It is hard to overstate the influence artificial intelligence has and will have in our lives. In 2020, the European Commission announced that being a “future-ready” business means being driven by innovative digital technologies. Being at the forefront of innovation is now a precondition for operating at all. Industry leaders end up setting the global pace for regulation, policy, and governance, while new technologies play an increasingly large and pivotal role in all aspects of our lives, from governance to commerce, healthcare to education, to art, culture, and philosophy.
With such high stakes, what does it take to bring Europe to the global forefront of AI? It’s not the first time European principles have had the potential to steer the course of a worldwide industrial and ideological transition. How can we ensure that democratic values like data security and user privacy are included in the development of these new technologies–technologies that are arguably already the foundation of our modern world? How do we stay ahead of the curve?
At ELISE, it’s all about connecting the right people. Our vision is for Europe to be a leader in AI developments, innovation, and regulation and today, being an industry leader means more than just being at the forefront of the latest technological developments, it means being the source of breakthroughs in AI. One of our key goals is therefore to support deeper connections between European industry and academia when it comes to research. By connecting and empowering Europe’s top researchers in AI and machine learning, we aim to facilitate outstanding, cutting-edge scientific research that drives change inspired by principles like explainability, transparency, and fairness, in all sectors of science and industry.
Industry-academic collaboration is not a new phenomenon. Companies often pair with universities to delve deeper into the science behind their products and stay at the forefront of developments in their sector. Academic researchers know that working with industry can also increase their capacity for research with access to datasets, funding, and real-life use cases. However, when it comes to research in AI, the line between industry and academia is increasingly blurred as industry turns its focus to foundational and theoretical research to accelerate commercialization. Symbiotic trade-offs like we’ve come to understand are not as straightforward, making collaborations more flexible, more agile, and less systematic. The key to a successful collaboration is not as obvious as it may have been even a decade ago.
To understand what makes a successful collaboration, we interviewed three ongoing partnerships in the ELISE network to see what set them up for success, and how their work has been shaped and enriched by the collaboration.
All three partnerships are part of the ELLIS PhD & Postdoc program, which supports excellent young researchers by connecting them to leading researchers and scientists across Europe. Each partnership is part of the program’s Industry Track, which is open to PhD students collaborating with an industry partner as a part of their degree. Typically, PhD students in this track have one ‘academic advisor’ from a European university, and another ‘industry advisor’ from a company based in Europe. Both advisors support the student in conducting cutting-edge research in AI and machine learning with the goal of publishing in top-tier conferences in the field.
As public-private partnerships like these become more prevalent across Europe, they have the potential to significantly impact both the quality of research conducted and the capacity and speed of scientific developments in AI and machine learning–a win-win scenario.
The ELLIS PhD & Postdoc Program supports young researchers in Europe with networking and training activities, it connects the top AI and Machine Learning talent in Europe and participants join a network of prestigious researchers in academia and industry.
There are differents benefits associated with this program:
Topic: Uncertainty quantification for machine learning system self-assesment in medical conversations.
PhD Student: Jakob Drachmann Havtorn.
Academic Advisors:
Industry Advisors: Lars Maaløe, Co-founder & CTO of Corti
Jakob Drachmann Havtorn is a PhD student in his second year at the Technical University of Denmark. He works on uncertainty quantification in machine learning for medical conversations, a topic he became interested in while working at Corti after his graduation. For Jakob it was important to pick a topic that is interesting for him but also as a project could be relevant for the company. Corti is a machine learning company, focusing on AI that aims to empower patient engagements. They are developing an AI-powered infrastructure that will listen in during patient consultations and remove mistakes when practitioners talk with their patients.
Once Jakob and Lars Maaløe, the Industry Advisor of the project decided on the topic, they took the initiative to write to a group of people in DTU about an industrial PHD project. Their call was answered by Jes Frellsen, associate professor at DTU. Jess was the external examiner / supervisor for the master’s thesis of Jakob, and when he saw there was a project with him he was very quick to say he wanted to be involved. As he said: “Jakob’s master thesis was one of the best I’ve ever read, so I was very happy to be involved”.
Lars, the industry supervisor, has a strong background in machine learning and a vivid interest in research. This shared interest between all parties has made the collaboration very smooth and fruitful. For Corti, being frontrunners of the research stack and focusing on a specific field gives them a competitive edge. Also the fact that the company has access to some data (e.g. recordings of phone call conversations) that are hard to obtain as an institute has made the project even more interesting for DTU.
Except for the common goal of impact-driven research that the two parties share, what made this collaboration possible was the Innovation Fund of Denmark for Industrial PhD projects. This fund allows startup companies and SME’s to get funding for research projects. The aim of this program is to create closer ties and collaboration between companies and universities and give a chance to young researchers to come closer to the private sector. Often we see that it is harder for startups or SME’s to invest in research and in collaborations with universities, but initiatives like the Industrial Researcher program helps with funding smaller companies to be able to invest in research and thus innovate.
- Lars Maaløe: “If we do not stay on top of the reasher stack we are left behind”
Topic: Methods for robust feature learning.
PhD Student: Sindy Löwe.
Academic Advisors: Max Welling, Professor & Research Chair at UvA
Industry Advisors: Maja Rudolph, Senior Research Scientist at Bosch Center for AI
The collaboration between UvA and Bosch is characterized by its longevity. The common goal of the two partner bodies (industry and university) is to develop a close collaboration on topics that are of interest for both. This collaboration is under the scheme of ICAI (the Dutch national Innovation Center for AI) and specifically the Delta Lab (public-private collaboration).
According to Max Welling, a professor and a research Chair at UvA and the academic supervisor in this collaboration, in the past the synergies between academia and industry were limited with the sponsorship of a PhD student from the industry that was most often already an employee of theirs. This meant that after the PhD was completed the research results were not always used by the company sponsoring the project and the researcher did not have the chance to continue building further their research.
Collaborations like the one between the University of Amsterdam UvA and Bosch are strong because they have managed to create a tight relationship where the two parties work together during the whole lifecycle of the project. The academic and the industry supervisors think hard together about the topics or the research direction of the project making sure that it is of interest to the industry as well. This way the industry coordinator is also more involved with the project, having monthly or even weekly meetings with the researcher and the academic supervisor to discuss the progress of the research. A new element that strengthens this relationship is that the industry that sponsors a Lab within ICAI has the opportunity to keep the IP of the developed research.
In this case Sindy Löwe, a PhD of the Delta Lab that uses notions from neuroscience to better machine learning algorithms, finds that this strong synergy gives her a lot of space to grow. She has biweekly meetings with her industry co-supervisor Maja Rudolph, Senior Research Scientist at Bosch Center for AI where she can discuss her issues but without a restriction or boundaries from the side of Bosch. According to Maja, even though the topic of Sindy is not motivated by a current project need, it gives the company the opportunity to be frontrunners of innovation and cutting edge research with application potential. The Uva-Bosch collaboration is continued with the launch of a new public-private research lab, the Delta Lab 2 focused on the use of artificial intelligence and machine learning for applications in computer vision, generative models and causal learning.
- Sindy Löwe: “There are no boundaries to my research from the side of Bosch”
Topic: Weakly-supervised multi-modal learning for scene understanding.
PhD Student: Antonín Vobecký.
Academic Advisors: Josef Sivic, Distinguished Research at CVUT
Industry Advisors:
The case of Antonín Vobecký is special since he is the second ever industry track phd student of ELLIS. With a vivid interest in computer vision for autonomous driving, during his master thesis he started his collaboration with Valeo.AI where there was an opportunity to join the Valeo.AI team in Paris and continue with a PhD project. What makes this project interesting is that Antonín is a PhD student in CKUT in Prague but he is also able to visit his team in Paris and work there.
A unique element of this partnership is its flexible nature. With three supervisors, Josef Sivic and Patrick Pérez focusing on strategy and David Hurych helping on daily problems, Antonin gets great input. Furthermore since Patrick is a researcher himself, the collaboration has been really smooth. For this collaboration to succeed, the different parties involved were looking for a topic that will serve as a win-win case for all.
Looking into the different issues that Valleo.AI is facing, and at the research interest of Antonín, they found that a topic that is centered around training the models in a way that we can use as few manual adaptations during the training as possible, would be of interest for both parties involved. Autonomous cars can record data (e.g images) that we later use. This is a cheap way to get access to data but on the other hand it can be really expensive for the companies to label.
With their research in this PhD project they try to see how they can use these data in an unsupervised way. This topic is an intersection on a really open and difficult scientific problem but has an impact addressing an important issue in Valeo.AI since it can save time and money from the labeling process.
To have fruitful public private partnerships with industry and academia you need an understanding that this partnership should benefit both parties and understand the value this exchange of knowledge can bring. According to Patrick, being able to interact with the engineers and the problems they raise and want to solve within the production could also be a source of interesting research problems. This is a notion that Josef also agrees with. For example one of his papers was inspired by his contact with industry. The open and even relationship that is created between the academic and industry supervisors, the shared “taste” and “curiosity” for research topics (in this case computer vision) and the flexibility to think together and plan are the ingredients for a successful partnership that can lead to rapid innovation.
- Josef Sivic: "I think in this case an important factor of success is the flexibility"
Jiayi Shen, Xingiun Du and Xiao Zehao are three students that are in the third year of their PhD and they are exploring the topics of multi-task learning, meta-learning and domain generalisation. Before their PhD all of them had an experience of doing internships with industry and what they valued the most from these experiences was the possibility of working on applicable topics. Thus, when they decided to continue with their education and do a PhD, an important element for them was to be within an institute and work on a project that gives them still the possibility to have a close contact with industry.
Being in the AIM Lab and having an industry coordinator gave them access to monthly discussions between the research group and the Inception Institute of Artificial Intelligence. Especially since all of them started their PhD during the Covid pandemic, being part of ELLIS gave them an extra opportunity to be in closer contact with the industry and the developments outside of academia even during the quarantine period. As they mentioned, the best experience so far was the ELLIS Doctoral Symposium that took place in Alicante. There they had their first opportunity to meet in person different companies from all over the EU, working on ML as well as discuss with ELLIS students about their topics and issues.
Networking with industry and understanding what are the prominent fields in the market gives them a competitive edge during their research.
When asked what is the best element of been an ELLIS Industry track student the gave us the following points:
All three of them aim to do an internship within an industry in the coming years of their PhD projects. According to Xiao, having the chance to work some months with industry will give them an even better overview of current trends and issues but also the opportunity to work on more applied projects. Some of them see their future within academia and some in the industry but all three of them believe that the ELLIS students are part of a strong network that potentially will impact and shape for the best the field of AI!
