Some brilliant people have invested entire careers in the investigation of whether true artificial intelligence (AI) is possible. The claims have been many…as have the disappointments. Today we see programs such as ChatGPT proposed for the role. Some of them have impressed legions of non-technologists. We the Geeks and Nerds can peek behind the curtain…and what we see there is still dubious.
One of the problems in this space is settling on a consensus answer to the question “What is AI?” Just now there isn’t one. Various “expert systems” optimized for the solution of particular classes of problems have tantalized us with the possibilities. But they’re generally agreed not to be AI, owing to their designed-in limitations.
Among the things that elude most people about questions such as this one is that it depends upon what we agree to use as an acceptance test. In his book on systems engineering and specification, the brilliant Tom McCabe put the matter plainly:
Some of the researchers in AI have proposed tests they favor, but so far none of them have commanded a consensus. Alan Turing’s proposed test – that the candidate AI must be able to fool a human being into believing that he’s conversing with another human being – has generated far more dissent than consent.
Today in the New York Post, NewsCorp CEO Robert Thomson issues some criticisms of contemporary thrusts at AI:
News Corp CEO Robert Thomson blasted the left-wing bias and inaccuracies spewed out by AI generated content — calling it “rubbish in, rubbish out” — even as he warned the technology threatens to kill thousands more jobs across the news industry.
Left-leaning media giants that dominate the news business have churned out stories for years that are not only riddled with errors, but also written with a left-wing slant.
Yet bots like the popular ChatGPT search engine will regurgitate the claptrap as fact, according to Thomson.
“People have to understand that AI is essentially retrospective,” the media executive said during an appearance at the Goldman Sachs Communacopia and Technology Conference in San Francisco on Thursday.
“It’s about permutations of pre-existing content.”
There’s a lot of truth to that. It also raises a key question: Can a program learn to seek input beyond the material it’s deliberately fed by its trainers? That’s a characteristic of human intelligence…sadly, more neglected than exercised. But once again, consensus is elusive, for it has been argued that human intelligence ought not to be the standard against which candidate AIs are measured. But what, then, is the standard we should use?
The question appears to loop back on itself endlessly.
The possibility of AI is of interest mainly because of its applications. Many such applications are “labor savers,” in that tasks currently undertaken by men could be relegated to AI systems equipped with the necessary devices for interfacing with and manipulating material objects. Indeed, expert systems of that sort are already in use; I’ve worked on a couple of them. Once again, the limited nature of the various tasks suggests that what we’ve achieved is not true AI…but there’s that pesky acceptance-test problem again.
Fiction tends to influence our conceptions of such things. A few examples:
- E. M. Forster’s novella “The Machine Stops;”
- Philip K. Dick’s novel Do Androids Dream Of Electric Sheep? which gave birth to the Blade Runner movies;
- Harlan Ellison’s award-winning short story “I Have No Mouth And I Must Scream,” which gave birth to the Terminator movie franchise;
- Julie Christie’s star vehicle Demon Seed;
- Alicia Vikander’s stunning performance in Ex Machina;
- Colossus: The Forbin Project;
…and many others. Note that the influence of the AIs in those fictions is depicted as generally negative. They murder, wage wars, overthrow governments, practice genocide, and kidnap and impregnate human women. As we already have an oversupply of demonstrated experts ready, willing, and able to do those things, AIs are seldom considered for those applications today.
I’m not about to propose an acceptance test of my own; I have too much to do and too many people to piss off already. But I’ll suggest that the “AI problem” is less of one than many think. If it’s potential applications that concern us above all else, then our focus properly belongs on applications. That, of course, reorients us toward expert systems and how sophisticated they can be made.
The Japanese have promoted the idea that AI-equipped androids could fill human roles that are currently unfilled and in many cases unfillable. We’ve all heard about the sexbots, the market for which appears to be healthy. Japanese developers are marketing companion androids, for those who lack sufficient human connection, and android nannies for working mothers. I have no information about the limits of such androids. But once again, it’s the application that matters…and the acceptance test is the specification.
My preferred application is yard work. The day some entrepreneur produces an AI-equipped android that can and will:
- Mow the lawn;
- Trim the shrubs;
- Wash the house and deck;
- Scoop up and bag the dog poop;
- And yell at the neighbor kids not to throw empty cans and bottles into my yard;
…with minimum required maintenance and guaranteed limited downtime, I shan’t ask “Is this true AI?” I’ll have my checkbook to hand. Hey, wouldn’t you?
2 comments
Pohl’s Gateway series also dealt with a god-like AI system that would allow people to upload themselves into the machine. It has been a long time since I read the books. The question that raises, to me is this- Say, one of the systems already in use decided to mimic you, or me, or anyone who spends considerable time on line. Over the years, now we’ve posted millions of words, lots and lots of images, and let our opinions be known on many different topics. Could the machine pull it off? Could a bot mimic any of us well enough to continue posting in our names after we were gone- either dead, or merely disconnected permanently from the web? Would it fool anyone? I’d say, that on a forum like twitter, it probably could. Long form might be a different matter, but long-form internet has already become a refuge for us old fogies. I know a lot of people who don’t have computers at all. They rely completely on the phone. That trend will only grow. I guess that comes back to the Turing test. In a constricted format the machine could probably pass. Soon all the formats will be so constricted.
JWM
Care for the elderly is a relatively benign use of AI – certainly, the AI bot would have infinitely more patience for repetitive questions or comments, be able to provide a presence that both monitors vital signs, AND is available for comfort-related interactions, including listening to attempts to communicate.
I’m less excited – a LOT less excited – at the idea of a Robby the Robot, who is intended to provide a large portion of child care.
Kids KNOW the difference between engaged interactions with emotional content, and that provided by unemotional caretakers. I would expect the first kids to be brought up that way to be unsociable adults incapable of functioning in a world of humans. That would quickly bring an end to the species.
The more lifelike the AI, the less the child would want to lose that connection. Unlike the story arc of Toy Story, the kid would not put that AI aside. In that case, their ability to connect with imperfect and flawed friends would be impaired
And, for parents, it would reduce the emotional connections with their child/children. It is partly the time we spend on our activities that gives them value. The AI-raised child would be an inscrutable and unloved toy for the adults – picked up when convenient, dropped when ‘ucky’ or something more exciting comes along.
TPTB have done their level best to interrupt the normal responses of parents. From the women who are disconnected from the birth pangs by spinal meds (or, completely disconnected from the process via surrogacy), to those women who don’t even consider breastfeeding (a tremendous bonding experience for both participants), to a push to turn over day-to-day operations to paid caretakers, all is designed to break the connection between mother and child.
And, the push of fathers out of the house is accelerating. Women are enticed with visions of attractive males just eager to provide their own personal love connection, which blinds them to the reality of life. With so little time with their own children, men detach, leaving them with no understanding of how men think or function.