Image: Mahdis Mousavi (2019)
In this session of the 2024–2025 ASCA Cities seminar on ‘Playable Cities’, Sam Hind (Manchester University) will give a guest lecture on “Playing Domains: Competition in Machine Vision Challenges“ followed by an open discussion. The session takes place on Friday 14 March 2025, 3.00-6pm in room 0.16 (E-lab), Turfdraagsterpad 9, Amsterdam. For more information and registration, please contact Linda Kopitz (l.kopitz@uva.nl).
The history of artificial intelligence (AI) is also a history of play. The so-called ‘Grand Challenges’ of AI aimed to spark new interest during the late 1980s. Raj Reddy, then President of the influential Association for the Advancement of Artificial Intelligence (AAAI), proposed six such challenges in 1988 aimed at the computer science and AI community. One stated challenge was building a ‘world champion chess machine’. After losing the first match in February 1996, IBM’s famous ‘Deep Blue’ computer beat the reigning world chess champion Garry Kasparov, 3 ½ points to 2 ½ in May 1997. Another mentioned by Reddy was the ‘accident avoiding car’. Reddy believed advances in machine vision and sensing at the time could deliver a fully-autonomous vehicle in the very near future. Situating it within the long history of competitions and challenges in science and technology, this talk seeks to examine the role of contemporary machine vision competitions in structuring the development of AI, building on recent work on the role of ‘challenges as catalysts’ (Hind et al. 2024) in the development of autonomous vehicles, and the ‘competitive epistemologies of benchmarking’ (Orr and Kang 2024) within AI and machine learning domains more broadly. In these cases, different kinds of collaborative play are integral to the (competitive) development practices of computer scientists and AI practitioners, as they manipulate different statistical relationships in underlying (visual, urban) datasets. The talk considers two specific aspects: firstly, drawing on work in play and game studies, the role of online environments such as Google Colab in offering a ‘sandbox’ for playing with machine learning models. Secondly, building on work in science and technology studies (STS), the consequences of what I call ‘playing domains’: when the representation of urban environments and the safety of actors within them is ultimately, arguably, threatened by the ‘playful’ work of machine learners.
Sam Hind is lecturer in digital media and culture at the University of Manchester, UK. His current research interests include machine vision challenges, autonomous driving, the history of computer simulation, and the platformization of automobility. He recently published Driving Decisions: How Autonomous Vehicles Make Sense of the World (Palgrave).
Readings
Hind, S. (2024) Driving Decisions: How Autonomous Vehicles Make Sense of the World. London: Palgrave. (Chapter 3, Training Decisions: Ground-Truthing the Interesting). https://doi.org/10.1007/978-981-97-1749-1_3
Hind, S., van der Vlist, F. N. and Kanderske, M. (2024). Challenges as catalysts: How Waymo’s Open Dataset Challenges shape AI development. AI & Society 0 (0): 1-17. https://doi.org/10.1007/s00146-024-01927-x
The readings can also be accessed via this link.
Co-organized by Carolyn Birdsall, Linda Kopitz and Alex Gekker.