Why AI isn't that bright

[LongBlade]

An interesting article about why computers really aren't all that smart.

Hector Levesque thinks his computer is stupid—and that yours is, too. Siri and Google's voice searches may be able to understand canned sentences like "What movies are showing near me at seven o'clock?," but what about questions—"Can an alligator run the hundred-metre hurdles?"—that nobody has heard before? Any ordinary adult can figure that one out. (No. Alligators can't hurdle.) But if you type the question into Google, you get information about Florida Gators track and field. Other search engines, like Wolfram Alpha, can't answer the question, either. Watson, the computer system that won "Jeopardy!," likely wouldn't do much better.

Full article: http://www.newyorker.com/online/blogs/elements/2013/08/why-cant-my-computer-understand-me.html

[TheCommandTent]

Interesting read.  However, I am all for computers staying stupid.  No need for them to be overdeveloping their intelligence and turn on us.   ;D

[Rayfer]

Interesting read.  However, I am all for computers staying stupid.  No need for them to be overdeveloping their intelligence and turn on us.   ;D

Or maybe they already have and we are living in the Matrix?

[SgtRock]

Siri and Google's voice searches may be able to understand canned sentences

All they do is look up in a database what they think is said and give a response. The truth is this is no intelligence in AI, it's been said many many times if you could just give computers the intelligence of a fruitfly computers would 1000% smarter and about 3-5% as smart as the dumbest person on earth. IBM's computer Watson "Jeopardy computer" has a bunch of build in databases like the complete IMDB (movie database), it required a tech (man behind the curtain) to type in the question in a very specific formant, it would then search all of it's databases, then use statistics to rate each found reference, the reference with the highest rating was then given as the answer.

[Father Ted]

It's the nature of the beast.  Essentially the computer is a "yes" or "no" machine, but we're not like that.

[BanzaiCat]

[FarAway Sooner]

Another interesting article on this topic can be found at http://www.theatlantic.com/magazine/archive/2011/03/mind-vs-machine/308386/.

I remember an old-time friend of mine (who was one of two co-lead designers on Imperialism 1 & 2) saying to me more than a decade ago, "Computers aren't going to be good at artificial intelligence any time soon.  The real way game designers ought to solve that is to try to create games where a computer can do the things that it's good at and most of that is what passes for real human intelligence."

AI opponents have a very hard time with games like the Heroes of Might & Magic series, where units' special abilities combine with hero-cast spells to create some crazy syntheses (e.g., casting "First Strike" on a huge pack of Griffins who already have "Unlimited Counterattacks" can make that stack almost undefeatable).  By contrast, games like chess (with a more finite number of moves and counter-moves) are far easier for computers to model.

The trick then, is to design a game where the math is more like chess and less like Heroes of Might & Magic.

[jomni]

Good single player game design hinges on rules and gameplay elements that are tailored for easy and effective  development of a cunning Ai.

Overly complex game rules and features would make things hard for the AI programmer. Ever wonder why most board game conversions (non solitaire)  fail? That's because the board game rules were written for humans to interpret, enforce, bend, or break.

Reminds me of World in Flames.   Takes forever to program the game and we don't know if the Ai will see the light of day.

[GJK]

Computers can't apply emotional responses to a situation, only logic and reason.  They also lack intuition.  That's why computers are all male too I think, but that is a personal opinion.  :)

[Mr. Bigglesworth]

Siri and Google's voice searches may be able to understand canned sentences

All they do is look up in a database what they think is said and give a response. The truth is this is no intelligence in AI, it's been said many many times if you could just give computers the intelligence of a fruitfly computers would 1000% smarter and about 3-5% as smart as the dumbest person on earth. IBM's computer Watson "Jeopardy computer" has a bunch of build in databases like the complete IMDB (movie database), it required a tech (man behind the curtain) to type in the question in a very specific formant, it would then search all of it's databases, then use statistics to rate each found reference, the reference with the highest rating was then given as the answer.

I'd say the difference is :

   
• how the database gets generated
• the size of the database
• the number of connections between data elements

How many years does a newborn file away continuous data from the senses before being able to use that data? A few people have developed a filing system of memories that lets them goto the memories when prompted on a date. For everyone else they are still filed away. Responses are generated from memory of those sensory experiences.

[SgtRock]

Can we even have intelligence without consciousness? I think the idea of if we pack enough transistors in a chip or connect enough computers together they'll some how become conscious is just dumb. There is just no way with current hardware and software, maybe when we have quantum computers but, I'll bet it won't be until we move away from silicon & software.  Develop something like a biological brain in a box but, then it won't be an AI it will be something else.

Interesting read.  However, I am all for computers staying stupid

I agree, the day a computer starts telling me what to do, it's headed for the land fill ;D