The whole subject of Artificial Intelligence (AI) has been taken up by the media channels following the release of ChatGPT in November 2022 created by the tech start up OpenAI. By January 2023 Microsoft has invested $10 bn into the Company and quickly built GPT functionality into its Bing search engine with a promise of its direct integration into Office. By March Google responded with their launch of their AI-powered chatbot named Bard. My concern is the media has labelled the subject as Artificial Intelligence (AI) when in fact these are the products of Machine Learning (ML).
Artificial Intelligence is a much more complex subject that
was first defined in the 1950’s. It was focussed upon the concept of a computer
acting exactly like a human. Could you be tricked into thinking the computer
you were conversing with was a human. It was viewed as computational psychology
looking for a computer to emulate the richness and subtlety of our mental
powers. Human intelligence is built upon cultural beliefs, individual ideas, interests,
purposes, choice, self-reference and self-knowledge. Various hypotheses exist
about how human mental processes use these building blocks to create what we
define as intelligence. ChatGPT and Bard are only tools built upon the huge
internet knowledge base and in no way reflect the complexity of Artificial
Intelligence. Let us explore the historical background to the evolution of Artificial
Intelligence (AI) and therefore see these new so called AI tools in their
correct Machine Learning (ML) category being just being a minor a branch of AI.
Artificial Intelligence is about making a computer have intelligent
behaviour like creativity, originality, autonomy and consciousness. These
intelligent computers will be able to converse with humans in a natural
language and understand speech and pictures. These will be computers that can
learn, associate, make inferences, make decisions and otherwise behave in ways
we have always considered the exclusive province of human reason. The significance
of artificial intelligence was as a tool it could amplify human thought. Intelligent
behaviour has attributes like deciding to search for a solution to a problem.
Then in so doing applying a guessing approach to cut down the ideas that need
to be searched. Followed by creating a solution then testing if it works. Then
if it does not work trying something else. This is building up specialist
knowledge based upon problem solving.
Artificial Intelligence to achieve its purpose needed to
move forward on many fronts including natural language understanding, robotics,
image and speech understanding, cognitive modelling and theorem proving. The
classic books on Artificial Intelligence cover Three Volumes edited by Avron
Barr and Edward A. Feigenbaun from Stanford University in 1981. These digitised
in 2012 and added to the Internet Archive and they are freely available to read
online or download.
The Handbook of Artificial Intelligence. Volume 1
https://archive.org/details/handbookofartific01barr/
The Handbook of
Artificial Intelligence. Volume 2
https://archive.org/details/handbookofartific02barr/
The Handbook of Artificial
Intelligence. Volume 3
https://archive.org/details/handbookofartific03cohe/
Edward A. Feigenbaum
sometimes referred to as the father of “Expert Systems” is well worth a read
see the Wikipedia entry below.
https://en.wikipedia.org/wiki/Edward_Feigenbaum
Having given you an insight into the true subject of Artificial Intelligence (AI) it should be noted it evolved many specialist branches all of which need to be fully developed before the we can say we have truly created AI. Many of these branches have evolved into sub-systems that stand independently from AI. They offer their own evolutionary pathways many of which will contribute parts to the creation of true AI. Many will go beyond “human capabilities” whereby they offer capabilities beyond that of a human. With the inevitable danger of them being an independent super intelligence that could come to harm us humans. No longer the talk just within the scifi (scientific fiction) community but the potential to become a reality.
In my career the 1980’s
were full of media headlines outlining the so called “Fifth Generation” of
computers on which Japan was become the lead.
First Generation Vacuum Tube Computers
Second Generation Transistors
Third Generation Integrated Circuits
Fourth Generation Large Scale Integrated Computers
Fifth Generation Artificial Intelligence
(Japanese lead)
In reality Japan did
not become the lead but it stirred America into responding to the threat. The
Japanese approach was very Government focussed and centrally structured. The
American response had some Government and Military initiatives, but it was once
again a very commercially driven response dependent on the innovation of individuals
supported by very free flowing venture capital. What the Japanese did do is define
very precisely their strategic objectives in terms of achieving the goal of
Artificial Intelligence. This certainly focussed American minds in particular that
of Edward A. Feigenbaum who wrote a book on the subject, called The Fifth
Generation, which became a justification for America to get its act together.
Not surprisingly the
American approach was more pragmatic being focussed upon Knowledge bases and Expert
systems. Something less theoretical but something offering immediate benefits.
The capturing of “experts” knowledge into databases that could be viewed in
various logical ways to derive “expert answers” became the practical side of
implementations. Originally this was knowledge acquired from human experts. But
the streaming of masses of data from other sources soon identified the power of
this type of approach. Out of these evolved Machine Learning. (ML)
David Bannister (Banno)
Written in 2023
First Published on Blogger 2024
No comments:
Post a Comment