21st International Joint Conference on Artificial Intelligence (IJCAI), Pasadena

Embodied Language Games for Autonomous Robots

introducer: Ramon López de Mántaras, Artificial Intelligence Research Institute - IIIA, Spanish National Research Council - CSIC
author: Luc Steels, Computer Science Department, Vrije Universiteit Brussel
published: July 22, 2009, recorded: July 2009, views: 516

Slides

Slides
0:00	Embodied language games for autonomous robots
2:25	Open-ended grounded language communication remains elusive
2:50	Issues
5:17	... on human-level natural language
7:00	Let’s get inspiration from evolutionary biology
7:43	Evolution of eggspots in cichlid fish
9:25	Methodology of evolutionary biology
10:22	Example of a language system (1)
12:06	Variation
13:31	Evolution of language strategies
14:23	Transposing the methodology of evolutionary biology to AI
15:18	Language Game
15:55	Sony Orio humanoids
16:23	Case study 1: Color
16:31	The Color Naming Game
16:37	- Example
18:38	Cone response curves - Demo
21:18	- Example Cont.
22:47	Deconstruct Spanish colors: 1. Comprehension and production (1)
22:51	Deconstruct Spanish colors: 1. Comprehension and production (2)
23:46	Deconstruct Spanish colors: 2. Learning of the lexicon (1)
24:19	Deconstruct Spanish colors: 2. Learning of the lexicon (2)
25:09	Deconstruct Spanish colors: 3. Learning ontology and lexicon (1)
25:44	Deconstruct Spanish colors: 3. Learning ontology and lexicon (2)
26:23	Deconstruct Spanish color strategy: 4. Self-organisation of lexicon and ontology (1)
27:08	Deconstruct Spanish color strategy: 4. Self-organisation of lexicon and ontology (2)
28:03	Example evolution ontology and lexicon for 5 agents (1)
28:41	Example evolution ontology and lexicon for 5 agents (2)
29:37	Semiotic network
30:57	Why it works:
32:43	Inspiration from evolutionary biology
33:10	Inspiration from evolutionary biology - Comparison
33:26	Color strategy based on hue+brightness
33:43	Color strategy based on brightness
34:04	Comparing two strategies
35:06	Ongoing competition
36:03	Case study 2: Body
36:11	Body
37:03	Action Game
38:55	Visual Feature Extraction
41:37	Recording of motor babbling
42:15	Exaptation after 1000 games
43:02	Case study 3: Space
43:12	The spatial reference game (1)
45:05	The spatial reference game (2)
47:32	Construction
48:11	Fluid Construction Grammar
48:27	Conclusion
50:43	Thanks to the teams!
50:56	- Questions

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.

Description

Artificial Intelligence methods and techniques have reached a high level of sophistication so that we can tackle difficult outstanding problems in science. In this talk, I will show how the question of the origins of language can be approached this way. This question has puzzled evolutionary biologists since Darwin and is still considered to be unsolved. I will outline a theory of language evolution by linguistic selection and then report a number of concrete experiments with humanoid robots that attempt to work out and validate this theory. The experiments all center around the notion of a language game, which is a routinized situated interaction that involves some form of language. Robots use linguistic strategies to evolve a communication system to deal with a particular class of language games. I will discuss examples of this and also address the question how new strategies can arise and how the robots can autonomously decide which strategies they will collectively use to bootstrap their language.