Skip to content

Commit

Permalink
ADDED: New draft chapter on artificial intelligence (AI) with Prolog.
Browse files Browse the repository at this point in the history 
This chapter was suggested by Anton Kochkov (@XVilka) in issue #5.
The Metagol reference was suggested by Stassa Patsantzis (@stassa)
in a HN discussion (https://news.ycombinator.com/item?id=17121028)
of The Power of Prolog. Many thanks!

This chapter will be updated with more use cases as they become
available. I am very interested in the chatbot example, RITS, and in
the formalisation of legislation. Please contact me if you want to
contribute to AI developments with Prolog. See also issue #5. I invite
you to add interesting suggestions and comments to that issue too.
  • Loading branch information
@triska
triska committed Jul 7, 2018
1 parent 845a224 commit 04e5b54
Show file tree
Hide file tree
Showing 2 changed files with 208 additions and 0 deletions.
205 changes: 205 additions & 0 deletions prolog/ai.html
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,205 @@
<!DOCTYPE html>
<html>
<head>
<title>Artificial Intelligence with Prolog</title>
<meta name=viewport content="width=device-width, initial-scale=1">
<meta name="description" content="Memoization">
<meta name="keywords" content="Prolog,AI,Intelligence">
<meta name="author" content="Markus Triska">
<link rel="stylesheet" type="text/css" href="prolog.css">
<link rel="stylesheet" type="text/css" href="toc.css">
</head>
<body style="padding-left: 5%; padding-right: 5%; padding-bottom: 3cm">

<br><br>
<br><br>
<h1>Artificial Intelligence with Prolog</h1>

<h2>Introduction</h2>

What is <i>intelligence</i>? Various explanations and definitions
have been put forward by different researchers.
<i>Intelligence</i> has been defined, among many other proposals,
as the ability to acquire and apply knowledge, and as the capacity
for logical reasoning and self-awareness. It has been related to
understanding, solving, planning, and several other concepts and
tasks. It has been categorized into different <i>kinds</i> of
intelligence, such as <i>emotional</i> intelligence, <i>social</i>
intelligence and so on.


<br><br>

The word stems from the Latin verb <i>intellego</i> "I understand,
realize, perceive, see, notice" (among other meanings),
from <i>lego</i> "I gather, collect, take, read" (among other meanings).

<br><br>

Since no generally accepted definition of <i>intelligence</i> is
available, it is even harder to define what <i>artificial</i>
intelligence is. Broadly speaking, we can distinguish
two different <i>categories</i> of artificial intelligence&nbsp;(AI):

<ul>
<li><b>strong</b> AI: This means a mechanism whose intellectual
ability is indistinguishable from that of a human.
</li>
<li><b>weak</b> AI: This means a mechanism that automates a
subset of human abilities.
</li>
</ul>

So far, we have seen a lot of progress in <i>weak</i>&nbsp;AI. For
example, machines have been built that
play <a href="https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)">chess</a>, <a href="https://en.wikipedia.org/wiki/AlphaGo">Go</a>,
<a href="https://en.wikipedia.org/wiki/Watson_(computer)">Jeopardy</a>
and many other games even better than the most able humans. Many
of us are routinely using partially self-driving trains and
subways, and we currently witness the first attempts at
self-driving&nbsp;cars.

<br><br> In contrast, <i>strong</i> AI has remained out of reach.
In fact, it has
been <a href="https://en.wikipedia.org/wiki/Hubert_Dreyfus%27s_views_on_artificial_intelligence">argued</a>,
and <a href="https://en.wikipedia.org/wiki/Chinese_room">further&nbsp;argued</a>,
that strong&nbsp;AI is <i>inherently</i> impossible to implement.

<h2>Approaches</h2>

Broadly speaking, we distinguish two possible <i>approaches</i>
towards implementing&nbsp;AI:

<ol>
<li><b>symbolic</b> approaches:
<ul>
<li>rule-based <a href="expertsystems">expert&nbsp;systems</a></li>
<li><a href="theoremproving">theorem&nbsp;provers</a></li>
<li>intelligent forms of <a href="sorting#searching">search</a></li>
<li>constraint-based approaches</li>
<li>etc.</li>
</ul>
</li>
<li><b>statistical</b> approaches:

<ul>
<li><i>neural&nbsp;nets</i></li>
<li><i>machine learning</i>&nbsp;(ML)</li>
<li>etc.</li>
</ul>
</li>
</ol>

Both approaches have been widely known among researchers, and have
also been used in practice, for many decades. Prolog is frequently
associated with <i>symbolic</i> approaches. Well-known
applications of this kind
are <a href="/wumpusworld/">Wumpus&nbsp;World</a>, <a href="/zurg/">Escape
from Zurg</a> and <a href="/various/conn4/">Connect&nbsp;4</a>.
Prolog can of course also implement statistical approaches. For
example, check
out <a href="https://dtai.cs.kuleuven.be/problog/index.html">ProbLog</a>,
<a href="https://github.com/friguzzi/cplint">cplint</a>
and <a href="https://www.cs.ubc.ca/~poole/aibook/code/ailog/ailog2.html">AILog&nbsp;2</a>
for <i>probabilistic</i> logic programming and reasoning with
uncertainties.

<br><br>

The approaches have different advantages and drawbacks: A symbolic
approach is typically amenable to formal <i>verification</i>. The
rules are explicitly available, and you can check whether they are
complete, and whether they express what you intend.
Using <i>inductive</i> logic programming (ILP), you can even let a
program <i>learn</i> rules based on positive and negative
examples. See for example
the <a href="https://github.com/metagol/metagol">Metagol</a>
system. However, not all tasks are amenable to symbolic
approaches, because the rules may be <i>too hard to express</i>.
For example, it is hard to express rules that determine whether a
musical&nbsp;piece is borderline atonal, or whether a photo
contains a painting of a flower.

<br><br>

On the other hand, statistical approaches have the ability to
acquire rules <i>implicitly</i>, based on
so-called <i>training-</i>, <i>test-</i>
and <i>validation&nbsp;sets</i>. This is convenient, because you
do not have to think about the rules yourself, and the machine may
derive rules that are too hard for you to encode. Unfortunately,
this also comes with significant drawbacks: You can no longer
reason about the rules explicitly, and the obtained mechanism may
produce colossal outliers that are hard to remedy. For example, it
may misclassify the picture of a penguin as a car if you change a
single&nbsp;pixel.

<h2>Prolog and AI</h2>

Prolog has very strong historic ties with AI. In 1982, Japan
started a very ambitious government project called the <b>Fifth
Generation Computer System</b>&nbsp;(FGCS) with the goal to create
a massively&nbsp;parallel computer, using <i>concurrent
logic&nbsp;programming</i> as the software foundation of the
project.

<br><br>

When I visited Japan, I asked a retired Japanese scientist who had
worked on the project why the project had failed. He said: "What
do you mean, '<i>failed</i>'? All my colleagues who have worked on
the project went on to become <i>professors</i>!" Another
researcher told me that the project achieved what was actually
intended: a way to input Japanese characters into computers.

<br><br>

In my view, a major achievement of the scientists who had worked
on this project was that they managed to convince the government
to fund an interesting and very&nbsp;ambitious project with
large&nbsp;sums.

<br><br>

By the end of the FGCS project, commodity hardware had become so
powerful that custom-built machines as those constructed by the
project were hardly needed anymore.

<br><br>

Now, several decades later, is it not time
to <i>try&nbsp;again</i>? Is it not time to convince <i>our</i>
governments and funding agencies that we want to benefit from the
power of Prolog to build applications, and do interesting
research?

<br><br>

Here are a few ideas:

<ul>
<li>Write a <i>chatbot</i> that answers questions to the administration.</li>
<li>Write an application that guides students through solving
tasks, so that it can be used in computer-aided education. See <a href="/rits/">RITS</a> for an example.</li>
<li>Find a way to <i>formalize legislation</i>, so that
questions about laws can be automatically answered. For an
example, check out a Prolog formulation of
the <a href="https://github.com/bitlaw-jp/the-constitution-of-japan">constitution
of Japan</a>,
and <a href="https://www.youtube.com/watch?v=YGG3e_1JX7Q">Reasoning
with Regulations</a> as instances
of <b>smart&nbsp;contracts</b>.
</li>
</ul>


<br><br><br>
<b><a href="/prolog">More about Prolog</a></b>

<br><br><br>

<b><a href="/">Main page</a></b>
<script src="jquery.js"></script>
<script src="toc.js"></script>
</body>
</html>
3 changes: 3 additions & 0 deletions prolog/toc.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -86,6 +86,9 @@ var toc = new Array(
{ link: "/prolog/memoization",
title: "Memoization"
},
{ link: "/prolog/ai",
title: "Artificial Intelligence"
},
{ link: "/prolog/horror",
title: "Horror Stories"
},
Expand Down

0 comments on commit 04e5b54

Please sign in to comment.