Slashdot Mirror
5 Predictions for 2012
Structured Audio writes "Mike Langberg of the Merc put up his 5
technology predictions for 2012. Well
chosen, although of course in 2012 speech
recognition will still be 10 years away :-)."
← Back to Stories (view on slashdot.org)
what on that list is impressive? He could've at least gone out on a small limb and mentioned fuel-cells.
Brevity is the soul of wit
-- Polonius
The characters, in the interest of drama, usually furrowed their brows, looked up at no one in particular, and assumed a deeper, "commanding" tone to their voice when addressing the computer.
Perhaps the computer uses a recognition algorithm based on all these factors to know when someon is talking to "it." My cat apparently has similar algorithms programmed in- I can speak in a conversational tone all day, even to a telephone, and the cat won't respond- but the moment I assume my "talking to kitty" voice, it snaps to attention.
Another interesting question about the ST computer- how did it route the person-to-person commnications before the individual spoke the receipient's name? You'd often here Picard's communicator pipe up: "Riker to Picard- you should come up to the bridge," or some such line.
One presumes that the communication did not go to everyone on the ship, only to be cut off when the word "Picard" was spoken. I always assumed the computer cached the outgoing communication until it was determined whom it was going to, and then retransmitted; the result should be a 1 second lag on the return to represent that, unless the computer subtly timeshifted the entire conversation to pad the lag into the spaces normally between words.
Marc Siry || interactive media professional, motorcycle enthusiast ||
I don't have any faith in speech recognition.
I'm already using it, and it drives me nuts. When I add money to my "pay and talk" cell phone account, I am forced to use their new (within the past couple of months) speech recognition menu. I am literally answering a robot's questions, and she makes me want to bang my phone on the closest solid structure near me.
Welcome to Rogers At&T pay as you go service, would you like to add money to your account today? Please say yes, or no.
Yes
I'm sorry, I didn't hear you, would you like to add money to your account today?
Yeeeees.
I'm trying to ask you a question. Please answer with yes or no.
YEEEEEES
this is the part where I wonder if swearing at the system will make it work. Maybe it recognizes "i said yes you piece of shit android" No it doesn't (I tried), but it usually takes about 5 tries, and I get into the "add money to my account menu" where i can then use the keypad (still) to enter in my P.I.N., new card number, etc.
You have paid for a total of 0 pages and so far 0 have been used up (0 today).
He obviously hasn't thought this one through; he's claiming that in 2012, anybody will be able to steal your credit card and then walk out of stores with anything they want, and there will be no impetus on the store to actually verify that it is you? Doesn't sound like a viable business model to me! Winona would then just claim that somebody stole her RFID and still steal from Saks!
"Freedom means freedom for everybody" -- Dick Cheney
This is wrong, depending on your definition of "proper."
/perfect/ speech recognition, as well as speech understanding, which is also known as "The Natural Language Problem," as long as we are both copying and pasting from link you posted.
What's AI-complete is
The reason perfect speech recognition is AI-complete is that it requires perfect speech understanding to choose between homophones (words that sound the same but have different meanings, such as to, too and two, or there, their and they're), and that problem is AI-complete.
Of course, most humans don't have perfect speech understanding, and hence also don't have perfect speech recognition.
Satisfactory speech recognition may indeed be as close as 2012. Maybe not, but it is possible. It has come a long way recently.
This has nothing to do with the Turing Test, and many people are of the belief that the Turing Test is a pretty silly milestone in AI anyway as it is a poorly formed, incredibly subjective measure of intelligence.
Passing the Turing Test is a matter of being able to fool a human into thinking you are human via a simple converstion, held with words only, sight-unseen. Some humans fail the Turing Test and some computers can already pass a limited variation of it.
Something that is AI-Complete is believed to require human-level intelligence to solve, and is an entirely different, and likely far more complicated problem.
Justin Dubs