Dutch Summer of AI 2022: Where data science and AI students get connected with companies like NS, KPN among others

The Dutch Summer of AI 2022 has officially started. The event is a seven-week program for students to apply their skills in artificial intelligence, computer science and/or data science for some of the largest companies in the Netherlands, like KPN, Nederlandse Spoorwegen (NS) and ABN AMRO. Around 150 students applied for the project, but only 100 were accepted to face the challenges set by the various companies.

During the seven weeks program, the students work as a group to face challenges from participating organizations. They are monitored by experts and can put their AI knowledge into practice. Co-initiator Dennis de Reus, head of AI at ABN AMRO, kicked off the day at CIRCL, the new eco-friendly restaurant of ABN AMRO, with an introduction to the project. “The Dutch Summer of AI 2022 is an opportunity with AI to close the bridge between companies and students and alumni. The reason we started this project is that we noticed that there is a large gap between corporates and academia in The Netherlands. We want to connect these two groups at an earlier stage. The idea is to build a cool solution for a challenge with AI. We’ve set three goals for this project: strengthen de Dutch AI ecosystem, drive innovation with AI at companies and close the gap between academia and business.” Next to working with a team for a company like Tata Steel, the students can be awarded for their solutions. Various award categories exist, like ‘most humanity-centered AI solution’ and ‘best team/best pitch.’

Keynote speaker

After the introduction by Dennis de Reus, it was time for a keynote speaker, Mo Gawdat, former Chief Business Operator (CBO) at Google (X) and writer of various best-selling books like ‘Solve for Happy.’ Mo Gawdat was the CBO at Google (X) for about five years, and he focused on rebuilding the current technology and creating predictive innovation. “My task was to get innovation to be predictable, make sure that what we invented be useful for the world. Sadly, I lost my child in 2014 and started to become a writer. I wanted to shift my career to write about humanity. I believe that when you become human, you become happy.”

His primary motivation was not to have his son remembered but to build billions of digital sons and daughters. That’s why he wrote Scary Smart, a book about the future of AI and how people can save the world. He started the presentation with a passage from the book. “We’ll be sitting in front of a campfire in 2049, and I’m telling a story about what happens between 2021 and 2049.” The book begins with two parts, a scary and an intelligent part. “The scary part was so scary that I wasn’t sure to publish it. If we do the right things, AI is not scary, but if we do the wrong things, it’s terrifying.” Gawdat talked about how much AI everybody in the room already has been using since they woke up. “I know that we’ve infiltrated the world with AI, it’s everywhere, and AI is the first of the three inevitable.”

He continued: “AI will happen. There is no way that any human in the world can stop this. This is what we call the technology development curve, every technology we’ve developed goes through years and years of development. When we find the breakthrough, it’s an inverted hockey stick. AI will happen, and there’s no stopping it. However, this is not scary.”

The second inevitable is that AI will be more intelligent than us. Gawdat talked about how by 2049, there will be a machine more intelligent than humans. “When AI figures something out, it moves so fast that our time is useless anymore. There is a new era in which machines are number one, and we’re second. Tech doubles every 18 months, and by 2045 AI will be a billion times smarter than us. A billion times smarter is a comparison between a fly and Einstein. We are the fly, and the AI is Einstein. But does Einstein have a reason to squash the fly or not? When a new being is smarter than us, the world’s rules will change, and we don’t know what will happen. Can we flip the odds in our favour?”

Thirdly, Gawdat talked about the scariest inevitable. “Bad things will happen. You realize this when you write code and it doesn’t do what you want it to. However, there will never be a RoboCop or something similar. That’s because we’re too irrelevant for AI. But mind dystopia will happen. The minute you develop an AI smarter than you, it will only get smarter and smarter. AI can side with the bad and good guys. There are already cybercriminals that research AI, so we need cops to monitor this. If there will be an AI versus and AI, we won’t be interesting anymore.”

After talking about the inevitable, students were allowed to ask some questions about AI. One student asked: “Will AI have consciousness?” Mo Gawdat answered: “We’re in a prisoner’s dilemma to improve AI for our good. Humanity will not implement control solutions because an AI can do this better. If we try to control the AI, it will get smarter than us.”

Gadwat then focused on the use of AI in our current society and that the AI’s now being developed to sell (ads), gamble (trading), spy (surveillance) and kill (defence). “I’m here to say to all you students: don’t work in this field of AI, don’t do it. Please develop the right code. Realize that you as an AI developer are the father or mother of this AI.”

Case introductions

After Mo Gawdat’s presentation, it was time to introduce all the companies the students would work for. Firstly, Tata Steel. Stef Lommen, a data scientist at Tata Steel, talked about the surface quality of the steel they produce. Quality is important because the customer strives for excellent quality and no defects. Tata Steel wants to develop a post-classification model that would enable them to prevent surface defects. This goal will be reached through unsupervised learning and an automatic finetuning convolution neural network. The success of this model will be defined if high-quality products are ensured for their customers by classifying defects accurately.

After Tata Steel, it was time for South Pole, a company that develops emission reduction projects and strategies that turn climate action into long-term business opportunities for companies, governments, and organizations. Their challenge is as follows: “Can we at scale measure from satellite imagery how much carbon is sequestered in a forest with an error of a maximum of 10%?” The students must develop a proof-of-concept AI model that predicts sequestered carbon using only remote sensing data.

The third case study is one of Shell, one of The Netherlands’ most prominent companies. Tim Mulder, a data scientist at Shell, talked about the climate target of the company. “Our target is to become a net-zero emissions energy business by 2050.” Their challenge is to enable the next generation of clean energy technology with AI optimizing electric vehicle charging and hydrogen and wind farm layout optimization.

Nederlandse Spoorwegen (NS) was the next company to introduce its case to the students. Yorick Fredrix, a data engineer at NS, talked about how AI is implemented at NS, for example, with short-term train crowdedness and graffiti detection. Their goal is to predict the number of passengers between origins and destinations every half-hour during events. But the students also need to figure out what an event is, like festivals, sports events and holidays like Kingsday in April.

Anastasia Khomenko, a data scientist at KPN, talked about marketing automation at the company and how this enables the discovery of the best possible offer to make customers happy. Their case is about how to deal with their ‘cold-start’ problem. This is about all the different marketing campaigns that KPN offers and how they make it harder for customers to make the right choice.

The following company that introduced its case study was DSM, a purpose-led health, nutrition, and bioscience company. Patrick Attallah and Telli van der Lei talked about the goal of DSM. “The student team should create a machine learning model which, taking the available information (structured and unstructured) as input, plus additional information such as weather forecasts, etc., that produces a recommendation about the purchase of fish oil.”

Company.info, a software company that is part of FD Mediagroep with Het Financial Dagblad and BNR Nieuwsradio, talked about the AI they use daily. For example, they use AI for news relevance, sentiment and tagging. Their case is about extracting and linking relevant entities in legal texts.

Next up, it was Dulcy van der Werff’s turn; she is a data scientist at Aegon Nederland. Their case is as follows: “Price elasticity and conversion of deferred annuity customers.” They want the students to solve this by developing an AI model that predicts this conversion rate.

The last company is ABN AMRO. Will Chien, a data scientist, talked about how the company wants to identify and seize opportunities in solar panels for small and medium-sized enterprises in the Netherlands. The students are asked to identify eligible clients via structured and unstructured data, estimate business value by the potential of solar panels and create a dashboard that integrates all information to facilitate decision making.

2048 1152 Pepijn van Vugt

Pepijn van Vugt

Pepijn van Vugt is editor for ai.nl who specializes in data, machine learning and artificial intelligence.

My name is HAL 9000, how can I assist you?
This website uses cookies to ensure the best possible experience. By clicking accept, you agree to our use of cookies and similar technologies.
Privacy Policy