You are currently browsing the category archive for the ‘Conference’ category.
Call for Participation: Interspeech Speech & Intelligence Competition
Interspeech 2009 in Brighton, UK will be playing host to the Loebner Prize in Artificial Intelligence on September 6th. Building upon this tradition the Interspeech Speech & Intelligence Competition will pit systems against each other to see which is most capable of imitating a human conversation through the audio channel.
In keeping with the theme of ‘Speech and Intelligence’ the systems will be asked to complete a simple, direct and urgent scenario within an artificial setting. In order to usefully advance the current state of the art the challenge will be limited, requiring only an urgent focus on the task at hand. However, we anticipate that successful systems will posses some of the following features:
- An understanding of and ability to convey basic emotions
- A simple model of the judges current cognitive and emotional state
- Simple domain knowledge, appropriate to the current scenario
- A reasonable TTS system
- An ability to comprehend simple speech with little training
This will enable researchers to use the competition to address current, pressing, research challenges. To add further interest, children from local schools will be asked to judge each entrant (ensuring natural, unpredictable conversation within the scope of the scenario) and the systems themselves will be competing against local actors playing a role within the scenario. It promises to be a lot of fun!
The parents at @bristol presented a really interesting opinion yesterday. They all seemed to say “So that guy is dressed as a robot that’s cool but now you’re telling me that everyone here is a real live scientist, that’s amazing!” Are we really such a secretive breed that our mere presence is enough to cause excitement. I’m not sure how to react to that, it’s either really depressing: a scientist simply talking to the public should not be a novelty. Or it’s really exciting: people are genuinely interested in who we are and what we do.
Anyway, as my academic carrer continues to roll along I really hope that public engagement with the sciences becomes the norm and not the exception. Walking with Robots is making a really valuable contribution to that ideal and I guess that this blog is a small consequence of that. As you can see from the previous posts Johnathan and Gia have put me through a bit of a blogging masterclass and hopefully I’ll find the time to keep things going. Ideally, I’d like to use this blog to both draw in my academic collegues, by posting my own research and the research that we discuss at Southampton, and to engage with the general public, by blogging about the many, truly amazing, recent scientific developments. The dream would be to get both sides (academic and public) talking to each other in the comment threads. Will I succeed, frankly I have no idea…
Detective Spooner: Human beings have dreams. Even dogs have dreams, but not you, you are just a machine. An imitation of life. Can a robot write a symphony? Can a robot turn a… canvas into a beautiful masterpiece?
Sonny: Can *you*?
Could AI create an artistic masterpiece? Gareth Hallberg, from Sussex university is already planning a public engagement activity to investigate this very question. The plan is to set up an exhibition in Brighton sometime around August/September. Bringing together a collaboration of artists and AI researchers they will be questioning the creative process. Focusing on the difference between “p-creatives”, those that can be creative within an existing field of study, and “h-creatives”, those that can define entirely new conceptual spaces, it promises to be a fascinating exhibition.
Just in case you were wondering what we all look like at the Walking with Robots conference, here we are! As the conference carries on we’ll be bringing you more profiles of these lovely individuals, detailing their research, motivations and interests.
The subtle manipulation of objects is a pressing challenge in robotics. To convey the problems researchers face Martin Postler is using the Walking with Robots workshop to create a glove to restrict movement. This forces people to think about the complexities present in apparently simple tasks (pictured below) and highlights the importance of good design within robotics.
It also raises the interesting question: To what extent is intelligence embodied? If we could build a robot with the physical senses and capabilities of a human being (artificial muscles, an incredibly sensitive sense of touch etc…), how much intelligence would we need? Personally, I feel that quite a lot or AI within robotics is developed to overcome the robots own physical limitations. As increasingly sophisticated engineering solutions become available we may discover that intelligence is simpler than we realised.
One of our delegates, James Wilson, seen here posing in a giant hamster wheel. Usually spends his time at Aberystwyth University constructing robotic models of infant development, in the hope that we will gain greater insights into both. When you think about it it’s a brilliant approach. As researchers in Artificial Intelligence we spend a lot of time inventing weird and wonderful learning algorithms. Instead, we could save a lot of time by trying to copy what we find in nature, infant development is a perfect example of this.
Infants are capable of incredible feats of learning and they start before they are even born. In the womb, unable to see, they move their limbs developing a sense of co-ordination. Back in Aberystwyth this is modelled by having infant robots move around completely blind, through trial and error they learn how to co-ordinate themselves and discover the limits of their physical abilities. Then they are “born”, suddenly given a sense of sight and a complex environment they are not overwhelmed because they have already developed an initial sense of self. The success of these robots, helps to develop greater insights into infant development. Establishing these beneficial symbiotic relationships between nature and simulation is one of the key goals of AI. Better robots lead to a greater understanding of nature, which leads to better robots.
I was fortunate enough to recently attend the Symposium on Language and Robotics in Aveiro, Portugal. Where I was able to present my extended abstract on the idea of ‘Speech Perception as Non-Symbolic Pattern Recognition’. It was a fascinating conference and I was amazed to discover that a whole range of people seemed to agree with some of my more ‘unfashionable’ ideas. Ultimately, the range of ideas flying around left me with the sensation that I had compressed 6 months worth of PhD research into a single weekend. I’m now definitely looking forward to Evolang in March.If anyone is interested my extended abstract can be found here:
and a modified version of my presentation can be found here:
simon's blog 