Without even thinking about it, we are using Artificial Intelligence technologies daily through smart search functions in our mobiles, personal assistants like Siri, and when using translation tools on our computers. This is just the beginning; new major commercial applications are just around the corner. So, does AI mean that we can sit back and relax? No! It’s the combination of man and machine that will deliver the best result, says AI researcher Fredrik Heintz.
While AI research dates back to the fifties, it’s only recently that we have seen major breakthroughs and widespread adoption of AI technologies. Why now? Because we now have all the fundamental components, according to Fredrik Heintz who conducts research in AI and autonomous systems. As a leading researcher in AI, Associate Professor of Computer Science at Linköping University, Director of the Wallenberg Autonomous Systems and Software Program (WASP) Graduate School, and President of the Swedish AI Society, he should know.
“Software has become a dominant factor with the digital and analog worlds merging,” says Fredrik Heintz. “We also have sufficient computing power and large quantities of labeled data, together with improved algorithms thanks to research in academia and industry. But while AI of today is making huge progress, and solves real problems, there are still many unsolved problems.”
While AI is good at learning single tasks, like classifying images, understanding speech or text, other tasks are more difficult. Machine learning is still brittle, AI can only find correlations and not causations, it gives opaque answers that lack explanations, and requires vast amounts of data and extensive manual labor. Someone still needs to curate the data, select and adapt algorithms and train the system. It is also difficult to test and validate; how do we know that the system will make the right decision in previously unseen situations?
However, AI’s limitations are not slowing down progress. AI is actively used in a variety of domains and in several impressive applications. IBM Watson beat the Jeopardy champion already in 2011, and Google DeepMind won over the South Korean Go champion in early 2016. Google’s self-driving cars are close to being commercialized, Ford has announced availability of autonomous cars for ride sharing in 2021, and intelligent robots are replacing more and more tasks.
What will the future of AI look like? Fredrik Heintz says there will be a gradual automation or smartification of different components, where we will see AI enabled systems and AI augmented humans. It’s important to remember that we are automating tasks, not jobs.
“I don’t believe it’s either or; it’s through the combination of man and machine you will achieve the best results,” says Fredrik Heintz. “It’s like with chess – humans have no chance against chess computers, but the quality of chess playing has increased since people started practicing against AI players, and teams consisting of humans and chess computers are better than both humans alone and chess computers. We clearly see that humans and computers complement each other.”