The Deep Learning Indaba is the largest machine learning and AI conference on the African continent. It brings together people from research and industry to build the future of AI in Africa. Our satellite workshop was about NeuroAI, what research happens at the intersection of neuroscience and machine learning, but also who gets to do it, and what it takes to grow that expertise on the African continent.
The workshop opening was by Dr. AZA Allsop (Yale University) with a talk on “Building African NeuroAI Capacity”, followed with keynote talks by Dr. Devon Jarvis from the Machine Intelligence and Neural Discovery (MIND) Institute, who spoke about “Understanding Language Evolution from Humans to LLMs”, and Dr. Jonathan Shock from the University of Cape Town, who turned the usual question around and asked what we can learn about ourselves by studying AI.
Then, in a hands-on tutorial, participants worked through a Colab notebook on training and evaluating foundation models for neuroscience, so everyone left with something they could keep using.
We closed with a panel discussion on future directions in neuroscience and AI, career paths, and capacity building on the African continent.
'’Thank you to all participants for the engaging questions and for the great atmosphere in the room. The excitement about machine learning in neuroscience was obvious. And the panel discussion was probably as empowering for those of us answering as for the people asking.’’
This was funded by the collaborative research center “Robust Vision – Inference Principles and Neural Mechanisms” (CRC 1233).
I got invited as a representative of the machine learning ecosystem in Tübingen and the Cluster of Excellence “Machine Learning: New Perspectives for Science.” Together with my colleagues from Cyber Valley, we showcased the science happening in and around the region as a science pitch. And as if taking the stage in front of Nobel Laureates didn’t add enough pressure, this was the challenge:
To prepare, I had the opportunity to work with a vocal coach on my presentation style and my voice (if you’re in or around Tübingen, I highly recommend Wortfabrik). I find it quite striking that we scientists are expected to present our work confidently on stage but we almost never receive any formal training for it. A few things I took away from the coaching:
Because the same day I spent with the Nobel Laureates they also had a panel discussion, it became the title of this post: How do we rebuild trust in science?
A few points stuck with me:
It’s both a blessing and a curse: as scientists we can get lost in the fast pace of discovery (I certainly do). Deadlines, papers, and (the best part) new results. But what is any of it worth if it never leaves your lab, or your bubble?
And who is going to push back on misinformation, if you’re the expert and you don’t say anything?
A recent Nature article looked at this and the numbers paint a more multi-layered picture: in the UK for example, trust in science hasn’t disappeared: 84% of people still say they trust science at least somewhat, and only 2% don’t trust it at all. What did change is how many say they trust science “a lot”, which fell from 63% in 2020 to 34% in 2026. Trust hasn’t disappeared. But people became more careful to believe everything they hear.
Communicating honestly is the whole point. I see science communication as part of my job. It’s the mechanism that rebuilds trust in science. And I wish more people saw it that way. Here is the ending of my 3-minute pitch, that also highlights why I think trust in science matters more than ever right now:
We are humans: we have a brain and we like to think that we are intelligent. But it seems like AI is catching up. And sometimes, AI is already the one having the better ideas. And yet AI can also be spectacularly wrong and hallucinate facts. It lies to us. And it does that with confidence. It can write thousands of lines of code for me that look perfectly fine. Until I’ve spent hours debugging it and discover it did the wrong thing after all.
We might be standing at the edge of the biggest shift in how science is done.
If we are, then being able to explain our research isn’t a soft skill. It might be the whole point.
The main question we explored was simple: can large language models (LLMs) accurately predict epilepsy diagnoses from nothing more than verbal descriptions of seizures. The kind of descriptions patients and doctors give, in their own words, in their own language?
This matters more than it might initially seem. Epilepsy diagnosis is often delayed, particularly in settings where access to specialist neurologists or EEG facilities is limited. A patient’s description of their seizure, what they felt, how they moved, what others around them noticed, remains one of the most clinically informative pieces of information available. But processing and interpreting those doctores notes is not trivial. That’s exactly where language models might help.
Alongside the research, I organized a workshop on Clinical Neuroscience & AI, bringing together researchers from Cape Town and Tübingen with the goal of building on an already existing collaboration.
The talks covered many topics, from using machine learning to find markers of cognitive health, to the vision for an AI initiative in South Africa, to computational psychiatry. Each one opened up a different angle on the same big question: what does AI-driven diagnostics actually look like when it’s embedded in real healthcare systems, with real constraints?
Organizing a workshop is always more work than it looks from the outside, the logistics, the coordination across time zones, and decisions to make, but I’m genuinely thankful for the experience. I learned so much, not just about event management, but about how differently people in different research contexts frame the same problems.
The Neuroscience Institute at the University of Cape Town gave me a research context I hadn’t experienced before. The proximity to the clinincal context made the effect our research could have feel very realistic. The question of how AI can support diagnosis, especially in resource-constrained or underserved settings, is a research field I want to gain more insights into. Cape Town also reminded me, that the best research happens in community. I’m looking forward to building on what started here.
Established in 2017, the Indaba has grown to become the largest annual machine learning and artificial intelligence conference on the African continent. This year’s conference operated under the motto “Urunana” — meaning “hand in hand” in Kinyarwanda. The theme emphasizes the strength of community and the importance of coming together to collaboratively shape the future of AI, both between African nations and the global community, and between industry and academia.
The conference hosted workshops, covering topics such as “Data Science for Health in Africa,” “Robotics,” and “TrustAI.” The NeuroAI workshop, where I had the privilege of presenting, explored the exciting intersections between neuroscience and artificial intelligence.
]]>In my talk “Thalamus Thought Traffic”, I explored how machine learning can help us understand complex computations in the brain and, vice versa, how insights from neuroscience can inspire the development of more robust AI algorithms.
The event was a fantastic opportunity to step outside the lab and communicate science in a fun and accessible way. Science slams are a great reminder that the stories behind our research matter just as much as the results.
The Science Slam – AI Edition was organized by the Tübingen AI PhD und Postdoc Assembly (TAPPA) and supported by:
During my Ph.D., I used machine learning tools to investigate neural mechanisms in the brain, employing diverse statistical methods to explore how single neurons and neural populations process sensory information depending on cell type and brain state.
A key achievement was developing a data-efficient vision model that integrates not only visual information and brain state but also cortical neural feedback. This work bridges computational modeling with biological mechanisms to better understand how the brain processes sensory information.
The spotlight also features one of my favorite moments: presenting my research during sunset 🌅 at the AREADNE conference on encoding and decoding of neural ensembles on a beautiful Greek island 🇬🇷
Read the full spotlight here: https://imprs.is.mpg.de/news/scholar-spotlight-lisa-schmors
]]>AREADNE is held on a comparatively small scale (~100 participants) and is deliberately designed to engage all attendees in discussion and community building. Each of the four days featured two oral sessions, interspersed with a long afternoon break that gave participants time to meet one another, share their research, and explore the island. It was a refreshing contrast to large conferences where it can be hard to have meaningful conversations.
The poster sessions, held outdoors at sunset after the evening talks, created a truly unique atmosphere. The setting made discussions feel relaxed and genuinely enjoyable. Science at golden hour on a Greek island is hard to beat.
One particularly democratic aspect of AREADNE: essentially every participant is required to give a poster presentation, regardless of career stage. This made it easy for early-career scientists to get in touch with leading researchers in the field. I presented my work on the functional organisation of visual input in the superior colliculus.
The conference’s focus on new and largely unpublished research contributed to the interactive, discussion-friendly atmosphere. Some personal highlights:
AREADNE was a genuinely great experience, both scientifically and in terms of community building. As a PhD student, it was particularly motivating to engage so directly with leading researchers in an environment that felt open and collaborative rather than hierarchical. I would highly recommend it to anyone working in systems or computational neuroscience.
]]>The history of CERMEL dates back to 1913, starting with the first Albert Schweitzer Hospital in Lambaréné. Today, it also includes a research laboratory focusing on malaria and other tropical diseases. Despite the challenges here in Lambaréné, such as its humid 33-degree climate and occasional lack of water and electricity, the dedication of everyone on campus is remarkable.
The commitment of the 15 master students impressed me deeply. Their motivation to learn and understand complex topics in mathematical modeling and biostatistics, essential for researching infectious diseases, was truly inspiring. I have no doubts that they will be the next generation of highly skilled African researchers who will successfully tackle the global challenges of future infectious diseases.
Julian Friedrich and Markus Kleinhansl are part of the PhD initiative Sustainability in Tübingen, which has actively engaged in various sustainability projects. They also created a survey focusing on the working habits and mental health of PhD students. Among their findings: 31.3% of participants work 45 hours or more, and one third of PhD students is at risk of major depression.
During the podcast, we discuss the causes of stress among PhD students and how this pressure can sometimes lead to crisis situations. We also talk about practical techniques to manage stress and overcome the “imposter” feeling many of us experience.
I joined the I AM SCIENTIST podcast team in 2022, started by Daniel Hölle and Philipp Hubert. Since then, the three of us have had a lot of fun working together and growing this project. In the podcast, we interview PhD students about the science they are passionate about, their day-to-day life as researchers, and what personally drives them.
All our efforts for this podcast happen in our free time, and we (so far) only produce episodes in German for our listeners there.
Recording this episode on mental health was an opportunity for me to raise awareness on important issues, reach more people, and make a difference.
]]>