Dagstuhl: A shining light in the Hochwald Black Forest

The Dagstuhl Institute

The Dagstuhl Institute, officially known as "Schloss Dagstuhl – Leibniz Center for Informatics," is a premier research center in Germany's Saarland region, dedicated to fostering innovation and collaboration in computer science. Its mission is to bring together world-renowned scientists, scholars, and industry experts to discuss cutting-edge research, share knowledge, and collaborate on groundbreaking projects. The institute hosts numerous seminars, workshops, and conferences. Last May, I traveled to Germany to participate in one of these initiatives, a symposium titled "Generalization by People and Machines." 

Getting to the Dagstuhl castle is not easy. Flying in from the US, the journey requires changing a couple of trains from a main hub (Frankfurt International Airport, in my case), riding a bus line that serves a constellation of villages in the area, or covering the same distance with a taxi. The reward opens before your eyes at arrival, with the beautiful silhouette of the Schloss, surrounded by the dense vegetation typical of the Hochwald Black Forest. Holding seminars in a remote location is not the fruit of a miscalculation, but the result of a successful design decision: experts from every corner of the globe love to go to the castle and spend some time away from their hectic professional schedule, discussing with peers during intense morning and afternoon sessions, which smoothly transition into dinner and late evening gatherings accompanied by cheese boards, locally-produced riesling, and refreshing weissbier.   

The Symposium on "Generalization by People and Machines"

Neural AI systems don’t generalize well beyond training distribution, a characteristic that is often needed in practical tasks–for example, safely maneuvering a vehicle on the road (one of the reasons why fully autonomous cars have not panned out yet, as hoped). Humans, on the contrary, are quite successful at generalizing from a few examples, and at filling gaps in experience with their abstraction and reasoning capabilities. What are the motivations for the discrepancy between human and AI generalization mechanisms? Can these be combined? How? For which applications?  

A strong tie with Carnegie Mellon

When compared to AI architectures that run on accelerated computing and are trained on a massive amount of data, human capabilities are even more impressive. They emerge from "bounded rationality," a notion distilled by the Turing Award and Nobel Prize recipient Herbert A. Simon, whose academic work and cultural influence helped with shaping AI into a truly interdisciplinary endeavor. Simon coined the notion of "bounded rationality" to denote how cognition operates within limited knowledge and under time constraints. Together with Allen Newell, he pioneered the first artificial intelligence programs focused on problem-solving and decision-making, paving the way for Carnegie Mellon’s enduring leadership in the field of AI. 

A source of inspiration for the Carnegie Bosch Institute

Among the memorabilia left by my predecessors at the CBI office, there’s a framed copy of a newspaper article from 1990 announcing the endowment from Bosch, which established our institute; I found the following paragraph particularly meaningful:

The Carnegie Bosch Institute will seek to foster international cooperation, aiming, ultimately, to bring together corporate, foundation and academic research partners from around the world to work side by side in developing new solutions for challenging problems.

Bosch, newspaper article from 1990

At a high level, CBI and Dagstuhl’s purpose are seemingly converging, although the histories of the two institutes took very different paths and emerged from unique ecosystems. In my brief visit to the Dagstuhl Institute, I could see what realizing the long-term vision of fostering cutting-edge research and innovation looks like. 

Our mission is to nurture a culture of research rooted in internationalization; our North Stars, Andrew Carnegie, and Robert Bosch, are formidable examples of the ingenuity, entrepreneurship, philanthropy, and foresight that we should aspire to in pursuing such an important mission.