For physicists, artificial intelligence is useful for classifying data and cleaning up distorted signals

Students won’t go wrong if they focus on artificial intelligence, says Edward Hæggström, professor of material physics.

Did you know that your dentist’s drill could soon be using artificial intelligence? Such a drill would be able to increase speed when encountering harder bone and slow down when working on a softer area. This way the dentist would not drill too deep by accident.

A device like this is not science fiction – it could easily be created through existing innovations in physics and programming. It would comprise a sensor, the device itself and artificial intelligence software operating in the background. The software would control the drill and decide what commands to send to it.

 “As physicists, we are specialised in measuring things and understanding the world through our measurement tools. In practice, all measuring instruments these days contain a software component that could take advantage of artificial intelligence,” says Edward Hæggström, professor of material physics.

For physicists to be able to develop ever-improving devices, they need help from software experts. The University is an excellent environment for developing ideas, as the experts are close at hand. For Hæggström, cooperation with both computer scientists and statisticians is part of just another workday.

 “I’m so old-fashioned that I actually walk up to people and tell them that I’ve come up with this solution, have you thought of anything better? Is there something that is still experimental but always works and may be more effective than existing methods?”

Cleaning pipes with ultrasound and artificial intelligence

One of Hæggström’s projects is Altum Technologies, a company based on University research, where he is a shareholder. The company makes a device that can clean industrial equipment with ultrasound. Sensors attached to the outside of pipes clear them of fouling with ultrasound and do not require that production be stopped during the cleaning process.

The cleaning method can be adjusted through artificial intelligence. With the help of AI, the ultrasound sensors need not be placed within millimetres of the correct location, which makes work easier. It may be difficult to precisely install sensors deep inside industrial equipment, as the relevant pipes may be in difficult-to-reach places, and there is always the chance of human error in installation. The artificial intelligence can recognise where the sensors are located and can optimise the cleaning process accordingly.

 “Artificial intelligence can also tell when the pipes no longer have to be cleaned, which saves energy,” Hæggström points out.

The method has been received with great interest, and in August, the company announced its expansion to the United States.

Artificial intelligence to clean and classify

The pipe cleaning process is a good example of how Hæggström thinks about artificial intelligence.

 “I’m fairly pragmatic by nature. Artificial intelligence is absolutely something worth researching and exploring. But we can’t let ourselves think that it’s all-powerful: it’s just a tool that can give certain things an edge.”

Personally, he uses artificial intelligence to complete two tasks: to filter noise from signals and to classify things.

It is sometimes necessary to filter the signals gleaned from different measurements before conclusions can be drawn from them. Software with artificial intelligence can try out different ways of processing the signal and determine when the signal has been optimally filtered. This will make it easier for physicists to develop equipment that can optimise itself.

Artificial intelligence can also help with determining how many classes a particular phenomenon should be categorised into. This is how artificial intelligence was used in the system designed to determine whether bus drivers and similar professionals are too tired to keep driving safely. Hæggström participated in the system’s development.  

“Philosophically speaking, I see artificial intelligence in terms of these kinds of everyday examples. Similar ideas can be used in cancer diagnostics or digital microscopy,” Haeggström explains.

We must understand algorithms even if we do not write them

According to Haeggström, physicists should learn to understand where artificial intelligence is useful and where it is not. Haeggström does not believe AI is appropriate in all contexts.

 “For example, if using artificial intelligence comes with a risk of a stock market crash or a similar disaster, we should be very cautious about using it. If unexpected and severe consequences are unlikely, artificial intelligence methods are usually very effective.”

According to Haeggström, it can be quite difficult to fully comprehend the workings of artificial intelligence, but it is important – particularly for physicists – to try to gain some understanding of it.

 “I am of the school of thinking that a physicist must always fully understand how the measuring instrument works. Otherwise we cannot be certain that the result is correct. The same goes for artificial intelligence. Physicists need not write the algorithms themselves, but they do have to know to a significant degree how they work and how they could accidentally be misused.”

Haeggström has one more tip for students:

“If students are pondering their specialisation, I would say they can’t go too wrong with artificial intelligence, no matter if their main interest lies in physics, chemistry or another discipline entirely.”

Edward Hæggström
  • Professor of material physics at the University of Helsinki’s Department of Physics.
  • PhD 1998, University of Helsinki, major in physics
  • Several academic elected positions, including chair of the University of Helsinki chapter of the Finnish Union of University Professors
  • Developed a nanonization technique that enhances the absorption of medical compounds.
  • Participated in the development of an electric space sail, a tool for measuring tiredness based on balance, monitoring methods for osteoporosis and a needle-free blood sugar measuring tool. Currently working on a new kind of eye pressure monitor.


University of Helsinki developing methods for artificial intelligence

The University of Helsinki is developing new methods for artificial intelligence, machine learning and data mining, which several different research groups are applying extensively for varying purposes. In this series, we highlight individual researchers to explain the research in the field and the ways in which it is impacting our lives.

Other instalments in the series have discussed the role of algorithms in drug development and how algorithmic bias can be minimised