Designers who become even more creative with a robot in their team, using Twitter to predict the stock market, and catching a single bacterium in the act of infecting a cell – these are all examples of how artificial intelligence has penetrated every corner of science in Zuid-Holland. Three researchers talk about the diversity of the work they do using AI.
Algorithm helps design teams work together more effectively
TU Delft | Computer scientist Catharine Oertel, Designing Intelligence Lab
Computer scientist Catharine Oertel is studying how an artificial system such as a robot can strengthen a team. “It will need to hear everything that is said but only remember the important bits, and perceive people’s movements and what they look at. It will need to be familiar with the relevant scenarios and be able to assess all important social and cognitive processes in a team against these scenarios. It will ultimately need to be able to influence the way the team cooperates in a positive way”, she says.
These are quite some demands. “Our holy grail is to achieve this degree of interaction with several people simultaneously and over a longer period of time”, Oertel continues. She has come to Delft for the coming five years to continue her search for this holy grail in the world of industrial design. Whereas designers used to be commissioned to simply design the best streamlined aircraft for the best price, today they also have to take account of such aspects as the working conditions in the mines where the raw materials are extracted and the threat of those raw materials becoming depleted.
How designers think
TU Delft’s strong reputation in the fields of industrial design and artificial intelligence come together in the Designing Intelligence Lab, one of the 24 TU Delft AI Labs currently being established at the University. Each AI Lab in Delft has two principal investigators. For the coming five years, Oertel will be cooperating with Senthil Chandrasegaran and be assisted by four PhD candidates in the Designing Intelligence Lab.
Oertel does not expect to have created a fully-fledged system that will give design teams wings within these five years. “But I do think we will make important steps towards the development of a real-time visual analytical system that will be a valuable addition to the AI toolkit of designers and other teams”, she says.
Competing in the AI age
Rotterdam School of Management, Erasmus University (RSM) | Ting Li, Professor of Digital Business
Professor of Digital Business Ting Li studies how companies can benefit from data at the Erasmus Centre for Data Analytics. You probably don’t realise it, but Google gets smarter with every search you make. “It’s a learning algorithm and it’s becoming more and more precise. The more often we click on a page, the higher that page will appear in the list of search results”, she explains.
AI is everywhere, says Li. It’s in cars and banking systems, in dating sites and all over the media. Companies often use recommendation systems – a form of AI – to recommend products, such as movies on Netflix, music on Spotify, and all manner of items on Amazon. “You might think that you never click on those recommendations, but 30 to 35% of Amazon’s sales come from smart recommendation systems that are increasingly being fine-tuned by their data engineers”, she says.
One of Li’s PhD students is helping an insurance company to understand why customers switch from the one insurer to another. “We analyse the data to find out whether customers are mainly motivated to switch by comparison sites such as Independer, by Google search ads, by links on other websites, or after being called by a sales representative”, Li continues.
Stock market forecasting with Twitter
In another project, Li is investigating whether you can predict the stock market using Twitter or Facebook. “Algorithms can browse through those media and distil sentiments. For example, based on the sentiment ‘Apple did well in the first quarter’ you might be able to predict how their share price will go”, she says. It’s not certain whether this will work yet; it could also be that the sentiment only becomes more positive after the share price has risen.
It is not just about making money. “You could also improve your service based on such data and AI systems. For example, I work with public transport companies who want to improve their fare structure based on travel patterns, or ensure that passengers do not have to stand.” Li describes the core of her research as follows: “I’m always looking to understand the relevance of information, be it for the benefit of companies, individuals or society as a whole.”
AI literally opens new worlds for the life sciences
Ariane Briegel, Professor of Ultrastructure Biology
Bacteria caught red-handed, deeply frozen just as they were about to cause Lyme’s disease. Professor Ariane Briegel is wildly enthusiastic about the wonders she observes thanks to three elements: a freezing technique, a camera-equipped microscope, and AI. ‘It’s fascinating. Every single cell is different.’
Bacteria follow their ‘nose’ towards a food source. Well, strictly speaking not their nose, but their receptor proteins. Neatly ordered like a honeycomb, Briegel has watched them prick through cell walls. On the outside they catch signals; on the inside they set the cell to work swimming in the right direction.
From 100 slices to 3D images with smart software
The camera in Briegel’s microscope takes approximately one hundred photographs of the same number of cell slices. Smart software then turns these hundred photographs into an understandable 3D image. Daar zit een grote verdienste van kunstmatige intelligentie. Een mens kan niets met honderden foto's van celplakjes. Dimensie-reductietechnieken uit de AI-wereld kunnen zulke data vertalen tot een beeld dat wij begrijpen. This has literally opened a new world for Briegel and her colleagues: they can observe in great detail what cell structures and organs look like and how bacteria ‘get moving’.
And the fun has only just begun. ‘When I completed my PhD twenty years ago, I was producing two datasets a day at most. These days it’s forty to fifty a day, far more than researchers could ever look at. And we want to tackle even larger projects.’ Like mapping the 1.5 kilograms of bacteria on and inside the human body. Infection, cancer, growth: with these techniques, we will soon be able to map in detail so many biological systems in the life sciences.'
Read more at universiteitleiden.nl/ai