2
$\begingroup$

This survey of Artificial Intelligence with Prolog provides a short perspective on the Prolog programming language and its position in the history of AI.

It proposes that a return to Prolog Research and Development as begun in the 1980s when Japan began work on a Fifth Generation Computer System. Has the reduction in cost and the increase in speed of computing platforms re-opened a door that was closed primarily because of technology immaturity?

$\endgroup$
  • $\begingroup$ One would assume that to create strong AI(or anything for that matter), one must have an idea on how to create/implement it. One might also assume that prolog is translatable to other languages, so I don't see how the lack prolog usage is impeding strong AI XD. It would quite the twisted reality if it was. $\endgroup$ – Daniel Oct 16 '18 at 20:08
1
$\begingroup$

The application of Prolog, the declarative programming language may or may not be closely related to AI strength. Prolog is a language that was designed to produce a Turing Complete backward chaining general programming language. The strength or weakness of Prolog programs, just as for programs in Scheme, Common LISP, Java or Scala with DRools rules engine, or neuromorphic libraries in Python or C++ depends on more than the language and libraries chosen.

The strength of an artificially intelligent system, from a mathematical perspective, is as ambiguous in meaning as IQ test results. Did human kind within one decade cross some critical level where they were once stupid and became intelligent or their weak intelligence became strong?There is no evidence for such discrete transition. Evidence indicates that DNA advancements, many of them, and cultural changes to increase intelligence capacities occurred over thousands or tens of thousands of years.

The same is true of general and narrow AI. All intelligent systems have strengths, weakness, generalizations, and narrows. This is true whether the intelligence arises out of billions of organic neurons and the brain's complex systems of chemistry, organelle structures, and plasticity capabilities or whether the intelligence arises from designs involving ones and zeros on a collection of interconnected silicon wafers.

Classifying systems into a four quadrant system is counter productive.

  • Weak narrow
  • Strong narrow — Extremely effective systems applicable to a narrow set of usage contexts
  • Weak general — Systems that show adaptive and proactive capabilities over a wide range of usage contexts but demonstrating only limited effectiveness in any given context compared to a typical human
  • Strong general

The binary system of weak = narrow and strong = general is even more limiting because of the existence of the middle two bullet points above in human populations.

Judgments about intelligence are ALWAYS

  • Relative to some standard,
  • Time dependent, and
  • Context dependent.

More importantly, there is more than one or two dimensions to intelligence — possibly hundreds.

Ken Thompson and Linux Torvalds, clearly the most successful operating system writers in terms of influence on currently deployed von Neuman computers may have great difficulty with writing a song that sustains sales over decades. Bono and Paul McCartney may have great difficulty with writing an operating system that is downloaded every minute for decades. Both features and breadths of intelligence have value in current culture and economics.

The appreciation of the Prolog programming language by those like the author of the page referenced in the question is not without reason. The design of the language provides an advantage for R&D over other languages in rules based contexts.

There may be value in reconsidering the value of the language in light of the significant improvements in computing platforms since earlier efforts were made to use inference to produce learning systems. The rule type used to learn other rules were called meta-rules, and meta-rules have produced some valuable behaviors in adaptive systems based on Java components such as DRools.

$\endgroup$
0
$\begingroup$

We have to separate between the Prolog language itself and research projects which are using it. The prolog syntax itself is easy to master. I've found on github a project in which somebody has written in Python a Prolog interpreter and the software needs around 253 lines of code. Is that github project an example for strong AI? Probably not, and i would guess many people can replicate the interpreter project and even try to shrink the needed amount of lines of code.

In the 1980s, many software was written in Prolog and with a bit search it is possible to identify commercial supplier for Prolog books, software and teaching material. If one of these projects has reached Strong AI level is unclear. As far as i know, the NARS software (Wang) is using Prolog. But the prolog part can be replaced with any other programming language, so Prolog is some kind of cargo-cult without real purposes.

$\endgroup$

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.