Artificial intelligence is becoming more and more powerful. The world’s leading expert in the field of deep learning, Prof. Yoshua Bengio, talks about the opportunities and risks of the technology. What potential does AI have for science and society? And how can we develop responsible use of artificial intelligence?
Bringing together over 1,000 machine learning researchers, the Mila Research Institute in Montreal, Canada is recognized worldwide for its significant contributions to deep learning. What is behind Mila’s great success?
Prof. Yoshua Bengio: When the Mila was in its earlier form at the University of Montreal, there were a couple of deep learning professors there. That was pretty special because there were very few people doing deep learning at the time. So we had a critical mass on a very specific topic. From my point of view, that was a decisive factor. In addition, we published some of the first works that would prove to be very formative.
Scientists in Canada have benefited from the intensive funding of AI research and the strategy for artificial intelligence there. Was that another factor that contributed to the Mila’s success?
Definitive. Both the Quebec government and the Canadian federal government have invested disproportionately in machine learning and AI compared to other sciences and fields. That was a political decision that the infrastructure that we have here, the research chairs and their extremely generous funding for professors made possible in the first place. We have succeeded in recruiting some of the best researchers for professorships.
In addition, the institutes are networked with society – with startups and companies, and that is an important part of our mission. We are in contact with governments and intergovernmental organizations from all over the world. And we are also investing in the question of how we can ensure that AI actually benefits society and is not misused.
What have you done to develop a responsible approach to AI?
We have in our ranks not only computer scientists, but also sociologists, philosophers and legal experts, because these technologies are already very powerful, and what scares me even more is the fact that they will be even more powerful in the future.
I believe that even in our countries, the economic, political and social context is not yet ripe for the potentially far-reaching negative effects of these technologies, for example in the military sphere or in controlling people.
We need to think hard about how we can develop technologies that are more compatible with human values and human rights. We must also point out the dangers to the media, citizens and politicians. In addition, we need to invest in applications that are potentially game-changing and could have a positive impact on humanity in the process.
One danger is certainly the manipulation of people with the help of AI.
Exactly. When combined with powerful AI, companies like Facebook and Google have the potential to manipulate us through personalized advertising. Manipulating people to buy one brand and not another brand may not be as serious, but it could also influence our political attitudes. And that’s really scary. How many of these are actually already there? I dont know.
I’m sure many companies are trying to push the technology in this direction. It’s good that governments around the world are trying to create rules of the game to make at least some of these things illegal.
To support this effort, get involved with an organization called The Global Partnership on AI.
This is an organization that is affiliated with the OECD but includes a larger group of countries that want to cooperate on such issues as responsible AI. We make recommendations to governments on how to reduce the potential threats, including from social media and the like. On the other hand, we also make suggestions on how we can get maximum benefit from AI, for example in the areas of environment and health.
Is there a global challenge where you would like to make a difference yourself?
Yes, namely the increasing problem of so-called antimicrobial resistance. We are already in a crisis. There are currently 1.2 million deaths per year. And that number will increase to 10 million deaths a year by 2050. The problem will cost an estimated $100 trillion by 2050 by crippling our healthcare systems. And the industry isn’t really working hard on it because it’s currently considered unprofitable.
How could AI do something about it?
For example, by helping to speed up the discovery of new antibiotics and reduce costs. Or to examine the area of antibiotics, which is not even considered in the current procedures. We must be able to produce new antibiotics faster than the pathogens mutate and become dangerous. This is a war between us and them.
In which other areas do you see concrete opportunities to solve major challenges with the help of AI?
There are so many. A few years ago we wrote an article on climate change and AI that just appeared in a trade journal. This one-hundred-page article provides an overview of all work on the diverse use of machine learning, primarily to contribute to climate protection, but also to climate change adaptation.
For example energy management, development of new materials, CO2 capture, methods to reduce our consumption, climate models, better modelling, how species are changing with regard to climate, how we can use nature to absorb carbon that is already in the ground is to be deposited or not released.
Health is an even bigger area. Humanity needs more minds that think about these challenges and opportunities and try out different ideas. It’s almost a matter of survival. Apart from all the things we could achieve, our work also makes us feel good. It motivates students and researchers. It shows us that we are all pulling together. It gives our work so much more meaning.
Read more: You can find the entire interview at helmholtz-hida.de.