Is there any way to measure Artificial Intelligence?

“Intel­li­gence can­not be mea­sured using a scale.” I assume we have all heard that. How about mea­sur­ing arti­fi­cial intel­li­gence, though?

It is true that the con­stituents of intel­li­gence can­not be added togeth­er to form a sin­gle, large num­ber. But why not mea­sure the AI’s per­for­mance through­out dif­fer­ent aspects of intelligence?

For instance, if an AI is able to sur­pass humans in an IQ test, this sug­gests that it is par­tic­u­lar­ly adept at pro­cess­ing tasks relat­ed to spa­tial rea­son­ing.[Spa­tial rea­son­ing is the abil­i­ty to think about and manip­u­late objects in three dimen­sions.]

We can say that an AI has good OCR skills if it sur­pass­es humans in captcha recog­ni­tion — and so forth.

Here are some like­ly ways to mea­sure the Intel­li­gence of an AI:

  1. The Tur­ing Test, of course
  2. Behav­ior recognition
  3. Exem­plar Based Sys­tem (EBS) in Nat­ur­al Lan­guage Processing
  4. Arti­fi­cial Gen­er­al Intel­li­gence (AGI)
A. The Turing Test
A. M. TURING’s abstract exam­ple for describ­ing dis­crete state machines {Source: Mind, Vol­ume LIX, Issue 236, Octo­ber 1950, Pages 433–460,}

The first step in deter­min­ing an AI’s intel­li­gence is to see if it can speak in human-like tones. This is The Imi­ta­tion Game, or the Tur­ing test. In an arti­cle writ­ten in 1950 and pub­lished, Alan M. Tur­ing explained it as follows:

“Let us sup­pose that a human inter­roga­tor car­ries out a test of intel­li­gence by try­ing to decide, for each suc­ces­sive out­put from the machine, which of four pos­si­ble con­trib­u­tors is responsible.”

This def­i­n­i­tion enables us to deter­mine that his def­i­n­i­tion con­tains the two com­po­nents of decep­tion and intel­lect. We say that a machine has passed the Tur­ing test if the lis­ten­er can­not iden­ti­fy from their dis­cus­sion which one is a machine and which one is anoth­er human.

The Tur­ing test has been a good way to judge the intel­li­gence of an AI, but it is still not reli­able. If a per­son can under­stand the AI very well and try to fool them, we will not be sur­prised. And if some­one can­not fool the AI, then this is already above human lev­el by definition.

B. Behavior Recognition

There are numer­ous approach­es to mea­sure behav­ior recog­ni­tion, including:

  1. The Visu­al-Gaze Response in Eye-track­ing technology/Biometric recog­ni­tion in palm print, face, and fin­ger­print to ensure that all peo­ple get equal treatment.
  2. The unique pass­word to assure fair and equal treat­ment for all people.
  3. Voice recog­ni­tion to make sure that every­one is treat­ed with respect.

Although it will be very dif­fi­cult for AI to manip­u­late peo­ple using the three afore­men­tioned ways, there are still oth­er tech­niques we can use to mea­sure the effi­cien­cy of the AI, includ­ing the online anti-phish­ing algo­rithm, the online anti-fraud algo­rithm, and the online adver­tis­ing algo­rithm. These three meth­ods are infal­li­ble to human decep­tion, but they are nev­er­the­less essen­tial for mea­sur­ing an AI’s intelligence.

C. The Natural Language Processing Exemplar Based System (EBS) for measuring Natural Language Processing (NLP)

Here are some exam­ples of EBS:

  1. Sen­tence cor­rec­tion: By using con­text, we can cor­rect the sen­tence as best we can.
  2. Spelling Cor­rec­tion: Terms which need to be cor­rect­ed are select­ed depend­ing on their close­ness to oth­er words, the con­text of the phrase, and how often that words appear in a cor­pus of languages.
  3. Error cor­rec­tion: Words that have been cor­rect­ed will be picked depend­ing on the fre­quen­cy in a cor­pus of lan­guages and their ortho­graph­ic sim­i­lar­i­ty to oth­er words in the corpus.

The EBS can be used to assess a machine’s intel­li­gence, par­tic­u­lar­ly in the area of NLP.

D. Artificial General Intelligence (AGI)

An AGI is hard to con­struct, but once it is, we can use it to mea­sure an AI’s IQ. An AGI that we can­not con­trol has an IQ that may be very dif­fi­cult to quan­ti­fy, but once we have pow­er over it, it will be the ide­al tool for mea­sur­ing intel­li­gence. The more intel­li­gent you are, you can con­trol your envi­ron­ment and oth­er peo­ple more effectively.

Arti­fi­cial intel­li­gence is becom­ing stronger and more intel­li­gent every day. Sci­ence has pro­gressed quick­ly and improved the com­fort of our life. But final­ly, we will cross a line beyond which tech­nol­o­gy will be able to dom­i­nate us rather than the oth­er way around.

The Accuracy Factor for measuring Intelligence

When assess­ing arti­fi­cial intel­li­gence and find­ing out whether or not the AI is capa­ble of mak­ing impor­tant deci­sions, accu­ra­cy is a key con­sid­er­a­tion. The abil­i­ty of AI to decide things will improve as it grows more accurate.

Tasks requir­ing human-lev­el intel­li­gence (like play­ing chess) become eas­i­er for machines to achieve as tech­nol­o­gy advances into an era of “machine learn­ing”. As a result, we can fore­see that machines’ judg­ment will improve.

A vari­ety of tech­niques, includ­ing Clas­si­fi­ca­tion Accu­ra­cy, Con­fu­sion Matrix, Log­a­rith­mic Loss Mean Absolute, F1 Score, and Area Under Curve, Error, Mean Squared Error and more, can be used to eval­u­ate a machine’s accuracy.

One prob­lem with mea­sur­ing machine intel­li­gence is that there are many dif­fer­ent fac­tors we need to take into account when eval­u­at­ing per­for­mance. These fac­tors include: how accu­rate the question/task is; how rep­re­sen­ta­tives each machine’s per­for­mance tru­ly is; and how well the var­i­ous machines com­pare with each other.

In addi­tion, if a cer­tain type of machine (e.g., a cog­ni­tive sys­tem or a sta­tis­ti­cal mod­el) is being used, we would need to take into account the hard­ware and soft­ware that it may be run­ning on.


Will different AI’s have different level of Intelligence?

Just like dif­fer­ent humans have dif­fer­ent lev­els of IQ, the same will be the case for robots. Unless, you know they have some kind of a col­lec­tive con­scious­ness or some­thing like that.

If we cre­at­ed each and every robot with equal lev­el of intel­li­gence, it would not make sense any­ways. If an intel­li­gent robot is built in the first place, it won’t just start build­ing oth­er robots more intel­li­gence than itself. Because who would, of course.

If the robot was intel­li­gent, it would love to stay at the top, rather than cre­at­ing some­thing more powerful/intelligent than itself.

The Tur­ing Test is not the only way to mea­sure intel­li­gence. There are many oth­er meth­ods that we can use to mea­sure it. Some of them can be based on face recog­ni­tion tech­nol­o­gy, pass­word ver­i­fi­ca­tion or online anti-phish­ing algo­rithm. But no mat­ter what kind of method you use, the Intel­li­gence score will vary a lot for every sin­gle machine and in many cas­es, it is also very controversial.

Why measuring intelligence of an AI is going to be really important?

Mea­sur­ing the Arti­fi­cial Intel­li­gence is going to be a key part of our future with AI. AI will be able to do many things, such as dri­ving cars and fight­ing bac­te­ria. But just because these machines can do it does­n’t mean they should. We need to know exact­ly how intel­li­gent each machine is and then decide whether we want them in our lives or not.

The most impor­tant thing is to under­stand that the future of AI is unpre­dictable, no mat­ter what the media says. We have a lot of peo­ple wor­ry­ing about how quick­ly AI can takeover our lives, but they don’t real­ly under­stand how AI real­ly works. In fact, we don’t even know what kind of AI we are going to unveil in the future.

Is there going to be a time when robots will evolve into hav­ing an intel­li­gence sim­i­lar to that of humans? Is that even pos­si­ble? Are we going to build robots with human-like con­scious­ness? If so, will they be able to cre­ate oth­er robots more intel­li­gent than them­selves as well? Will these new species of super-intel­li­gent robot over­lords replace humans? Aren’t you afraid of yet?

Well, being able to mea­sure the “INTELLIGENCE” of an Arti­fi­cial Intel­li­gence is going to be the key for us to rest assured and choose the future we want. If it devel­ops in an unreg­u­lat­ed man­ner, the sce­nario is going to be dis­as­trous.

Relat­ed Readings:

Final lines,

So, the more we know about AI and how it works, the bet­ter we can take con­trol of our future. We need to under­stand them in order to be able to man­age them. And the best way of mas­ter­ing AI is to mea­sure their intel­li­gence through var­i­ous tech­niques. The tech­niques include word analy­sis and machine learn­ing as well.

Leave a Reply

Your email address will not be published.

Join our NewsletterDaily Glimple of Future

Our blog, "Daily Glimpse of Future", strives to make the future much clearer than it is today. Join our newsletter for free now.