OFAI is delighted to announce its 2024 Spring Lecture Series, featuring an eclectic lineup of internal and external speakers.
The talks are intended to familiarize attendees with the latest research developments in AI and related fields, and to forge new connections with those working in other areas.
Lectures will take place at 18:30 Vienna time, usually every other Wednesday. All lectures will be held online via Zoom; in-person attendance at OFAI is also possible for certain lectures. Attendance is open to the public and free of charge. No registration is required.
Subscribe to our newsletter or our RSS feed, or bookmark this web page, to receive further details for the individual talks.
24 April 2024 at 18:30 CEST (UTC+2)
Jordan Kodner (Stony Brook University)
Is it Language or Task Design? Reinterpreting language models' recent successes in morphology and syntax learning
The success of neural language models (LMs) on a wide range of language-related tasks may be in part due to their ability to induce human-like representations or understanding of natural language grammars. Humans are, after all, gold-standard language learners. For the past several years, researchers pursuing this question have developed a number of methodologies for testing the grammar representations learned by LMs that have reached generally positive conclusions. I will take a critical look at such studies in this talk. While modern LMs are clearly extremely impressive, and clearly do often capture important aspects of natural language grammars, the methodologies of many popular studies have unfairly overestimated the capacities of LMs when it comes to their ability to induce human-like representations. Focusing on questions of hierarchical syntactic representations and generalization in inflectional morphology, I will discuss how unintended biases in data-splitting, artificial training or test data, overly simplistic evaluations, weak or absent baselines, and faulty interpretations, have conspired to overestimate the abilities of LMs. While the conclusions of this study are largely negative in terms of the current state-of-affairs, they are also optimistic. By employing more thorough and rigorous methodologies, we have developed a better scientific understanding of the nature of LMs and representations of the grammar. In identifying weak points for current models, we points towards research areas where greater improvements may be gained.
29 May 2024 at 18:30 CEST (UTC+2)
Jim Young, BSc, PhD (University of Manitoba)
Designing Human–Robot Interaction
How should we interact with a robot? How can we give it commands? Get information from it? Robots' real world, often collocated and autonomous presence, provides a range of new and exciting opportunities for re-envisioning interaction with technology. In this talk, Dr. Young will present his team's work on exploring novel interaction with robots through a range of projects over the last 12 years. A key focus of this work is aiming to solve HRI problems through novel interaction design rather than technological advances, re-conceptualizing problems to make them simpler. Further, Dr. Young's team explores the limits of robots' abilities to use emotion and human social interaction techniques, for example, to deceive and manipulate people. Finally. Dr. Young will introduce his lab's current projects on re-designing domestic companion robot interactions with a focus on simplicity and deployability.
How to attend: Attend online via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.
You can add this event to your calendar.
05 June 2024 at 18:30 CEST (UTC+2)
Dr. Margherita Pallottino (University of Geneva / OFAI)
Fii uses across Arabic varieties
The particle fii (etymologically the preposition ‘in’, with allomorph bi-) is universally attested across Arabic varieties. Alternative uses of this element have been identified in single dialects. For instance in the Levantine dialects fii is used in existential constructions (Jarad, 2015) or as a modal (Cowell, 1964), whereas in the North African region fii is an ingredient of progressive constructions (Brahim, 2007). The existing literature gives us a picture of the multifunctionality of fii in Arabic but it does not provide any indication about the geographical distribution of these usages, nor about the possible correlations among them. The research described in this talk fills this gap surveying 11 different uses of fii previously described in the literature, across 11 different Arabic varieties extending from Morocco to Iraq. The resulting picture shows a leopard-spot distribution of certain functions of fii suggesting that other forces drive the emergence of fii usages across varieties beside the influence of geographically proximity. The main question tackled by this talk is, therefore, which factors can explain the distribution of fii functions. The validity of the competing hypotheses, historical accident or grammar driven, are evaluated with mathematical methods run on the data gathered in the study.
19 June 2024 at 18:30 CEST (UTC+2)
em. O. Univ.-Prof. Dr. Hubert Haider (Universität Salzburg)
Multiple targets – Ambiguitäten als Kriterium für LLMs
Laut AI-Index-Report 2024 wurde die menschliche Leistung als Baseline im Fall von Natural Language Inference bereits von LLMs überschritten. Es gibt aber einen sprachlichen Aufgabentyp, bei dem diese Systeme schlechter abschneiden. Das ist das Erkennen und Berücksichtigen syntaktischer Ambiguitäten, was der Vortrag zu zeigen sich bemühen wird.
How to attend: Attend online via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.