Posted by David Gledhill
"The old order changeth, giving place to new..."
The last visiting speaker we had was the son of one of our members and showed how very much technology had progressed in the space of one generation and this night we had another, similarly impressive example.  Paul gave a very clear and comprehensible explanation of what AI is, and where it fits in the development of computers.
Paul trained and worked in London before coming to New Zealand an 1998.  Since then he has held senior positions in many places, including Datacom, the Colonial Bank of Fiji, and Kiwibank.  He completed his MBA in 2005 and won the CIO Award in 2019.  His text for the evening was from Arthur C Clarke, the doyen of Science Fiction:
"Any sufficiently advanced technology is indistinguishable from magic."
He started by showing an example of the capability of AI.  From one still life photograph, even if in black and white or a hand sketched drawing with a brief sound bite of speech AI could construct a totally convincing video of that person saying whatever the computer operator wanted.  By example he showed some video clips of Donald Trump saying  totally made-up things.
Paul went on to explain the digital mathematics behind the capabilities of AI.  These are very complex, and new to most of us but he somehow made them understandable.  The basis of most of our classifications is a grid system, from the streets of Adelaide to flowcharts and spreadsheets which follow set rules.  In this instructions have to be clear and logical.  But in real life such is not the case.  Thus in the English language there are rules such as "i before e except after c."  But then there are words like neighbour which make it hard for computers.  As for converting written words to speech the task is even more difficult, as George Bernard Shaw suggested "ghoti" must be pronounced as "fish"! (gh as in enough) etc.  Or the classic English: "The spirit is willing but the flesh is weak" machine translated into Russian then back into English becomes "the vodka is good but the meat is rotten".  The reason is that human brains work on a web system, where points or concepts have many different regular and irregular connections. A classic example was the ancient  Rosetta Stone, a triangular piece of stone discovered by Napoleon's scholars in Egypt with a message carved on it in three languages, one of which, hieroglyphics, had been unintelligible.  By reading the three accounts in total scholars were able to work out the meanings of the hieroglyphics and the Demotic text.  By using the web approach to neural connections computer experts can feed untold numbers of examples of language, or anything, into powerful computers and require them to develop their own rules of comprehension and recognition.  This makes the computer using AI far more potent.  In 1997 it took Big Blue, an IBM supercomputer to beat the current world chess champion because it, and all lesser computers, were using the grid system of yes/no instructions and decision trees.  By 1916 a far less powerful computer, using AI, beat the current champion of GO, a far more complex board game, using rules it had taught itself using AI.
"Old" famous computer firms like INTEL made grid based chips, which are still useful, but the new generation of AI chips are making new firms like Nvidia the way of the future.
The implications of this new technology are impressive or frightening or both, depending on your point of view.  In 1800 90% of jobs in USA were based on agriculture, by 2000 the figure is 2% because machines could do the work of people.  Similarly a very high proportion of jobs in New Zealand are based on the grid system of technology but with the new approach of computers using neural networks most of the existing  technicians, manual workers, lawyers and teachers, even doctors (the list goes on) will be unnecessary.
In summary, by 1945 Alan Turing had developed the concept of AI, by 1964 computers were operating using the grid system, in 1999 machine learning started, in 2011 deep learning appeared and by 2020 generative AI.
Paul did a brilliant job of explaining what AI is, how it came about, and what it can do.  He was thanked by Barrie, his father who gave a them in a passage constructed by chat gpt.