"I want to build bridges between people and machines"
The ongoing development of artificial intelligence is often presented as a race between humans and machines. ETH professor Menna El-Assady takes a different approach: as head of the Interactive Visualization and Intelligence Augmentation Lab at ETH Zurich, the young scientist wants to develop an AI that can be used interactively and that shapes its capabilities only in collaboration with humans.
It doesn't look good. When the world's powerful gathered in Davos at the beginning of the year for the World Economic Forum (WEF), they were faced with gloomy predictions for the future. The latest edition of the "Global Risk Report", compiled by the organizers of the forum, points to a new danger in addition to issues that have been worrying politicians and business leaders for years - climate change, armed conflicts, uncertain economic development: rapid technological change.
The technology referred to here is computer technology, or more precisely, artificial intelligence (AI). It therefore comes as no surprise that AI was one of the most discussed topics at the WEF 2024 in Davos. Numerous panel discussions, lectures and media conferences revolved around this key field of software technology. For many media representatives and WEF participants, the focus of interest was on a young American: Sam Altmann. He is the CEO of the US company Open-AI, which has become world-famous within a very short space of time thanks to chat GPT. The American manager represents AI in the perception of many media consumers.
Artificial intelligence as a tool
But AI also has another face: it is the face of a young woman who works as an assistant professor at ETH Zurich: Menna El-Assady. She wants to create an AI that does not replace human abilities, but complements them. An AI that proves itself as a tool in the hands of a human. An AI that can be used interactively, that not only provides answers, but also allows questions to be asked. "I see it as a personal challenge," she says in a face-to-face interview, "to build bridges between people and machines."
At ETH Zurich, within the Department of Computer Science, El-Assady heads the Interactive Visualization & Intelligence Augmentation Lab. Research here focuses on the interfaces between humans and AI systems. Collaboration between humans and AI agents is a "co-adaptive process", emphasizes El-Assady. This means that humans must also continue to develop.
"We want to design interfaces that help users to get an idea of what a model can do," explains El-Assady. "These models can do a lot, but they can't do everything. People need to develop an understanding of what data and basic assumptions are contained in a model. Only then can they correctly interpret the answers they receive. We call this 'human empowerment': people should be enabled to use AI in a meaningful way. It's about taking away their fear of technology, but at the same time keeping a certain skepticism alive so that they don't believe everything.
From Alexandria to Zurich - Menna El-Assady's path to interdisciplinary research
Having grown up in the Egyptian metropolis of Alexandria, El-Assady came to Konstanz University at the age of 18. "I actually wanted to study physics at first, but then I switched to computer science after just a few weeks." Initially, she was fascinated by the mathematical basics of the field, the algorithms; then she became interested in natural language processing.
El-Assady speaks three languages fluently and has learned to navigate different cultures. This multicultural background has sharpened her sense of the different tonalities in political debates. She calls this "framing". Because she sourced information on current events from different channels, she noticed how big the differences in "framing" can be and was interested in exploring this insight as part of a computer-aided argumentation analysis. As part of her doctoral thesis, she was able to work on a project that involved analyzing the transcripts of political debates with the help of computers. The aim was to process the discussions, in which dozens of citizens had participated over many days, in such a way that the back and forth of the arguments could be explored interactively on the screen at various levels of abstraction. The methods of visual analytics, a sub-discipline of computer science, made it possible for people to grasp large amounts of data and complicated argumentation patterns at a glance.
This is how the biography of this scientist brings together disciplines that are otherwise separate: Natural Language Processing, Machine Learning, Visual Analytics, Interface Design, Human-Computer Interaction and more. Each of these disciplines is the center of its own scientific community with its own journals and conferences. In her research and teaching, El-Assady wants to unite these disciplines and thus create the conditions for interactive machine learning and artificial intelligence that places people front and center. El-Assady is convinced that you cannot have one - human-centered AI - without the other - interdisciplinarity. Interdisciplinary approaches are particularly valued at ETH Zurich, says El-Assady. She also had offers from other universities, but chose ETH Zurich because the AI Center fosters exchange between disciplines.
Can a university research laboratory in a small country compete with large American tech companies? Yes, believes El-Assady. There is a global trend in AI research towards open source models. On this basis, ETH Zurich, together with other Swiss universities and in particular with the support of the Swiss National Supercomputing Center in Lugano, is very well able to develop its own Large Language Models (LLM). El-Assady refers to the Swiss AI Initiative launched at the end of 2023, with which Switzerland aims to position itself as a leading global location for the development and use of transparent and trustworthy artificial intelligence, and to a network of international collaborations in the field of artificial intelligence, including the European Laboratory for Learning and Intelligent Systems (ELLIS) and the International Computation and AI Network (ICAIN).
AI at the turning point: opportunities and risks of the key technology of our time
The WEF's Global Risks Report sees the greatest global risk in the mid-term as "foreign and domestic actors using misinformation and disinformation to further deepen social and political divides". An "explosive increase in fake information", which can be produced and circulated on an industrial scale with the help of AI, is on the horizon, the report says. What should be done? Legislators are called upon to act, according to the Global Risks Report. However, there are also AIs that can counteract the mass spread of fake news. Not by simply trying to neutralize misinformation with mass-produced "correct" information. But rather by providing people with tools that enable them to quickly and interactively sift through large amounts of data, complicated issues and widely ramified discussions. This is the AI that El-Assady is championing.
Such AI supports people, multiplying their intellectual powers in the same way that a bicycle enables a person to move faster without taking the wheel out of their hands. This contrasts with the AI of the privately organized American tech giants, where humans are a kind of cargo that is transported from A to B at breakneck speed inside a windowless capsule: Once they reach their destination, they rub their eyes, not knowing what has happened to them.
This article was first published in the ETH AI Center 2024 Annual Report
Mennatallah El-Assady is an Assistant Professor at the Department of Computer Science, ETH Zürich. She heads the Interactive Visualization and Intelligence Augmentation (IVIA) lab. Prior to that, she was a research fellow at the ETH AI Center; and before that, she was a research associate in the group for Data Analysis and Visualization at the University of Konstanz (Germany) and in the Visualization for Information Analysis lab at the OntarioTech University (Canada). She works at the intersection of data analysis, visualization, computational linguistics, and explainable artificial intelligence. Her main research interest is studying interactive human-AI collaboration interfaces for effective problem-solving and decision-making. In particular, she is interested in empowering humans by teaming them up with AI agents in co-adaptive processes. She has gained experience working in close collaboration with political science and linguistic scholars over several years, which led to the development of the LingVis.io platform. El-Assady has co-founded and co-organized several workshop series, notably Vis4DH and VISxAI.