Why the world is not in the toilet (yet)

"Knowing whether humanitarian action in a conflict area is effective requires feedback loops. Until now, these have predominantly been qualitative: observations, interviews, and status reports compiled by field analysts to monitor their work. Today, the large amount of data collected and automated, machine-learning-based event analysis can be combined to provide a quantitative measure of conflict intensity. This allows NGOs and humanitarian actors to monitor the impact of their actions and make more informed, data-driven decisions for future planning.
Furthermore, the current online media landscape provides unprecedented access to publicly available information on conflicts. Manually processing this open-source content has become increasingly challenging and time-consuming, and it exposes staff to potentially distressing material. Artificial Intelligence-based technologies can automate the analysis of conflict-related data, such as photos of documents, military tags, and written reports from war zones, and help to clarify the fate and whereabouts of missing persons."

What if engineering’s biggest problem were not only technical, but cultural? From classrooms to labs, engineering often rewards emotional silence, self-reliance, and competition, then treats them as signs of merit. But these norms do more than shape who feels they belong. They also shape what technology gets built, which problems seem worth solving, whose needs are treated as urgent, and whose voices are ignored. The result is technology that makes life easier for some while making it harder for others, while steering society in directions we too often accept as inevitable. So I want to open up a different question: what might change if we trained engineers not just to solve problems, but to care—about consequences, vulnerability, dependence, and the people left to live with the fallout?

Biodiversity is under increasing pressure from climate change, deforestation, and invasive species, making conservation more urgent than ever. But that raises a key question: where should we focus our efforts first?
While we don't know exactly where species live, we can estimate their distribution. By combining species observations with climate data, we can map where species could live. We can also use climate change scenarios to predict how these distributions might shift in the future. This information can then be used to guide conservation decisions.
In this talk, I’ll show how we used machine learning to map the distributions of more than 10,000 tree species across the globe. These maps help us investigate how forests are composed, assess the impact of historical forest loss and predict effets of climate change. We find that the impacts aren’t the same everywhere: some ecosystems are much more vulnerable than others.
These maps also let us ask new questions. For example: are trees already living close to their thermal limits? By combining our models with lab measurements, we found that in tropical forests, more and more areas are reaching temperatures in which trees may no longer be able to perform photosynthesis. This trend is likely to worsen with climate change.
Overall, this work highlights two key points: we need to protect the forests we still have, and when we restore degraded ones, we need to do it in a way that helps them survive in a warmer future.