Building Life-Mimicking Machines

As a child, Ahmed Abdalla enjoyed studying and reading books on just about any topic. In high school, he excelled in biology, although his favourite subject was physics. Studious as he was, at the national exams for university admission, he ranked 27th among the half a million students taking the exam across Sudan. With such results, he had the numbers to access the best academic courses in the country. “I wanted to challenge myself by choosing a competitive programme, and I chose electrical engineering. In hindsight, that was a good choice, because I like computing and the topics covered were broad,” he explains. 

In normal conditions, his bachelor’s degree could have been completed in five years, but it took Abdalla seven years to graduate because of the political instability in Sudan. “In 2019 my university was closed for one year following protests and the revolution. The current war started as I was working on my final thesis, which meant that we were displaced inside the country. I was very lucky to complete my bachelor’s degree in 2023,” he says.

Despite the hurdles, Abdalla’s motivation didn’t fade. During his bachelor’s degree, he specialised in control engineering and developed an interest in reinforcement learning, both widely applied in modern robotics. “I have always been interested in how people and machines interact with the world. One thing we have in common with machines is that neither has access to the world as it is. Like them, we use input information – in our case coming from our senses – to define our own representation of the world, and then base our behaviour on that,” he explains. 

After graduating, Abdalla joined a master’s programme in machine intelligence at Senegal’s African Institute for Mathematical Sciences (AIMS). While his training as an electrical engineer had prepared him to work on very applied problems, AIMS gave him a more theoretical background. Because the master’s programme is sponsored by Facebook and Google, many faculty members are from Europe and the US, and work at some of the biggest tech companies. “It was exciting to be lectured by people with such expertise,” he says, adding, “I believe it contributed to my acceptance here at ICTP.” 

It is during his time at AIMS that Abdalla heard about the ICTP Postgraduate Diploma Programme. Another student, also from Sudan, who attended the master’s just one year before him, was stuck in Senegal for a few weeks as she was waiting for her visa to move to Trieste and join the ICTP Diploma Programme. “We stayed in touch after she left and as she was telling me many nice things about ICTP, I started checking it out,” he explains. Like many of his fellow students, Abdalla chose to join the ICTP Diploma Programme because it promised to give him strong theoretical foundations and the possibility to be exposed to different topics before deciding to specialise in one. 

The start, however, was bumpy. Abdalla’s visa arrived late and he managed to reach Trieste only one month after the programme had already begun. “I had a lot of catching up to do. Not coming from a physics background alone would have been challenging enough, but having to go through a whole month’s worth of lessons at the pace of this programme was a lot,” he recounts, adding, “I also worried that by the time I arrived, everyone would have already found their social circles and I would find it difficult to fit in.”

Instead, when he arrived, ICTP felt welcoming. “I always felt supported and encouraged. At the beginning, I was very intimidated, also considering the level of the professors, but I learnt a lot and this experience at ICTP has made me change the way I think about problems,” he explains. Despite the initial fears, he also quickly became friends with his fellow Diploma students in the Quantitative Life Sciences section. “They are like brothers and sisters to me now. They are definitely the best part of this experience,” he says.

For his final project, Abdalla worked on Finite-State Controllers, interpretable models for studying decision-making with limited memory. These have been applied by QLS researchers, for example, to investigate the behaviour of animals. Although the project focuses on a relatively simple system, it has allowed Abdalla to develop mathematical and computational tools that he plans to build on during his PhD. Next autumn, Abdalla will start a four-year doctoral programme in Theoretical and Scientific Data Science at SISSA, in Trieste. “I am happy that I will continue to work with many of the people that I have met this year and to stay in Trieste, a city that I like so much,” he concludes.