but there was also good media coverage of the industry’s challenges and progress. It is always good to remind ourselves that AI, like other technologies, evolves continuously. Hard-working researchers and innovators will continue to advance the state of the art, but the advancements tend to result in evolutionary – not revolutionary – change.

It seemed like there was less “rise of the robots” or “the Singularity is near” hype in the past year. Perhaps there is more understanding of the complexities of AI, and an appreciation that the dream of “artificial general intelligence” is extremely distant. Perhaps some of the visible setbacks in autonomous vehicle projects helped illustrate the complexities of AI.

On the bright side, we have continued to see growth in a broad range of useful AI applications. If you missed the report from McKinsey Global Institute this past year that cataloged AI use cases across 19 industries, I highly recommend that you read it. This paper covered more than 400 use cases across all major industries. The estimated economic value of these applications in is in the trillions of dollars.

When you look at this wide range of use cases, you see that AI can become very good at specific, narrowly defined tasks. Availability of training data is of course one of the key factors. We are all familiar with voice assistants from Apple, Google and other large players. These benefit from the enormous amount of voice data available for training and the ability to continue learning based on continuous user feedback. We’ve seen AI become a part of our daily lives by enabling voice assistants, recommending movies, and helping automate routine tasks at work. But when should we expect to see AI actually eliminating jobs? That will be a slow, gradual process. And I expect that new jobs will be created by AI along the way.

What lies ahead in 2019? We will certainly see further broadening of AI’s impact across industries and different aspects of our lives. I predict that as AI touches our everyday lives in more specific ways, people will better appreciate that AI is effective in very specific tasks, not in complex reasoning or social interactions that people are good at.

I also predict that we will see more “edge of the network” use cases appear in 2019. Many of today’s familiar AI use cases run just fine in the cloud. But for low-latency applications, for example controlling fast-moving robots, you need AI intelligence at the edge. The appearance of power-efficient AI chips for embedded applications will enable AI at the edge, where real-world problems require real-time processing.

Thanks for reading my blog, and I wish you a happy 2019!