Recently in AI Category
Interesting view of the Loebner/Turing from a judge this year. His approach is very much to find out if the bot is a human or not, rather than to see if he can have a "normal" conversation. Again this pushes me to the belief that we shoudl have a tiered approach to this:
- test where testers don't know there are any bots
- test where the testers know there are bots but have "normal" conversations
- full gloves-off test
His questions were very much reasoning tests of the third kind:
- if we shake hands, whose hand am I holding?
- I have a jewellery box in my hand, how many CDs can I store in it?
- he four capitals of the UK are three, Manchester and Liverpool. What's wrong with this sentence?
How many times do you ask those in normal conversation?
- test where testers don't know there are any bots
- test where the testers know there are bots but have "normal" conversations
- full gloves-off test
His questions were very much reasoning tests of the third kind:
- if we shake hands, whose hand am I holding?
- I have a jewellery box in my hand, how many CDs can I store in it?
- he four capitals of the UK are three, Manchester and Liverpool. What's wrong with this sentence?
How many times do you ask those in normal conversation?
http://www.ddj.com/architect/184416017?pgno=1
Be interesting to see how we can build this into the AI emotion engine - and probably useful for the motivation engine too.
Be interesting to see how we can build this into the AI emotion engine - and probably useful for the motivation engine too.
http://technology.newscientist.com/article/mg19926696.100-rise-of-the-ratbrained-robots.html?DCMP=NLC-nletter&nsref=mg19926696.100
I can remember speaking to Prof Warwick about this and some other interesting applications of the "technology"about 9 months ago - interesting to see it now maturing.
I can remember speaking to Prof Warwick about this and some other interesting applications of the "technology"about 9 months ago - interesting to see it now maturing.
Spent the day working on our ChatbotIF interface to Second Life for a change. This sits between our Discourse chatbot engine and libsecondlife. Worked my way through a list of things we'd flagged up over the past few weeks. Improvements/new features include:
- better point to point movement by estimating travel time for each segment
- full IM handling, including routing IM via Discourse and setting the bot to decline IMs
- full friend handling, including setting the bot to decline Friend requests
- a "fidget" function when idle to play random animations and movements
- ability to set whether to ignore or listen to objects on channel 0
- ability to control movement from Discourse so the bot can more easily show people around
- accept teleport lure's form owner
Always more to do though!
- better point to point movement by estimating travel time for each segment
- full IM handling, including routing IM via Discourse and setting the bot to decline IMs
- full friend handling, including setting the bot to decline Friend requests
- a "fidget" function when idle to play random animations and movements
- ability to set whether to ignore or listen to objects on channel 0
- ability to control movement from Discourse so the bot can more easily show people around
- accept teleport lure's form owner
Always more to do though!
Just loved this diagram in the IET's Engineering magasine - I think accompanying a future-gazing article by Ian Pearson at BT.
Not a bad looking book from the contents. Even got a chapter by Pei Wang on NARS. Might pick it up - after waiting to see what comment is like on Robitron.
The home of stigmergic systems
Might be something interesting here.
"It is a very clever strategy used by nature to get colonies of insects to self-organize, tell each other where to find food, create sophisticated messaging systems and build complex architectural structures."
Gwyn%u2019s Home » The $20,000 Question: An Essay by Extropia DaSilva
"We think a lot about the nature of the brain, and whether computational substrates can be dense enough to enable thinking within them. I know exactly how that's going down, I think SL is dreaming. It could be looked at as one collective dream." - Philip Rosedale
Cracking essay by Extropia DaSilva on AI in SL
Yesterday I kicked off a new AI project. For the last few weeks we've been getting a mimic bot up and running in SL. A mimic bot is one that learns purely by hearing what people say to it, and then re-saying those things later and then remembering the response. You end up with a very natural chatbot, but one that hasn't got the faintest idea whatit is talking about. We think there'll be a great market for them in SL as drunks at the end of the bar! e also added some explicit learning, and the aim is to make them a simpler alternative to our Discourse based chatbot.
But walking back from the BCS/RSI event with Rod Brooks (Director CSAIL at MIT) I had a sudden brain-wave as to how this mimic technology could be used to do something that I've been thinking about for ages - and do it incredibly elegantly and in a way that for the moment is only possible in a virtual world.
So yesterday I found the time to cut the code and put it into operation. It's a real struggle not to say a lot about it now, but I want to get a few months data before making it public. I'm not saying it will revolutionise AI (it won't), but it might really open peoples eyes as to the possibilities.
