Skip to main content
GameDev.net gamedev.net
🔒 Locked

Things computer can't do and humans can

Started by Lode Jan 12, 2006 at 11:05 AM 61 replies 24.4k views
Original Post
Lode
Lode
There are so many things that computers can't do properly while for humans they're very simple: -speech recognition -recognising objects in images that are in different positions, rotations, ... -seperating background from foreground -having a conversation that makes sense and actually knowing what a text is about -make a computer game from scratch -... Why is it that our brains can do these things, while it's impossible for us to come up with proper algorithms that can do this in a way as reliable as a human can? Is it because computers need to be faster first, or more parallelism is needed? Do you think that one day good algorithms will exist? [Edited by - Lode on January 12, 2006 11:29:51 AM]
dorix
dorix
I'd have to guess that it's because we don't fully understand how we do these things ourselves yet.
Fruny
Fruny
Quote:
Do you think that one day good algorithms will exist?


There are things computers cannot do, regardless of how much power you give them, and how powerful algorithms you design.

Can humans solve the Halting Problem?
"Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it." — Brian W. Kernighan
capn_midnight
capn_midnight
think outside the box

because computers are always IN the box! Hooooo!

edit: I think GDNet must have a bad joke filter, because the site 500 errored out when I tried to post this. if 404 is "file not found", is 500 "funny not found"?
levjs
levjs
The obvious answer, computers are unable to recognize or recreate humor or emotions, and there's no algorithm for that.


Quote:
edit: I think GDNet must have a bad joke filter, because the site 500 errored out when I tried to post this. if 404 is "file not found", is 500 "funny not found"?

Though maybe the GDNet filter programmers thought of one. [smile]
markr
markr
Spread butter on bread.

No, really. It's very difficult to build a robot which can butter bread. It just doesn't have the ability to spread the butter evenly without tearing the bread.

Mark
levjs
levjs
Quote:
No, really. It's very difficult to build a robot which can butter bread. It just doesn't have the ability to spread the butter evenly without tearing the bread.

I take it you've tried? [smile]

[Edited by - levjs on January 12, 2006 2:29:30 PM]
Lode
Lode
Quote:
Original post by AnonymousPosterChild
Love.


That's not a skill.
Agony
Agony
Quote:
Original post by levjs
But it's something computers can't do.

[overly-cynical]Neither can humans.[/overly-cynical]
"We should have a great fewer disputes in the world if words were taken for what they are, the signs of our ideas nly, and not for things themselves." - John Locke o
Max_Payne
Max_Payne
Quote:
Original post by levjs
But it's something computers can't do.


I think they can. Just not the kind of love I'd want.

Looking for a serious game project?
www.xgameproject.com
Stonicus
Stonicus
It's possible to have puters that can do speech and text recognition and can identify objects in pictures, but it's more than just an algorithm. It takes a human many many many years of training and processing input to be able to do it. I would take a guess and say the algorithms needed to accomplish it are probably simple. It just requires an insane amount of time and data to give the computer enough scenarios to use as a comparison. Also, a lot of what we can recognize in a picture is due to context. If you see something fuzzy in a picture, but the fuzziness is under the wing of an Air Force Fighting jet, your brain won't even consider that the fuzziness is a picture of a rocking chair, but probably a missile of some sorts. A computer would require a sense of context to aid in it's ability to eliminate possibilities.

It's easier to fill in the blanks using context.
"S__ir__y __ _e_v_n" What is this saying?

Here's a hint: Led Zeppelin.
How much easier was it to figure out?
Kallisti
Kallisti
You ought to read two books:

  • Goedel, Escher, Bach by Douglas Hofstadter

  • The Emperor's New Mind by Roger Penrose



Two rather opposed perspectives on this kind of thing. Both fun books, too.

(And if you manage to reconcile them, let us know.)
Discordian, yo.
Perfunction
Perfunction
Computer telephony companies such as Genesys have some excellent speech recognition software. Although their software is designed for call-center applications, I'm sure that if it were ability learn it would have amazing results. I think the only thing holding back our speech recognition efforts is the expansiveness of human memory. We start absorbing words from the day we are born so by the time we get to a mature age we have a huge bank of vocabulary and phrases.
Stonicus
Stonicus
It's also a HUGE step from recognizing the speech to processing it.

"Uh, yeah, I was... I called up to see if I could, uh, lemme see here... Yeah, I want to upgrade my DLS Cable modem with you guys. Hmmm, like, how do I do that?"

That would be a nightmare to programatically figure out and interpret.

AndreTheGiant
AndreTheGiant
Quote:
Original post by Lode
There are so many things that computers can't do properly while for humans they're very simple:
-speech recognition
-recognising objects in images that are in different positions, rotations, ...
-seperating background from foreground
-having a conversation that makes sense and actually knowing what a text is about
-make a computer game from scratch
-...


You could think of these as things that computers cant do...

or you could think of them as things that humans cant do. If a computer cant recognize speech, its because theres no human smart enough to program it to do so. By the way, computers can do most of the things in your list.

As far as a computer "knowing" what a text is about, that gets into a philosophical debate about what the definition of what the hell you mean by "knowing".


Sorry, but this topic gets a big thumbs down for lameness
Dmytry
Dmytry
I think that should be
Things that humans can't do(now):
write algorithm for (really good) speech recognition.(well, in fact it can be done)
write algorithm for recognising objects in images that are in different positions, rotations.(it is done already though, simply doesn't work with databases of really many objects)
write algorithm of having a conversation that makes sense and actually knowing what a text is about.(harder than above, also actually knowing is ill-defined)
write algorithm that makes a computer game from scratch.
Write algorithm that can write such algorithms.
etc...

edit: whoops, will teach me to read the thread. Beaten by above.
Stonicus
Stonicus
I don't think it's so much the algorithm, but the database of previous experiences. For the most part, the brain uses brute force on a constantly updated and sorted database. Our "memories" and "experiences" are constantly weighted and sorted by chronology (recent events) and commonality(how often do we use this data). Which is why, when watching a football game, people see every remotely roundish object (most notably empty 2 litre bottles) as footballs and start throwing them around. The algorith itself would be very simple. It's the database of knowledge that is hard to fill.
Dmytry
Dmytry
Well, you need some algorithms to create and process those "databases". Database itself is not a problem, in days of terabyte storage.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.

Sign in to reply to this topic.