Artificial intelligence becomes part of daily lifeTUT's long-standing lecturer Heikki Huttunen took up an associate professorship in the Laboratory of Signal Processing in August 2017. His research interests focus on machine learning, which is evolving at a breakneck pace.
Back in the day when Heikki Huttunen first enrolled at the University of Tampere, scientists were exploring the challenges of character recognition using neural networks. Later they moved on to more complex tasks and, for example, taught computers to tell the difference between a cat and a dog in a photo. This is all in a day’s work for students of today, which shows how far machine learning has come.
“Now one of the hot topics in machine learning is video recognition that allows computers to recognize scenes in a video. This is much more complex than image recognition,” says Huttunen.
Another current topic is visual regression. How can a computer look at a photo and tell how many people it sees or how old they are? Researchers at TUT have also developed a live age recognition demo that habitually attracts curious and amused crowds at fairs and exhibitions. It assesses the age of people standing in front of a screen.
“It already works pretty well, even though people try to trick the system by pulling faces and fooling around. It is not easy for a computer to detect a person's age, because the years show on our faces differently. Age recognition works with clear photos, but the success rate depends on lighting and image quality,” Huttunen says.
A special focus of his research is system deployment, or turning ideas into practical and functional everyday solutions. He has a long history of working in the corporate world alongside his academic career. He has worked, for example, at Visy Oy – a company developing technologies that automatically recognize license plates.
“Visy's license place recognition systems are deployed in ports all over Europe”.
Human-level performance has already been achieved
A long-time goal of AI scientists was to achieve human-level performance, whereby a computer could perform a task as well as people. We are already past that milestone. Where is machine learning heading now?
“Current artificial intelligence is so-called weak AI. It can equal or even surpass human performance at narrow tasks, such as recognizing traffic signs or playing chess. Weak AI cannot generalize learning between activities. Strong AI is still in its infancy, but it will be equivalent to human intelligence.”
“A child can look at foreign traffic signs and deduce their likely meaning, even though the signs look different than the ones at home. AI cannot do this – at least not yet,” explains Huttunen.
The democratization of artificial intelligence is spurring research forward. To a surprising degree, code is no longer locked in a safe but openly shared. Even corporate giants, such as Google and Facebook, have made codes openly accessible. The academic community also embraces open access.
“In principle, anyone can do what the major corporations are doing and develop their solutions further. Software and publications are openly accessible. You can purchase or rent computing resources at a relatively low cost. Skills and good ideas are the things that matter.”
Heikki Huttunen (born in Jyväskylä, Finland, in 1971)
Master of Science, 1995, University of Tampere
Doctor of Science in Technology, 1999, Tampere University of Technology
Assistant Professor, TUT, 1999-2003
Machine Learning Scientist, Visy Oy, 2003–2005
Lecturer, TUT, 2005-2017
Family and hobbies:
E-biking, badminton and fishing
Wife and six-year-old son
Growing interest from the insurance industry
The tenure track professorship offers Heikki Huttunen more freedom to conduct research and expand his research group. Huttunen will now work full-time at TUT.
“Or let’s say that I work 110 per cent: 100 per cent at TUT and a little in a company. It is a win-win situation: my first-hand experience of business and customer needs will keep my feet as a scientist firmly on the ground, and I will be in a position to quickly transfer the latest knowledge to industry. Students are also very interested in industry experiences,” Huttunen says.
The newly appointed Associate Professor is not planning to give up teaching, which he enjoys. In the next academic year, he will continue to teach at least the introductory course on signal processing and an intermediate course on machine learning.
“Our student numbers have doubled each year. Last year we offered a machine learning course at the University of Tampere. This kind of collaboration provides an avenue for fruitful dialogue between students from different fields, such as statistical sciences, economics and actuarial science.”
Huttunen is actively involved in collaboration within Tampere3. He serves on TUT’s Academic Board and Faculty Council and has a front-row seat to the development of the Tampere3 higher education community.
“It’s a great opportunity to contribute to both the substance and environment of Tampere3. It is essential that we bring the Tampere3 project to a successful completion. The synergy potential is substantial, and the new higher education institution will draw talented students. Students will not have to decide what to do with their lives at the age of 18. Tampere3 will help them keep their options open longer.”