Blog: Artificial Intelligence — The Definition & Tests
The Biggest Problem is that We/Humans don’t have a definition of What is Intelligence?
Our inability to (i) understand, (ii) define, and (iii) create intelligence probably stems from the limitations of our brain and our presumption that we are intelligent (as a species). The latter may or may not be true once we (i) understand and define intelligence. Lets hold on from discussing that for a while.
The Human Species is very Narcissist because it believes (i) It is the most intelligent creation of our GODs. And that (ii) The Definition of Intelligence itself is bench-marked against them.
On similar lines…
The New Era Science is very Narcissist because it believes (i) it has some magical ingredient because of which whatever it has achieved in the last ~400 years could have never been achieved by anyone else in billions of years of the existence of this Planet.
ALL the attempts to define intelligence can be broadly categorized into two categories
- By using “Observed” Behaviour as the proof/evidence of Intelligence
- By “Internal” states & functionalities of humans & machines by which we can understand if its Intelligent
Without attempting to and defining [Artificial] Intelligence and succeeding. The Scientists and Academics simply went ahead and benchmarked it with Human Capabilities. And we said…
In the field of artificial intelligence, the most difficult problems are informally known as AI–complete or AI-hard, implying that the difficulty of these computational problems is equivalent to that of solving the central artificial intelligence problem — making computers as intelligent as people, or strong AI.
We defined 3 broad terms to describe “Capability Levels” of [Artificial] Intelligence (Mostly Marketing & Sales, Laymen Jargon)
Artificial General Intelligence (AGI) is an AI that matches human intelligence.
Artificial Superintelligence (ASI) is an AI that exceeds human capabilities.
Lets look at some definitions of [Artificial] Intelligence. With our comments alongside them.
1 One of the first and foremost definitions and tests of [Artificial] Intelligence was proposed by Alan Turing in 1950.
The Turing test, is a test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human.
The way it works is as follows…
Suppose you are having a chat session with a person and a computer, but you are not told at the outset which is which. If you cannot identify which of your partners is the computer after chatting with each of them, then the computer has successfully imitated a human. If a computer succeeds in passing itself off as human in this “imitation game” (or “Turing Test” as it is popularly known), then according to Turing, we should be prepared to say that the computer can think and can be said to be intelligent.
Comment— “So Turing side-stepped the question of somehow examining the internal states of a computer by instead using its behavior as evidence of intelligence. By the same reasoning, we have assumed that in order to say that a computer understands English, it just needs to behave as though it did. What is important here is not so much the specifics of Turing’s imitation game, but rather the proposal to judge a capacity for natural language understanding in terms of observable behavior.”
2 “general intelligence” refers to the capacity for efficient cross-domain optimization
Comment — This stems from the idea that niche intelligence would be limited to one domain and general intelligence would be across domains. It is a sort of discrimination or classification between two categories of intelligence. But nothing more.
3 the ability to achieve complex goals in complex environments using limited computational resources.
Comment — This definition stems from the scientific experience that Intelligence is a hard problem, and like the hardest problems in Computer Science, it will definitely be NP, NP-Complete or NP-Hard. And hence would take exponential amounts of compute and time. And hence this definition asserts not just that intelligence is the ability to solve complex goals in complex environments. It says we should be able to do it with limited computational resources. It assumes Humanity would make a breakthrough and solve NP-Complete Problems in Polynomial Time for this to happen. (Which is exactly what Automatski has achieved in the last 2–3 decades. And why Automatski is a front runner in delivering A.I.)
4a branch of computer science dealing with the simulation of intelligent behavior in computers. — merriam webster
5 The goal of work in artificial intelligence is to build machines that perform tasks normally requiring human intelligence. Nils J. Nilsson
Comment — These definitions don’t even attempt to define Intelligence in terms of the two broad categories we have seen earlier. They simply say A.I. is a practice. Its an effort towards “a” goal.
6 “Artificial intelligence is that activity devoted to making machines intelligent, and intelligence is that quality that enables an entity to function appropriately and with foresight in its environment.” Nilsson, Nils J.
Comment — Here Nilsson extends his own definition of [Artificial] Intelligence, mentioned in the paragraphs above. And says [Observed] how something intelligent will function in its environment.
7 Artificial intelligence is concerned with the attempt to develop complex computer programs that will be capable of performing difficult cognitive tasks. Eysenck, Michael W. (1990)
Comment — Just like Nilsson’s extended definition above, Esyenck says that we will need to develop [Internal] extremely complex software which will be [Observed] capable of performing complex cognitive tasks. Since we have not been able to create [Artificial] Intelligence Esyenck assumes rightly or wrongly that it would be far more complex than anything we have now, asserting — which is why we have not been able to develop it so far.
8 My definition of AI is any algorithm that is new in computer
science. Once the algorithm becomes accepted then it’s
not AI, it’s just a boring algorithm. — Dick Keene
Comment — This statement says “Any New Algorithm” is considered A.I. but once when its novelty fades away it relegates back to the larger collection of usual algorithms. This statement has a basis in the fact that every organization and individual who develops anything seemingly new hypes it as A.I. (since mankind doesn’t have a definition for that) and once everyone else catches up. It just becomes a well known trick in the bag of tricks which humanity already has.
9 AI is making computers act like those in movies — Ralf Brown
Comment — This statement stems from the observation that Hollywood/Science-Fiction is usually a pre-cursor to innovation. In a way it defines the end goal [Observed] implying in reverse that [Artificial] Intelligence would lead to it.
We can see some more definitions in the article below. But like earlier. They all reason in similar ways. And ultimately fall back to the presumptuous benchmark with Human Capabilities.
1 — Any definition with broad utility
2 — Any definition which is non-ambiguous
3— Or, any definition which is both necessary and sufficient.
The world (including marketing) seems to have hit peak-A.I. Marketers are either marketing to us with A.I. (with a view…digitalwellbeing.org
Lets look at some of the tests proposed for [Artificial] Intelligence.
The Turing test ($100,000 Loebner prize interpretation)
One specific interpretation is provided by the conditions for winning the $100,000 Loebner Prize. Since 1990, Hugh Loebner has offered $100,000 to the first AI program to pass this test at the annual Loebner Prize competition. Smaller prizes are given to the best-performing AI program each year, but no program has performed well enough to win the $100,000 prize.
The exact conditions for winning the $100,000 prize will not be defined until a program wins the $25,000 “silver” prize, which has not yet been done. However, we do know the conditions will look something like this: A program will win the $100,000 if it can fool half the judges into thinking it is human while interacting with them in a free form conversation for 30 minutes and interpreting audio-visual input.
Comment — this is very funny in many ways. While a scheme has been proposed. The scheme has not been defined in entirety yet. This is adhoc, ambiguous, and has literally no utility.
The coffee test
Goertzel et al. (2012) suggest a (probably) more difficult test — the “coffee test” — as a potential operational definition for AGI:
go into an average American house and figure out how to make coffee, including identifying the coffee machine, figuring out what the buttons do, finding the coffee in the cabinet, etc.
If a robot could do that, perhaps we should consider it to have general intelligence.
Comment — This test basically says we would have created [Artificial] Intelligence when and only when. When like GODs we can create Intelligent Robots and Machines in our own image. Anything less is “Not” A.I. This is adhoc, ambiguous, and has literally no utility.
The robot college student test
Goertzel (2012) suggests a (probably) more challenging operational definition, the “robot college student test”:
when a robot can enrol in a human university and take classes in the same way as humans, and get its degree, then I’ll [say] we’ve created [an]… artificial general intelligence.
Comment — Again! This test basically says we would have created [Artificial] Intelligence when and only when. When like GODs we can create Intelligent Robots and Machines in our own image. Anything less is “Not” A.I. This is adhoc, ambiguous, and has literally no utility.
The employment test
Machines exhibiting true human-level intelligence should be able to do many of the things humans are able to do. Among these activities are the tasks or “jobs” at which people are employed. I suggest we replace the Turing test by something I will call the “employment test.” To pass the employment test, AI programs must… [have] at least the potential [to completely automate] economically important jobs.
Comment — Yet Again! This test basically says we would have created [Artificial] Intelligence when and only when. When like GODs we can create Intelligent Robots and Machines in our own image. Anything less is “Not” A.I. This is adhoc, ambiguous, and has literally no utility.
Universal Intelligence Quotient
The first premise is that Intelligence should be a universal concept. Applicable to both machines and humans (+ other species). And any general measure of intelligence aka I.Q. should correlate very well with both Observable and Internal aspects of Intelligence. And hopefully should be comparable across Species.
Universal Intelligence Quotient Test
The Test(s) of Universal Intelligence should derive from the definition of the concept of intelligence. And should reflect the definition.
In the absence of any usable definition of intelligence. The current standard IQ tests are hence therefore by definition inappropriate and irrelevant.
The Automatski — Pyramid of A.I. [A Definition of Universal Intelligence]
The Biggest Problem in A.I.
Artificial Intelligence has ‘NOT’ been invented by mankind yet. Atleast not outside Automatski.
The lowest level of existence of Intelligence is just a set of ‘skills’. Skills can be audio, visual, mobility, flight etc. Skills by themselves mean ‘nothing’ in terms of intelligence, but they are the foundations of intelligence, upon which intelligence is built.
The first level of machines is the ability to control and execute the skills it has. Example a machine could mow a lawn or clean the floor by itself.
Inference is the first real level of intelligence. This is when machines or humans, collectively called Actors from now on, can make inferences from data, or the world around them. When they can figure out the model (model generation) or the essence of something, a situation, a problem etc. And then figure out the best course of action to achieve their goals, solve the problem etc.
The next level of intelligence is when Actors, based on their goals, can generate their own hypothesis (speculate reasonably) about a situation, and how to achieve those goals. And take appropriate actions.
Which is further extended by a collection of heuristics, hypotheses, guesses, approximations, fuzzy and probabilistic methods.
Then comes the next jump in intelligent thinking. The way humans do. Reasoning using Logic but also similarities, analogies, types and categories, meta-learning (higher order learning) and learning transfer from one situation and domain to another.
The next two levels of intelligence basically involve being able to do multiple tasks using some shared and common + specific knowledge of those tasks. And being able to work with multiple objectives in multiple contexts at the same time.
The next level of Intelligence involves evolution which is simply put as Learning, Unlearning and Relearning. But this is not possible without a self capability to assess the Quality of your own knowledge and learning. But at this level we achieve this by using an external actor which guides our learning, unlearning and relearning. In a somewhat occasional reinforced fashion.
The next level completes this area by implementing Self Assessment, Reflection and Reasoning.
Till now everything was done using some internal representation. But at this level we start with the concepts of Vocabulary, Language, Concepts, Beliefs, and Desires.
The next step in the evolution of intelligence is the use of natural interfaces. Till now the system could have used fixed, system interfaces to perform its functions.
And now the next step is Universal Mobility capabilities.
The next two levels involve becoming self aware, and creating a Mind and Soul.
Now our intelligence is ready to be Autonomous.
And it can now collaborate with other Actors, using shared goals and purposes.
The last and the final stage of Intelligence is Reproduction and/or Replication.
The Automatski — A.I. Pyramid is both necessary and sufficient, quite comprehensive. It is reasonably unambiguous and utilitarian. And forms the basis of the worlds first Artificial General Intelligence developed at Automatski. And hence is proven.
It also makes for a Universal Definition of Intelligence. And related Tests, both qualitative and quantitative.