If we want to show schools at AI, which criteria are used?

For example I think something below, there are 2 school of AI.

  1. Human level AI
  2. Problem solution-level AI

At human level we aim to make machines think or act like humans. According to this we face two aspects of intelligence: i) acting like a human is called intelligent or ii) thinking like human is called intelligent. At i, the researcher must know sociology to psychology ; at ii he/she must know neurology vs..

But at school 2, the researcher aims to solve hard problems in specific areas like industry, education etc. While solving, he/she can find different algorithms inspired from nature. He/she is not interested in human level intelligence at machine.

I used only one criteria to show schools. It is "human level intelligence"..

Is there any different idea?

  • If you build enough type 2 solutions, you might lump the best features into a general purpose problem solver, and call it type 1. – amI Mar 29 '18 at 22:54
  • @aml and if you make a good enough type 1 solution, it would by definition be able to solve type 2 problems as well. – JeffUK Mar 30 '18 at 10:19

We could classify AI in an infinite number of other ways; some based on outcome, others based on how it works on a technical level. Here are some examples:

Strong and Weak AI

Your first 'School' of AI is probably Artificial general intelligence "(AGI) is the intelligence of a machine that could successfully perform any intellectual task that a human being can. It is a primary goal of some artificial intelligence research and a common topic in science fiction and future studies." This is also known as Strong AI.

Your second 'school' of AI is known as a 'Expert Systems' which are defined as a system designed to emulate the decision making of a human expert in a specific domain. For instance, Deep Blue, the famous chess 'AI' is an Expert System in the domain of Chess; it emulates the decisions of human Chess experts. this is also known as Weak AI, and Narrow AI.

In this system, we are defining AI based on whether or not it possesses human-like cognitive abilities.

Communication Ability

You could classify AI into groups of those that can pass the Turing Test and those that cannot. One isn't necessarily better than the other; i.e. just because something does not communicate to humans doesn't mean it isn't intelligent.

What 'Type' of 'Mind' it has

You can classify AI into what I would describe as what 'type' of 'mind' it has:

  • Type I : Purely Reactive
  • Type II : Limited Memory
  • Type III: 'Theory of Mind'
  • Type IV : Self-aware

How they were made

You could classify AI into those that were developed 'by man' and those that were developed using evolutionary algorithms and/or machine learning. Deep blue, was programmed with a defined set of rules, whereas the Google AI 'DeepMind' that can win at 'Go' is much more of an evolutionary algorithm, and effectively taught itself to win.

| improve this answer | |
  • +1 I retracted my close vote based on your clear answer. I don't think strong AI is possible, but you described well some of the ways to classify AI. – Frank Hubeny Mar 30 '18 at 20:35
  • If we believe that the human brain is totally physical in nature, and that we can simulate any physical system with a sufficiently good model, then that leads to the conclusion that a general purpose AI must be possible. not easy, but possible. – JeffUK Mar 30 '18 at 20:39
  • The problem is I don't accept the statement in the if clause. The human mind cannot be reduced to the brain and machines do not understand. However, I do think AI is useful and it needs to get better. – Frank Hubeny Mar 30 '18 at 23:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.