There's a brilliant article that discusses this problem here: http://yudkowsky.net/rational/bayes. About a quarter of the way in there's a look at a few different ways of expressing the quantities. It seems frequencies are good (3 in every 1000 innocent people will be IDed as a terroist or 299 in every 300 people identified as a terrorist are innocent). People intuitively focus on the expected outcome - postive test result == terrorist and negative == not a terrorist. Maybe the way to make it clear is to tell them the non-intuitive statistics (299 in 300 that appear guilty are innocent while 1 in 30,000 that appear innocent are guilty). The issue is that if you tell someone "Q given P" (positive-result given is-terrorist) they always fall into the trap of thinking "P given Q" (is-terrorist given positive-result). Saying the test is 90% accurate is saying "Q given P 90% of the time". No one understands prior probability yet figures like this always ignore it.
Does anyone else from NI/Scotland remember when you first got into computers and were quite shocked that FTP was used as an acronym? I was surprised that lots of peole didn't get cross about it. Turns out that for the entire rest of the world FTP has always been file transfer protocol. If any of you ever go to Belfast or Glasgow you might wonder why there are protocol acronyms written on walls in large letters.
Only skimmed the article but I think it won't happen.
Too big a change. Like saying everyone in the USA should drive on the left of the road like UK. In principle you pick a day and say "Tomorrow we all switch sides". In practice far too comprehensive a change.
Without a law change of similar magnitude outlawing analogue 'old' technologies this won't happen. I think such an attempt would struggle as people will not obey laws they deem unreasonable, for instance speed limits.
Are there non-communication based ways of judging anything?
Just because you have to communicate with something to deem it intelligent doesn't mean the turing test is a fair measure of machine intelligence.
Forgeting the whole philosphical thing about weak and strong AIs I think the turing test is pretty good. Having said that I think a machine should have to convince *everyone* that it is human before being deemed intelligent to some level.
BBC's Tomorrows World programme did some big Turing test and 50 odd percent of people thought the computer was a person. Yeah, 50% of non-techy people who will ask the predictable questions. A lot of people could be fooled by Eliza if they don't have the knowledge to guess were the computer might trip up (jokes, emotions, questions about my previous comments, dodgy context etc). Doesn't mean Eliza on my Amstrad CPC 6128 was intelligent.
When I pronounce a/an I use the same rules you describe - I agree it sounds odd the other way. I am still sure that written formal English has different rules and that the commonly used rules for a/an are modernised versions of the originals.
Legality wise I assume the author is American and as far as I know their data protection laws are very scant, in comparison to the UK certainly. The point is that the book is available to an international audience and so the author can't really talk abou the legal issues.
Ethics is an interesting issue. Sure spam, whatever is a pain in the arse but one still requires a well strucured database to do it. Ethics are not an issue when it comes to writing a good database, whether the database is being used to store personal data for your hosipital patients or people you plan to kill is not an issue at a design level.
It seems to me to be a whole separate issue. Web design books don't discuss the ethics of creating an inflamatory site. I don't dispute the DB creator *should* be aware of these issues. I dispute whether you should criticise the author for omitting this topic.
Lots of people have mentioned the Big Blue thing but as far as I know Kasparov acknoledges he made a silly mistake in that final game. Basically I think the computer was lucky.
That certainly explains why IBM aren't so keen on a rematch.
From The Register (http://www.theregister.co.uk/content/2/13558.html )
The site, currently sitting on 3Com's public server, announces Audrey,
3Com's Palm-connectable appliance, aimed at women. According to the
site, Audrey will ship in a range of pastel colours - shades of Apple's
iMac here - and sport a 6.25in x 4.75in (the resolution is 640 x 480,
judging by the site's HTML) touch-sensitive screen; built-in 56kbps
modem, speakers and microphone; serial and USB ports; and a wireless
keyboard. Audrey is based on National Semiconductor's Geode CPU.
Audrey will ship with its own Web browser and personal information
management features. 3Com is basing some Ergo products on the
Palm OS, others on QNX's Neutrino realtime OS - we suspect the latter
is being used for Audrey, though the site gives no clues.
Slashdot Mirror
There's a brilliant article that discusses this problem here: http://yudkowsky.net/rational/bayes. About a quarter of the way in there's a look at a few different ways of expressing the quantities. It seems frequencies are good (3 in every 1000 innocent people will be IDed as a terroist or 299 in every 300 people identified as a terrorist are innocent). People intuitively focus on the expected outcome - postive test result == terrorist and negative == not a terrorist. Maybe the way to make it clear is to tell them the non-intuitive statistics (299 in 300 that appear guilty are innocent while 1 in 30,000 that appear innocent are guilty). The issue is that if you tell someone "Q given P" (positive-result given is-terrorist) they always fall into the trap of thinking "P given Q" (is-terrorist given positive-result). Saying the test is 90% accurate is saying "Q given P 90% of the time". No one understands prior probability yet figures like this always ignore it.
You're right. This is what bothered me about the sig in the first place, 1690 being the Battle of the Boyne and all that. It seems a bit out of place.
--
1690 - Never forget.FTP.
UID> 47? Beware.
Does anyone else from NI/Scotland remember when you first got into computers and were quite shocked that FTP was used as an acronym? I was surprised that lots of peole didn't get cross about it. Turns out that for the entire rest of the world FTP has always been file transfer protocol. If any of you ever go to Belfast or Glasgow you might wonder why there are protocol acronyms written on walls in large letters.
Only skimmed the article but I think it won't happen.
Too big a change. Like saying everyone in the USA should drive on the left of the road like UK. In principle you pick a day and say "Tomorrow we all switch sides". In practice far too comprehensive a change.
Without a law change of similar magnitude outlawing analogue 'old' technologies this won't happen. I think such an attempt would struggle as people will not obey laws they deem unreasonable, for instance speed limits.
"Sources tell KING5 News that the group that put the sculpture up in Magnuson Park is not the same group that took it down."
Blatantly it's people who thought it would be funny to move it. Foggy night, few drinks, 10 students...
Are there non-communication based ways of judging anything?
Just because you have to communicate with something to deem it intelligent doesn't mean the turing test is a fair measure of machine intelligence.
Forgeting the whole philosphical thing about weak and strong AIs I think the turing test is pretty good. Having said that I think a machine should have to convince *everyone* that it is human before being deemed intelligent to some level.
BBC's Tomorrows World programme did some big Turing test and 50 odd percent of people thought the computer was a person. Yeah, 50% of non-techy people who will ask the predictable questions. A lot of people could be fooled by Eliza if they don't have the knowledge to guess were the computer might trip up (jokes, emotions, questions about my previous comments, dodgy context etc). Doesn't mean Eliza on my Amstrad CPC 6128 was intelligent.
When I pronounce a/an I use the same rules you describe - I agree it sounds odd the other way. I am still sure that written formal English has different rules and that the commonly used rules for a/an are modernised versions of the originals.
You mention legality and ethics.
Legality wise I assume the author is American and as far as I know their data protection laws are very scant, in comparison to the UK certainly. The point is that the book is available to an international audience and so the author can't really talk abou the legal issues.
Ethics is an interesting issue. Sure spam, whatever is a pain in the arse but one still requires a well strucured database to do it. Ethics are not an issue when it comes to writing a good database, whether the database is being used to store personal data for your hosipital patients or people you plan to kill is not an issue at a design level.
It seems to me to be a whole separate issue. Web design books don't discuss the ethics of creating an inflamatory site. I don't dispute the DB creator *should* be aware of these issues. I dispute whether you should criticise the author for omitting this topic.
That certainly explains why IBM aren't so keen on a rematch.
(I) (B)lame (M)icrosoft?
The site, currently sitting on 3Com's public server, announces Audrey, 3Com's Palm-connectable appliance, aimed at women. According to the site, Audrey will ship in a range of pastel colours - shades of Apple's iMac here - and sport a 6.25in x 4.75in (the resolution is 640 x 480, judging by the site's HTML) touch-sensitive screen; built-in 56kbps modem, speakers and microphone; serial and USB ports; and a wireless keyboard. Audrey is based on National Semiconductor's Geode CPU.
Audrey will ship with its own Web browser and personal information management features. 3Com is basing some Ergo products on the Palm OS, others on QNX's Neutrino realtime OS - we suspect the latter is being used for Audrey, though the site gives no clues.
Apple Mac System 1. Doesn't look too different to System 7 to me...