GoodAI was proud to support the Animal-AI Olympics which was organized by the Leverhulme Centre for the Future of Intelligence, Imperial College London, and the University of Cambridge. The project aimed to benchmark AI agents against multiple animal species using a range of established animal cognition tasks.
The competition had over 60 entries from across the world and you can watch some of the agents perform in the testbed as well as testing it for yourself in your browser here.
During testing, AI-agents were exposed to animal cognition tasks, which were split into 10 categories.
- Food
- Preferences
- Obstacles
- Avoidance
- Spatial Reasoning
- Generalisation
- Internal Models
- Object Permanence
- Numerosity
- Causal Reasoning
Participants received only limited information about the tests (category names and brief descriptions) and while the competition was running they could test their agents on a representative testbed of 300 problems. However, they could only see their scores per category as feedback.
The overall winners of the competition were:
- 1st Place: Trrrrr
- 2nd Place: Ironbar
- 3rd Place: Sirius
You can find the full table of results for each task and find out who won the category prizes here and find out details of some of the tests here.
