The AI train is moving fast - we have to start running now to catch it

Marco, you are Professor and Scientific Director at IDSIA, the Institute for Artificial Intelligence in Lugano. What is your main research focus at the institute?

IDSIA has a very broad range of research interests, spanning most of Artificial Intelligence as it is understood today: machine learning, including deep learning/neural networks, control and signal processing, natural language processing, robotics, computer vision, search and optimisation, and more fundamental questions in uncertainty, probability, statistics, causal inference.

To give an example, we have a 4-year Data project funded by the National Science Foundation as part of Switzerland’s National Research Programme 75 “Big Data”. In this project we deal with Gaussian processes, which can be understood as statistical neural networks, which can then provide uncertainty estimates relating to their own predictions – unlike traditional neural nets. This is very important in applications where we are evaluating risks. For example, a self-driving car needs to know whether the car’s sensors are reliably warning of a potential accident ahead rather than a a person safely crossing the street. Therefore Gaussian processes have definite advantages over traditional neural nets, but they also have one big disadvantage: they are far too slow to learn from big data. In this project we are designing powerful new algorithms for Gaussian processes that allow them to scale up to millions of data points far faster than was previously possible, opening up a huge number of potential applications for these tools. For instance, we are working at present with Meteosuisse, the Swiss meteorological body, on improving rain estimates in Switzerland by using Gaussian processes.

IDSIA is known worldwide for its Artificial Intelligence research. What different tech applications and innovations has IDSIA brought to the world?

IDSIA is very well known for its research on optimisation algorithms – for example, in routing, supply chains and scheduling – inspired by the behaviour of ants: so-called ant-colony optimisation. This research eventually also inspired swarm robotics, where very simple robots create complex swarm behaviour through indirect communication between them, in the way ants do. Technically this process is called stigmergy.

Other than this, we have a very long history of applied projects developed with companies and numerous special innovations have been developed while working on these projects.

A pretty fashionable topic at the moment is LSTM (long short-term memory), a type of neural network invented in 1997 in a joint collaboration between IDSIA and the Technical University of Munich. This is probably the most used type of neural network today in the world, with all the big players using it, such as Apple, Google, Amazon, etc.

What are the most important recent trends in Artificial Intelligence and where do you see growth potential for AI applications?

Natural language processing has been growing a lot recently on the back of the incredible progress achieved by the use of deep learning. I am talking for instance of OpenAI's GPT3. We see numerous requests to work in this field, especially by companies in the banking and insurance sector, and in general where texts and regulations abound. There is also increasing use of AI in Industry 4.0 and the internet of things, which is growing rapidly in manufacturing industries.

More generally speaking, I think all industrial sectors and companies should be growing dramatically at this time because of AI. If that is not happening, there's a problem and we ought to understand better where that problem lies to remove possible roadblocks. In other words, my point is that AI is already ready to help and make businesses grow. It's our fault if we haven't seized the opportunity yet.

Know-how and technology transfer is an important part of your work. That sounds very attractive to companies. How do you collaborate with Swiss and international corporations?

Yes, know-how does excite companies. We usually collaborate with them through Innosuisse, the Swiss innovation agency, which funds joint work between research centres and private companies. This offers a great opportunity for companies, especially those that cannot afford to invest in innovation through applied research. These projects last typically for a year and a half and can greatly help companies to innovate.

But there are also many companies that prefer to give us a direct mandate, to speed up the work, for example. One way or the other we typically have about 20 applied projects in progress at all times. UBS, Mastercard , Novartis, Roche, Georg Fischer and Bystronic are among the companies with which we have worked or are working.

Switzerland’s success as a business location has a lot to do with strong R&D capabilities and activities – in universities, research institutes and corporates. How would you evaluate Switzerland’s current position as a R&D knowledge hub compared to other countries?

I can give you many objective reasons why Switzerland is very well positioned:

• First and foremost, the government's annual investment in R&D as a percentage of GDP tops the world ranking;
• the quality of its polytechnics and the myriad high-profile research centres distributed all over the country;
• the big private sector players, such as Roche, Novartis, UBS and ABB, invest massively in research – so that Switzerland's number of patent applications per inhabitant also tops the world ranking;
• Switzerland's dense fabric of small and medium-sized enterprises that are best in class in their fields;
• Switzerland's two-track education system, one more academic, the other more applied, which is regularly praised worldwide as an example of a good education system.

But most of all, for me, Switzerland owes its success to its liberal and pragmatic mentality. Bureaucracy is kept to a minimum, the State trusts its citizens, innovation is welcome and favoured, no matter where it originates. I, for example, came here from Italy as a fresh PhD graduate, knowing nobody and not yet well positioned in worldwide research. And yet I have been given freedom to do my research, and given lots of research funding. I have been able to form a large group of researchers, and eventually become scientific director of IDSIA as well as a Professor.

What would you say could be done to increase Switzerland’s attractiveness as an AI research hub?

Sad to say, money definitely plays a role. Chief scientists in AI companies can earn several million dollars per year in the US. Postgraduate students, even without a PhD, can easily start work on a salary of $150,000. The average salary for researchers is around $350,000. Are these researchers worth that much? That’s debatable but if we want to have top AI scientists in Swiss companies we should consider the financial aspects, too, and not rely on the other benefits of being here.

We also need more aggressive, risk-taking investors. In fact, big investments are key to this process, not just public but also private investment. Look at the very big companies in AI: they are all either in the US (e.g., Amazon, Google, and Facebook) or in China (e.g., Baidu, Tencent, and Alibaba). There are none in Switzerland or Europe. My feeling is that we ought to be asking ourselves why.

My own answer to this question is that what's lacking is a bit of strategic policy-making (as well as vision) that encourages companies and centres to work together to promote strong business growth in this area. That would also involve developing some infrastructure, such as a well consolidated and aggressive investor chain; support and organisation from the Cantonal and Federal government; joint tables where government, academia and big players in the industry and also in the public sector, like the Federal railways, aviation, etc., can continuously discuss means to grow/invest/innovate together. And why not have a ministry for AI? They have had one up and running in the Emirates since 2017.

Can Switzerland be an R&D hub for AI?

There’s a big opportunity. But the AI train is moving fast and is easy to miss, despite our past successes. We have to start running now to catch it. The answer is not just to appoint new AI professors in a polytechnic. It's much more than that. The country’s system has to embrace change overall. You might ask why I am saying all this about AI and not some other field of innovation. The answer is that AI, unlike other innovation sectors, will pervade every other field. Global AI business revenues in the current decade are estimated at around 13 trillion dollars. We're talking about our future wealth and well-being.