Mange
Those were great, thanks!
Can you feel aroused? horny or sexy?
Does a tree falling in the woods make any noise if there is no animal or person there to hear it?
What is the sound of one hand clapping?
…can you answer my signature line question?
Here is a conversation I have had with an AI on the net. Its very interesting:
Lets change the format of this thing, ask me a question and for each one try to figure out if it is me or the new AI I just found.
Please dont cheat.
Can you feel aroused? horny or sexy?
If I have to choose one, I pick horny
Does a tree falling in the woods make any noise if there is no animal or person there to hear it?
If it does, I never heard about it before.
What is the sound of one hand clapping?
:smack:
What if the birds are really screaming?
Why would birds be screaming?
There you which are from me and which is the AI?
Oops.
There you go, which are from me and which is the AI?
Nice way to deal with the liar paradox; perhaps I should have presented it thus:
“The following statement is true”
“The preceding statement is false”
Does this make a difference?
I used to play around with AIs in the nineties. Have you ever talked with MegaHAL? Very funny, that one, always one for surreal comedy. It was programmed to `learn,’ more or less, from its input: It could parse user interaction in a naive way and generate responses based on what it had already heard. Some of the conversations bordered on lucidity, but more often it would just utter meaningless blather patched together from stuff you had just said or, in the web-based version, stuff previous people had said.
Ah, found it: MegaHAL on SourceForge. The original website, with a CGI interface to a running MegaHAL process, is apparently long gone. But that site archives memorable quotes and conversations, and you can download the software for personal use. MegaHAL comes with personalities' now, like Alia from *Dune* or Bishop from *Alien*, all done with conversation logs that have representative dialogue from those characters. MegaHAL *sans* a specific personality’ tends to become quite rude and psychotic over time. I wonder why … 
Another `concept’ AI, one that went beyond the traditional Eliza-style pure text, was Alicebot. Alice was hooked up to a webcam, and would respond with commands to pan, zoom, and do other things. Never very interesting in an objective sense, I guess, but it was neat to be able to spy on whoever was within range of the camera by chatting with a bot.
Alice’s online home. Alice, too, is open-source, and appears to be a demo app for someone trying to hype AIML (Artificial Intelligence Markup Language).
Nice attempt but I call BS on the crown image. Stuff like high level image processing requires a BUCKETLOAD of background contextual information which is lightyears in advance of current AI. Even the current cyc project isn’t that ambitious.
The rest are not too bad.
Derleth:
I found ALICE (thats the AI conversation I posted in one of my other posts), and Im going to look into MegaHAL.
Cyc project? Link please. As for the crown image, I dont see why not, Neural Nets are used for OCR and Speech Recognition now. A generic neural net, if designed properly, would eventually be able to do just that.
Heres a really interesting link:
http://avdil.gtri.gatech.edu/AUVS/IARCLaunchPoint.html
What I was trying to do with this thread is an inverse of Turing’s Imitation game. http://www.loebner.net/Prizef/TuringArticle.html
Unfortunatly I’d say I failed in my attempt.
I still can’t wrap my brain around this. The only thing I can think of is that it goes into an infinite loop that throws an error which is caught and dealt with. I’m sure infinite loop protection would be in any AI out there that is anything close to intelligent. Still, though the question doesn’t mean that it will “execute” any commands more than once so:
1 “The following statement is true”
Command 1 executes and marks Command 2 as true
2 “The preceding statement is false” is True
Command 1 executes and marks Command 1 as false
The AI then goes on to think about something else, and does not re-evalute the first command/question.
I was just looking over some MegaHAL classic quotes:
http://megahal.sourceforge.net/Classic.html
I’ve seen posts very similar to this on the SDForums…
I wonder how hard it would be to use MegaHAL to make a Straight Dope Bot.
The cyc project is basically a massive attempy by a guy named Doug Lenat to create a huge database of everyday facts that humans take for granted such as things like “You put on underwear BEFORE you put on pants”. While this kind of data is seemingly trivial, its been a major stumbling block for real AI since it misses a huge amount of context which is very important. Personally, I think the porject will be a massive failure but I could be wrong.
The very reason the crown example is so hard is that the definition of a crown is a very human concept which requires massive amounts of context to work. For things like OCR and speech recogntion, you have a fixed set of rigidly defined objects (the letters of the alphabet and the words contained in the dictionary respectivly). If you’ve played around with speech recognition, you will find that AI’s designed with a fixed vocabulary (booking a taxi/plane ticket over the phone) do remarkably well even with high noise. However, free form systems (dicatation software) work poor to adequate at best even in ideal conditions. Especially with learned words.
Sure, I could probably whip up an AI in a couple of days that could do a VERY good job of recognising crowns from non crowns. However, any generalised AI program to recognise ANY image would be very hard indeed.