50 years of AI
The term "Artificial Intelligence" (AI) was coined at a conference in Dartmouth College, NH, in the summer of 1956, so this year has seen several celebrations, e.g. in Boston, Monte Verità, Paris... so Brussels couldn't be left behind!
The AILab of the VUB organized a one day event at the Royal Flemish Academy for Arts and Sciences, with invited talks by people who have done advancements in AI in Flanders... so it was basically VUB people (i.e. ex-students of Luc Steels) and a couple from Leuven. You can see the full info here (in Dutch).
So, during and after the presentations, I realized how the expectations of AI have changed. In the 50's, people were predicting that we would have robots with humanlike intelligence in a few decades. In the 80's, a new wave of AI began to spread, namely with behaviour-based systems, neural networks, dynamical systems theory, and artificial life. Many of these were also aiming at humanlike intelligence (from the bottom-up), severely criticizing the approaches of GOFAI (good old fashion AI, based on symbol manipulation, knowledge modelling, planning, etc.). The fact is that "new AI" expectations haven't been reached anyway, and now most people make hybrid systems. This is because knowledge-based systems, behaviour-based system, connectionist systems, conceptual systems, etc. aim at different aspects of cognition. (See Cognitive Paradigms: Which One is the Best? Cognitive Systems Research 5(2):135-156, June 2004).
Then, can we say that AI has been a failure???
Not at all! It didn't fulfil its sexiest expectations, but it has made great technological advances, and not only in playing chess! The thing is that it has been absorbed by other domains, which now use AI techniques, without calling them AI... In engineering, computer science, linguistics, genomics entertainment... Many cars and factories have AI-based self-diagnostic systems, all the automatic translators and recognizers are AI. Special effects in films, strategies in games, multi-agent systems, modern programming languages and techniques... all of them have AI roots. Only that we don't call them AI anymore...
The AILab of the VUB organized a one day event at the Royal Flemish Academy for Arts and Sciences, with invited talks by people who have done advancements in AI in Flanders... so it was basically VUB people (i.e. ex-students of Luc Steels) and a couple from Leuven. You can see the full info here (in Dutch).
So, during and after the presentations, I realized how the expectations of AI have changed. In the 50's, people were predicting that we would have robots with humanlike intelligence in a few decades. In the 80's, a new wave of AI began to spread, namely with behaviour-based systems, neural networks, dynamical systems theory, and artificial life. Many of these were also aiming at humanlike intelligence (from the bottom-up), severely criticizing the approaches of GOFAI (good old fashion AI, based on symbol manipulation, knowledge modelling, planning, etc.). The fact is that "new AI" expectations haven't been reached anyway, and now most people make hybrid systems. This is because knowledge-based systems, behaviour-based system, connectionist systems, conceptual systems, etc. aim at different aspects of cognition. (See Cognitive Paradigms: Which One is the Best? Cognitive Systems Research 5(2):135-156, June 2004).
Then, can we say that AI has been a failure???
Not at all! It didn't fulfil its sexiest expectations, but it has made great technological advances, and not only in playing chess! The thing is that it has been absorbed by other domains, which now use AI techniques, without calling them AI... In engineering, computer science, linguistics, genomics entertainment... Many cars and factories have AI-based self-diagnostic systems, all the automatic translators and recognizers are AI. Special effects in films, strategies in games, multi-agent systems, modern programming languages and techniques... all of them have AI roots. Only that we don't call them AI anymore...
Comments