There have been a couple of whizbang developments in the world of computers and technology the past two weeks that haven’t been widely reported. Such reporting that exists has been breathless and perhaps a tad extravagant. This week it was the haiku producing artificial intelligence bot called ChatGPT.
“Ask it anything and it will give you an answer?! It is so cool!” seems to the most common response of those who are playing with it. I don’t want to throw cold water but… ask me anything and I’ll give you an answer as well. And, according to the developers, the probability of me being right is about the same as the chatbot.
I will concede that the software writes better haikus than me but that is because I don’t know what a haiku is. Perhaps if I did, I could give the bot a run for its money. As impressive as its haiku writing ability is, Google tells me that “haiku producing software” has been around for at least four years so even in that realm it is not very unique.
Having watched videos of people putting the software through its paces, I remain mostly unimpressed. When I was younger, the whizbang development of the day was Yuri Geller scrunching up his face and bending forks without touching them. It was amazing to see. Unfortunately, that is all Mr. Geller was able to bend and, after the excitement of watching a fork bend, the inevitable thought was, “Why bend the cutlery?”
So why is ChatGPT suddenly such a big deal? I think perhaps because it has been asserted that this software will replace Google and destroy the current internet search engine business. Following this promising line of investigation was ultimately disappointing for me, however. It turns out that the program is not so much a threat to Google as Google is to itself. People are tired of wading through Google ads and so now it is cool to hate Google. I have hated Google ever since they started manipulating the “Google ads” algorithm preventing me from monetizing the internet. ChatGPT is not a Google-slayer after all. Digging deeper into the virtues of ChatGPT reveals that it is especially good at answering code writing questions.
“ChatGPT is so impressive. Surprisingly it’s also super useful for coding-related questions. Would prefer an answer like this over going through multiple six years old Stackoverflow threads,” said Felix Krause.
I feel your pain, Felix. There is nothing worse than going through Stackoverflow threads.
Are you getting the sense that chatbots are kind of made-for-purpose? That shouldn’t be too surprising because it takes an enormous amount of input data to train a computer program to sound “innovative” in its responses to general questions. Therefore, to deepen their knowledge, the breadth of knowledge is constrained, and the software is written to purpose. You may have discovered this from reading the comments section of any popular webzine. The bot entries are easy to spot as they don’t have spelling mistakes. Unless, that is, they are programmed to mimic the other idiots offering “higher level” commentary.
The fundamental issue is whether those responses are truly innovative or simply contrived to sound innovative. Are we dealing with a clever Ouija board or the Delphic oracle? Is this software a step change in general artificial intelligence? Is it garbage in = garbage out or has the paradigm shifted to garbage in = pearls of wisdom out? The way I answer this question is to ask, “How much money am I willing to bet on the numbers ChatGPT gives me for tomorrow’s Powerball draw?’” If my answer is less than the price of the lottery ticket, then I guess I don’t believe it is a paradigm shift.
Please don’t get me wrong. I am impressed by this kind of software. The billions being invested in improved general artificial intelligence may one day pay off but until then I think we mostly have good tools for cheating at Trivial Pursuit. Emphasis on “trivial”. Felix will disagree with me.
The other whizbang technology announcement was the demonstration of Neuralink by the ubiquitous Elon Musk. Let me say that I find the energy and courage of Mr. Musk far more interesting than the idea of planting a signal processor in the brain of a macaque monkey so that it can play flawless games of pong using only its brain to move the computer paddles. From buying, stripping, and rebuilding Twitter to demonstrating remarkable man-machine interfaces, Mr. Musk is one busy dude.
The question centers on whether his monkey is bending forks or is this a preview of something more substantial? Much of the reporting on the Neuralink demonstration has been as breathless as that on ChatGPT but it has come from two angles. To some people, Mr. Musk is stepping into the cyborg world with nary a word of discussion whether, as a society, we want those doors opened. To other people it is gee whiz technology that will unseat Google… or something. The sticking point is the nature of the man-machine interface. A very good treatment of this issue is offered by Joe Walsh at:
I tend to be a reactionary conservative when it comes to messing with the meaning of “human” but what about the imbedded batteries and electrodes that keep many hearts ticking? Do pacemakers reduce our humanity? I don’t remember being consulted on that technology but nor am I disturbed by it. Perhaps I should be. Or is there something unique about putting wires in the brain that causes my discomfort? If Mr. Musk can use his imbedded processor to allow a paraplegic to walk again, will I squawk about the dangers of the man-machine meld? What if his processor will allow a human brain to be “borrowed” by the computing cloud while its carapace is sleeping? What if super soldiers, beyond wearing exoskeletons, can be created by merging machines and humans? Is there a line to be drawn?
Perhaps the difference between brain implants and spark generators for the heart centers on the desire to move beyond simply overcoming physiological deficiencies to creating new binary beings that are neither human nor machine. Mr. Kurzweil wants his machines to become conscious at the singularity – when machines can process information at the volume and speed of the human mind. Mr. Musk wants man and machine to lean on one another to achieve that singularity sooner. In the mind of Mr. Musk, this merged being will somehow protect the integrity of humankind.
Perhaps they both need to reread the Island of Dr. Moreau and Frankenstein or, the Modern Prometheus and then pass a test on both. When one’s aim is to create a new being in one’s own image, one is stepping into deep spiritual waters with significant moral implications. Shouldn’t those of us who will have to live on this strange Moreau-like island have the option of voting off those who won’t bother to ask us what we think of their new world before they build it?
At the very least it is a public interest issue. If I can’t dig on your property without asking your permission, shouldn’t I ask your permission before creating potential monsters who have been trained to climb over the fence into your yard? Must we wait until the fence has been climbed to have that conversation?
My reaction to the rapidly changing context of the world we live in is, “It is just one darn thing after another” and I am frankly getting tired of it. Sometimes it reminds me of, as a child, sitting in the back seat of the family car on a long vacation trip. Inevitably one of us would start to poke another causing all kinds of caterwauling, a swing and a miss from my father and an injunction to shut up and behave or the car would be stopped. Why did we poke at each other? We knew the inevitable consequences. Because we could.
I think as adults we must learn to handle our spiritual boredom with a bit more nuance. Are you listening Mr. Musk? Let’s talk.
Thanks Keith. I will check it out. I often wonder if the bots work well when playing people like me that are lousy at poker. Maybe they rely on an assumption that the opponent will act rationally.
Good article Doc. You should check out the lex Friedman podcast (a recent one) talking to the AI developer about their system that can beat poker players in poker, which is a different problem than playing chess or GO.. with poker you have bluffing which is essentially dishonesty. It’s super interesting and will provide a lot of insight into where this stuff is all going. Also the AIs are now able to beat all top notch poker players, being now able to bluff better than humans.