Not surprisingly

given the appalling lateness of the July-August column, this catch-up column also covers two months, though it might have been more accurate to put a three months' tag on the previous one. Alternately, perhaps the Spy himself should be deemed artificially intelligent, given his unseemly forgetfulness. What did you say my name was, again?

The "promised land"

forecasts for AI have been around for decades, and its optimistic prognosticators are now having an unqualified jubilant field day touting the success of bots such as ChatGPT. It seems the vigorous debate over its negative and even frightening possibilities and potential flaws have all but been forgotten in the new rush to hail and enthrone the latest media super celebrity. Well, there is noting new in that--or in the fleetingness of popularity. Ask any elected official or once media darling whose star has faded.

Intrigued, and for his own bemusement, the Spy tried it out, asking it who he himself was, supplying only his name. It came back with his pitching statics from the Cubs and Padres. Hmmmm. Missed that game. So, he added "the computer scientist" and it did much better, returning the bio of the man with the same name at an Australian university. Adding in "at Trinity Western University" finally elicited the right answer, along with much additional information, including that he had written science fiction novels. Great stuff! It even provided the the titles of his books. Unfortunately, they were all written by someone else with completely different both given and surnames. It was nice to get the information, for he had never before heard of those particular authors and it was interesting to have discovered them in his own AI-generated biography. Given that he has been hanging out electronically, so to type, since before there was an Internet, this seemed….inadequate.

The incident reminded him of a faculty discussion in which he had participated concerning the marking if essays in an interdisciplinary studies capstone course TWU once offered. These were all scored by at least two professors, and if their judgements differed significantly, a third would be engaged. One co-marker was never more than 3% off the Spy's mark, but another often differed by as much as 40%. The latter's main criterion was correctly presented English, whereas the other agreed with the Spy that some papers had much correct form, were even eloquent, but sadly, had almost no substance. In the subsequent debate at a general faculty meeting, and after several comments concerning correct grammar, punctuation, sentence and paragraph structure, and general style, the Spy asked "What about substance?", and the chair of the English department hollered out "There's gotta be some," eliciting modest applause.

There is no doubt of the success of ChatGPT when it comes to correctly slinging words. It stands out so obviously in purported student essays that it's almost painful to behold. Who actually talks or writes like that outside the highest levels of arcane academia? Stirring words together into a tasty stew of eloquence in a style that warms the heart of any professional grammarian, language professor, lawyer, or political speechwriter is certainly its forte. There are plenty of tutorials on proper English and a plethora of good examples for patterns it can be algorithmically programmed to mimic.

Even when it comes to looking up and presenting technical content, such machines can do reasonably well, at getting it approximately right most of the time. After all, most habitual liars and people with political, religious, or social axes to grind do not know enough about technical matters to post their patented nonsense on the net for the bots to find and use. So, if one wants a passable little computer program in some obscure language (programming notation) to work out the mean, standard deviation, and confidence level of a data collection, tell you the molecular weight of a complex hydrocarbon, explain Newton's laws, relatively or the undecidability of the halting problem, one is reasonably likely to get back usable results, though even then, they should be checked with a separate verification search. But, as the Spy's own little experiment illustrates, when it comes to personal, social, political, religious, or other debated or nuanced matters, it fails dismally, and the essays it writes are usually eloquently empty of content and fail to exhibit thoughtful critique or creativity.

In the classes on mathematics and computing science, the factual information is fairly straightforward and the results of not following proper planning and software engineering regimens, or using the correct grammar in the programming notation, are depressingly predictable. The proof, the program, the electronic circuit, and the letter grade are all dismal.

These new bots do constitute sophisticated advances on the expert systems we have had for some time. Given a boatload of curated factual information and some heuristic rules for patterns of human thinking on the matter at hand, expert systems can be quite useful at handling some kinds of medical diagnoses, legal matters, or playing chess, for instance.

But even there, the operative word is "curated", and it applies to both the data and the heuristics provided. We all know about GIGO (garbage in, garbage out) do we not? The main problem with the current Internet, and the reason why it is not the prototype for the all-encompassing Metalibrary of knowledge and information that the Spy was forecasting as far back as the 1980s (and still is) boils down to "who curates the data and algorithms, why, and how?" Bluntly put, far too much of what gets thrown up (metaphor intended) on the Internet is nonsense, and no one really cares.

FWIW, the Spy's solution has always been that there has to be a peer review system, with a network of two-way citation links, peer reviews, and ratings, along with an accounting system of micro debits and credits for content creators to be reimbursed. Moreover, there may have to be separate networks for scholarly publications, news organizations, political discourse, and personal postings, as the current cross postings of all these pollutes any possibility for data credibility in many fields.

Even in his fiction, the Metalibrary is not wide-openly available to the AI character, which has access only to curated data. After all, if purportedly intelligent human beings can fail to discern truth from lies (witness our political past and present), how could this be capably programmed into an artifcially intelligent machine?

After all, the Internet even hosts modern (and apparently serious) flat earthers who must also believe the moon voyages were hoaxes. Who'd have imagined that ridiculous calumny against Christian beliefs invented merely a few centuries back would actually ever turn out to be believed by anyone? Then there are the endless conspiracy theory true believers, election deniers, slanderers, worshippers of false prophets, and fraudsters, plus the political, religious, media, and social celebrities who can purportedly do no wrong, say no lies, and be excused for any crime provided they back ones pet cause(s). If humans are often incapable of sifting truth from lies, fact from rumour, fantasy from reality, how could a machine programmed and fed its data by people do so?

There are also problems,

as the Spy has for decades pointed out, with the purported, and oft science-fictionalized goals of harder AI-namely the notion that we may someday build machines that can think equivalently to a human being. What, after all, is the human mind? Putative prophet Marvin Minsky famously postulated "The mind is a machine made of meat." His point was that said machine could therefore obviously be reverse-engineered and its functionality duplicated. For some enthusiasts, that dream parallels the idea of somehow scanning the human brain and downloading its contents and operations into a backupable artificial construct, thereby achieving transcendent immortality as a cyborg--a lofty goal indeed.

On the other hand, consider the counter of Mathematician-Philosopher-Physicist-Astrophysicist Roger Penrose, whose provocatively titled AI book "The Emperor's New Mind" puts the case that Godel's theorem (it is impossible to build a finite logical system that is both complete and consistent) which is equivalent to the undecidability of the halting problem for finite state machines--and that is what computational devices are--shows that hard AI is not in fact possible. He argues that the known laws of physics are inadequate to explain the phenomenon of consciousness and that thought cannot be simulated algorithmically because it isn't algorithmic. The Spy is inclined to side with Penrose on this. The latter's well-earned 2020 Nobel Prize for his long and celebrated polymathic career perhaps lends a certain additional cachet to his arguments.

And there are other issues

that the most optimistic SF authors (sometimes writing in the newspaper or in online fantasies such as those like this column might want to consider. Here, for instance, are a few questions from the exercises in his textbook on technology, ethics, and society that he asked his students to tackle as far back as the 1980s.

If a machine passes Turing's test, claiming for itself self-awareness, does that mean that it is in fact self-aware? How can one tell?

If the Metalibrary or any AI device behaved in every other way as though it were intelligent, how could one determine if it were independently self-aware?

What is intentionality? Could a machine have it? Why or why not? (The same question could be asked of love, intuition, good judgement, common sense, etc.)

Should an AI device that is capable of duplicating the results of human thinking be regarded as equal in status to human beings?

What legal and/or social rights should AI devices have? Does your answer depend on their appearance or only on their "thinking" capabilities?

Should an employer give a job to a human applicant in preference to an AI device that can do it better for the same salary? Why or why not?

You are the pastor of a small church and are visited one night in 2065 by a mobile AI device that tells you: "I've been listening to your sermons and I want to become a believer and a member of your church." What do you do and why?

Finally, who curates the data to which a bot has access for answering factual questions and rendering diagnoses, opinions, and solutions to problems? Oh, and who chooses the curators--politicians, electors, religious leaders, lawyers, social influencers, celebrities, media professionals, the tech industry, military leaders, scholars, self-styled experts?

Oh, and by the way, he tells his students in the ethics and technology course that unlike in mathematics or Physics, which are not for the most part about opinions, if all his students do is parrot back the lectures and text (which he wrote) they would be doing well indeed to earn a C+. For an A they have to credibly argue against his views and ideas, even if they actually agree with him. Isn't that what a university is all about--searchingly scrutinizing ideas? True, there are factual things that are indisputable (at least unless and until someone performs a repeatable experiment refuting their universality or at least points out conflicting evidence), but the knowledge and even the epistemology of a given discipline do undergo change, (hopefully maturing) with time, and almost everything needs to be re-examined for credibility by every generation. Yes, many truths are indeed universal and timeless, but other "truths" are only fashionable and transient, and may even be delusional.

THIMK

Your computer has no intelligence, artificial or otherwise. The real thing resides in the programmer. Or, have you hugged your computer yet today?

--The Northern Spy

Opinions expressed here are entirely the author's own, and no endorsement is implied by any community or organization to which he may be attached. Rick Sutcliffe, (a.k.a. The Northern Spy) is professor of Computing Science and Mathematics and Assistant Dean of Science at Canada's Trinity Western University. He completed his fifty-third year as a high school and university teacher in 2023. He has been involved as a member of or consultant with the boards of several organizations, and participated in developing industry standards at the national and international level. He was co-author of the Modula-2 programming language R10 dialect project. He is a long time technology author and has written two textbooks and ten alternate history SF novels, one named best ePublished SF novel for 2003. His various columns have appeared in numerous magazines and newspapers (dead tree and online formats) since the early 1980s, and he's been a regular speaker at churches, schools, academic meetings, and other conferences. He and his wife Joyce celebrated their fiftieth anniversary in 2019 and lived in the Langley/Aldergrove/Bradner area of B.C. from 1969 to 2021, where he latterly continues alone, depending heavily on family and friends to manage.

URL s for Rick Sutcliffe's Arjay Enterprises:

The Northern Spy Home Page: https://www.TheNorthernSpy.com

opundo : https://opundo.com

Sheaves Christian Resources : https://sheaves.org

WebNameHost : https://www.WebNameHost.net

WebNameSource : https://www.WebNameSource.net

nameman : https://nameman.net

General URLs for Rick Sutcliffe's Books:

Author Site: https://www.arjay.ca

TechEthics Site (Fourth edition of text; the fifth is in preparation) : httpss://www.arjaybooks.com/EthTech/index.htm

Publisher's Site: https://www.writers-exchange.com/Richard-Sutcliffe.html