Imagining the future of artificial intelligence

Join us for a workshop June 5 on the future of artificial intelligence, at the annual Science|Business Framework Programme conference in Brussels.

Google’s Deep Mind software took just 40 days to become the best ever player of the ancient game of Go, and commentators heralded it as a major milestone for deep learning, a field of artificial intelligence (AI). The achievement highlighted how computers equipped with the right algorithms can now quickly teach themselves to achieve a specific goal.

This approach works well within a well-defined set of parameters, such as a board game. But what about when something unexpected happens, such as a change in the rules of the game?

In the work environment, unexpected things happen all the time, limiting the usefulness of today’s AI systems. But The Curious AI Company, founded in Finland in 2015, is seeking to address this challenge. “The AI we have today is unable to handle new situations: It is unable to analyse what is happening and the reasons.” says Harri Valpola, CEO at Curious AI, which has been supported by Business Finland and raised $3.7 million in a second funding round in September 2017.

“We want to build a human-like AI that can solve problems and that can respond to something unexpected,” he says.

‘A computer that can imagine’

Curious AI is doing this by combining multiple AI systems to create digital co-workers that can interact with human beings in a natural manner. “We are looking for a computer that can imagine,” Valpola says. “We are taking a lot of components and building something bigger. We are using both supervised learning and unsupervised learning to create a cognitive architecture.”  Valpola describes this kind of general AI as “a lofty goal”, noting that “narrow AI is all the rage”.

Narrow AI that can perform tightly defined tasks, such as identifying an animal, is being widely deployed thanks to advances in processor technologies and the vast amount of digital data now being collected by Google, Amazon, Facebook, Baidu and other Internet players. Deep learning systems can use this big data to hone speech recognition, text translation and facial recognition software because there are so many samples available and the rules don’t change. But with an evolving industrial process and many other forms of knowledge working, historic data may not be relevant or only partially relevant.

Although fears are growing that AI will make many jobs redundant, it is also conceivable that digital co-workers will enable people to perform roles that were previously beyond their capabilities. Automated assistants could also increase individuals’ productivity, enabling them to work shorter days, while benefiting the economy as whole. Consultants say demand for so-called robotic process automation systems, which have been around for a decade or so, is growing rapidly.

Pilot projects

After two years developing its core technology, The Curious AI Company now has two pilot projects underway. One is to build a solution that can understand what is in a document. The other is to create a system that can help a human operator manage industrial process controls, such as those in a chemical plant with hundreds of components generating thousands of measurements every day. “We plan to get them into the market and then evolve them as we find out what problems need solving,” Valpola says.  “We can create computer-human teams that are able to communicate efficiently, so they can ask each other: What do you mean? Or what would happen if we did this?”

In the case of a complex chemical plant, Curious AI’s system would observe the processes and learn how they work, so it could help the human operator optimise the plant’s performance. As the processes evolve over time and there is limited historical data on which the AI can learn, the solution needs to be able to adapt to change. “It would not be enough to collect 20 years of historical process data because so much would have changed in that time and so much would change in future,” notes Valpola.

Curious AI is the second company Valpola has co-founded. He helped set up ZenRobotics, which develops brain-inspired artificial intelligence for robots, in 2007 following 10 years at Aalto University teaching and studying artificial brains. Zen Robotics is now focused on robotic waste separation.

Having conducted extensive research in academia, Valpola is convinced general AI needs to be developed and refined through commercial deployments designed to solve real world problems, rather than theoretical challenges. “We see it as very important to build something functioning early on,” he says. “General AI is not going to come out of some research lab, it will emerge from a long succession of increasingly complex systems.”