On New Year's Eve, it is customary to take stock and make plans for the future. So we decided not to break the tradition and contribute to the general trend.
We all now often hear the phrase "artificial intelligence". Internet, television, advertising - an avalanche of information falls on us from everywhere. Home appliances with artificial intelligence, voice assistants in the phone. We are promised that artificial intelligence will solve all our problems and make our life better.
And we decided to evaluate how intelligent the existing AI systems really are, and whether they can even be called intelligence.
Currently, artificial intelligence is usually divided into two types - weak / narrow AI and strong / general AI (in the literature it is often referred to as AGI).
Narrow AI are algorithms designed to solve one specific problem. For example, pattern recognition, a borrower's bank assessment, or text translation. We will not consider such algorithms in this article - they have their own ratings and quality criteria.
A strong or general AI is an algorithm, or a set of algorithms, capable of acting under conditions of uncertainty and finding solutions to problems for which it was not originally programmed to solve.
What does an ordinary person think about when it comes to artificial intelligence? Most likely, he represents a system that communicates with him in a pleasant female voice, performs his tasks and answers questions. To a large extent, this image is shaped by science fiction films. For example, the Skynet system from the beloved Terminator movie with Arnold Schwarzenegger.
What is closest to this idea in our daily life? That's right - voice assistants: Ok, Google, Alexa, Siri and so on.
We want to assess how close the existing voice assistants are to general artificial intelligence, for example, to the very Skynet (as we imagine it).
You can say - “Why reinvent the wheel, everything has already been invented before you. There is also the Turing test in its various variations ”.
The Turing test is there - it's true. But:
1. , . , , , , . , . .
2. – «» « ». .
3. . ?
? . , , - . , - , , . – , .
, ( , ), ( , ). , .
() . , , «» ( , ) :)
, , , , , , . IQ-. IQ- .
IQ- () . .
IQ- , , , . , – , , , . , .
, -
1. - , . , - – , , . . (, ).
2. – , , . , , , .
3. - , . - .
4. — , . ().
:
1. – .
2. – .
3. – , - .
4. – . , , , , . «» , .
, , «» «, » - , , , . , . « 3», -, .
. 2000- «» – , . «» - . , : , . .
, . ?
, «» , –
, . ?
- ( - – , ) , - . GPT-3, , – « ».
, , – , . ( ) . , .
, , - - , , « », .
, , 1+.
2.0
, .
, – , . : , ? , , , . , .
?
: «» , ?
, - ( , , , , , ). – , . – .
- , . :
1. – . , , , , . , – - . , «if». . , , - .
- . , . . , . , .
– , . – « ». , . . «» - .
, , ( ), . , ? ? , . , , , .
« ». , – ( ) . , – , - , – . , « ?», «». , , .
, – « ». , . , , .
, . ? ? , , . , ( ), - , . . , – , , . .
, . , , . .
. , . - . .
, , , . , , 7 . . , «» . , , «» .
, . , «» . «», . , . , – .
? – . , . - . , , , .
, , 7- , , . ? , . , .
, , . , ? , «».
?
, , . , . - . 5-7 , . , .
- , , . , . 5-7 . , , : - – , ( , ) .
– . – 20 - , . , . , , . , , .
, .
, – , - .
, . , ( ). , , .
, ? , , , , , . , , – , , , .
, , :
1. ;
2. ;
3. .
, , . , , – , , .
( , ), , , ( , ). .
, - , .
, - 2 . . , , . «, »- .
AI of the third level will reason differently - Man will fill the entire planet with garbage -> most of the animals and plants will be destroyed -> the state of the "Planet Earth" object will become negative. This cannot be allowed, but what to do? Find a new way to recycle waste or destroy humanity?
I wonder what he will choose.
Fourth level - Imagination
At this level, AI will learn to create new objects with previously unknown properties in the real or virtual (if we all move there together by that time) world.
And then the fun begins. Or the worst thing. We'll see...