This talk will take place on Microsoft Teams. Contact the organizer for more details.

Over the last decade, advances in machine learning have taken the computer science community by storm, enabling new applications and pushing the envelope on existing ones. Even on tasks that are traditionally viewed as falling in the domain of logical reasoning (e.g., reading comprehension tasks), deep neural models are now the state-of-the-art. Furthermore, logic and learning are perceived by some as being distinct or even opposing approaches. At the same time, while various algorithmic and hardware limitations that inhibited deep learning solutions in the past have been successfully addressed, other fundamental problems arise, such as problems concerning fairness, explainability, and controllability. In this talk, I will discuss a few problems at the intersection of machine learning and logic, including providing deep models with means to access structured knowledge.

Zoom link: https://uva-live.zoom.us/j/82670894282

Prof. dr. Sonja Smets (Institute for Logic, Language and Computation):
"When agents learn new information they have to be very careful, because the fact of learning information may interfere with the reality that is being learned."
Interviewed by: Dr. J. Ashley Burgoyne (Amsterdam Music Lab).
Zoom link: https://uva-live.zoom.us/j/82520968502

Originally scheduled speaker: Justin Clarke-Doane (Columbia) on "Russell’s Regressive Method in Mathematics and Philosophy"

Type systems are a useful tool to prevent programmers from making obvious mistakes, but they are generally quite limited in what they can describe. We've taken a look at a session types, a type system aimed at describing communication protocols, and checking that programs adhere to these protocols. Where previous work on session types has treated them as syntactical objects, we find that protocols have a natural notion of state and characterize them as coalgebras, i.e., state-based machines. In doing so, we retrieve natural definitions for type-equivalence, subtyping and duality of types as coinductive relations between states. In my talk I'll explain what session types are and present our syntax-free description of protocols as states of a coalgebra (without assuming prior knowledge of coalgebras).

Zoom Meeting ID: 878 6558 7983 (link: https://uva-live.zoom.us/j/87865587983)

All human societies have languages capable of expressing the richness of human thought. To what extent is this achievement an historical accomplishment, similar to mathematics or science, and to what extent does it rely on our evolved cognitive capacities? I study these questions by looking at language creation in different communities, including Nicaraguan Sign Language (a new language only 50 years old), homesign systems, and laboratory-created communication systems. I will present results on how a new language comes to have recursion and quantifiers like “some” and “all." In both cases, I find evidence for rapid emergence of linguistic structure within a few generations. One possible explanation for these findings is that features that emerge early are those that reflect underlying shared semantic structures that are universal (or nearly) in languages. In contrast, the features that emerge later (e.g., grammatical morphology) may be those that vary across languages and require convergence and iterated learning.

This talk will be given in American Sign Language (ASL) and interpreted into English. If anyone would like to attend the talk and have it interpreted in Sign Language of the Netherlands (NGT), please send an email to F.Roelofsen at uva.nl in advance and we will do our best to arrange an NGT interpreter.

Professor Joel David Hamkins from the University of Oxford will be at Φ-Math to present his upcoming book Lectures on the Philosophy of Mathematics. The presentation will contain an overview of the book's contents and motivation with a focus on selected philosophical problems tackled in it, followed by a discussion/questions from attendants.

The Nordic Online Logic Seminar (NOL Seminar), will be organised monthly over Zoom, with talks on logic topics of interest for the broader logic community.
The first talk will be given by Dag Prawitz on Monday, March 22, 16.00-17.30 (UTC+1).

Contact the organizer to join the talk on Microsoft Teams.

Weekly social event for PhD students, postdocs, and generally non-permanent research people at the ILLC. Meetings last between 30 and 45 minutes and feature a short (generally non-academic) presentation by one of the attendants on a topic of their choice followed by an open informal discussion.

Zoom link: https://uva-live.zoom.us/j/85133392465

In many languages grammatical relations are to some extent lexically restricted, in the sense that certain verbs or verb classes take different argument coding frames than others. While such constraints are well studied for case marking, they have also been reported for grammatical relations defining other types of constructions, including a range of voice- and valency-related constructions and some clause-combining constructions. This hybrid (on-line/on-site) workshop aims to unite scholars from different (sub)disciplines, bringing together descriptive, comparative, corpus-based, and experimental studies, as well as studies that compare linguistic data with genetic and/or socio-historical evidence. Together, we hope to further our understanding why lexical restrictions should exist, how they are processed and acquired, and why/how/where they persist in languages.

In many languages grammatical relations are to some extent lexically restricted, in the sense that certain verbs or verb classes take different argument coding frames than others. While such constraints are well studied for case marking, they have also been reported for grammatical relations defining other types of constructions, including a range of voice- and valency-related constructions and some clause-combining constructions. This hybrid (on-line/on-site) workshop aims to unite scholars from different (sub)disciplines, bringing together descriptive, comparative, corpus-based, and experimental studies, as well as studies that compare linguistic data with genetic and/or socio-historical evidence. Together, we hope to further our understanding why lexical restrictions should exist, how they are processed and acquired, and why/how/where they persist in languages.

As a young mathematician, Brouwer gained prominence by proving a number of fundamental theorems in topology, the most famous of which is his eponymous fixed point theorem. However, Brouwer would later come to reject this theorem as not intutionistically acceptable. In this talk, I will (briefly) sketch the proof of Brouwer's fixed point theorem as it is given in most textbooks, and present an alternative, 'almost constructive' proof based on Sperner's lemma. I will also discuss how this second proof can be used to obtain intuitionistically valid variants of the fixed point theorem.

Zoom link: TBA. Please check website - Zoom link will be provided the day before.