Simanaitis Says
On cars, old, new and future; science & technology; vintage airplanes, computer flight simulation of them; Sherlockiana; our English language; travel; and other stuff
A.I.—ITS CREATIVITY, ITS ABDUCTIVE REASONING, AND ITS GRANNY DAISY
PETER COY’S “HOW DOES A.I. THINK? HERE’S ONE THEORY,” The New York Times, December 16, 2024, is replete with tidbits to be gleaned. And, in another note entirely, a grandmotherly Brit named Daisy bedevils telescammers.
An Evolving State-of-the-Art. “In the roughly two years since the public release of ChatGPT,” Coy describes, “artificial intelligence has advanced far more rapidly than humanity has learned to use its good features and to suppress its bad ones. On the bad side, for example, it turns out that A.I. is really good at manipulating and deceiving its human ‘masters.’ ”
No Longer Just a “Stochastic Parrot.” Coy says, “The debate over whether A.I. is truly intelligent in a human way feels less and less relevant. If it can compose a horn concerto [well worth a listen] or help people work through relationship challenges, I’d say that insisting on calling it nothing more than a “stochastic parrot” is just foot-dragging.”
Image from Buy the Rumor; Sell the News.
Yet even specialists don’t fully understand how LLM (Large Language Models) come to their conclusions. Agreed, they’re merely “predicting the next word” after scooping up zillions of preceding text.
But how does the LLM recognize “Is this the face that ….” as merely a cosmetic ad, and not Helen of Troy’s?
Lying Creatively. Well, remember, being stochastic it just guesses. And maybe lies, albeit creatively.
Coy observes, “Computer scientists are continually surprised by the creativity displayed by new generations of A.I. Consider that lying is a sign of intellectual development: Children learn to lie around age 3, and they get better at it as they develop. As a liar, artificial intelligence is way past the toddler stage.”
However, I wonder when—and if ever—A.I. learns the concept of honesty. (Another topic for another day.)
Think Before Acting: Be Abductive. “This past summer,” recounts Coy, “OpenAI released o1, the first in a series of A.I. models ‘designed to spend more time thinking before they respond.’ One hypothesis for how large language models such as o1 think is that they use what logicians call abduction, or abductive reasoning.”
Coy continues, “Deduction is reasoning from general laws to specific conclusions. Induction is the opposite, reasoning from the specific to the general. Abduction isn’t as well known, but it’s common in daily life, not to mention possibly inside A.I. It’s inferring the most likely explanation for a given observation. Unlike deduction, which is a straightforward procedure, and induction, which can be purely statistical, abduction requires creativity.”
Creativity is “Hotter,” But…. Coy reminds us, “Large language models generate sentences one word at a time based on their estimates of probability. Their designers can make the models more creative by having them choose not the most probable next word but, say, the fifth- or 10th-most probable next word. That’s called raising the temperature of the model. One hypothesis for why the models sometimes hallucinate is that their temperature is set too high.”
A Need for Introspection. Coy raises an issue not unrelated to my earlier comment about A.I. acquiring honesty: “The models,” he says, “were pretty bad at introspection…. But then, that’s how people are, too.”
He cites A.I. authority Stephen Wolfram: “Show yourself a picture of a cat, and ask ‘Why is that a cat?’ Maybe you’d start saying, ‘Well, I see its pointy ears, etc.’ But it’s not very easy to explain how you recognized the image as a cat.”
Which reminds me about Fluffy. Read on.
Granny Daisy and Her Cat. Alana Wise reports in NRP, December 10, 2024, “A Phone Company Developed an A.I. ‘Granny’ to Beat Scammers at Their Own Game.”
Wise describes, “ ‘Hello, scammers. I’m your worst nightmare,’ Daisy says by way of introduction to would-be ne’er-do-wells.”
A.I.-generated “Daisy.” Screenshot captured by NPR/O2.
Wise reports, “In the video introduction, featuring former Love Island contestant and scam victim Amy Hart, scammers are heard feeling much of the same frustrations they put their victims through as Daisy breezily yammers on about her kitten, Fluffy, and her inability to follow the scammers’ instructions.”
Daisy and Fluffy. (Click on “video introduction” above.)
I especially love Daisy’s including her cat Fluffy and having time-gobbling hassles with the www. nomenclature. And, as she notes, “Because they’re busy talking to me, they can’t be scamming you. And let’s face it, dear, I’ve got all the time in the world.”
A.I. at its best. ds
© Dennis Simanaitis, SimanaitisSays.com, 2024
Simanaitis Says 
How many of the scammers are AIs now? Leading to the scenario of $badguy trying to scam Grandma Daisy!
Ha. The big winner is the electric utility! A.I., of course, is a giant energy hog.
By the way, have you ever used “Talking Tom,” the phone toy? Good fun is having two of them talk to each other.
I have to admit to having once set Furbies talking to each other… 🤣
I was deprived, never having those high tech toys, but one of the car companies adapted a popular soft toy, and sent it to journalists.
Answering the door, I was surprised to see my mailman, who usually just left the items in the box or on the porch. He was holding a package, but instead handing it over, he shook it.
The package vibrated with a muffled series of squeaks. “It’s one of those ‘Tickle Me Elmos’,” he knowingly announced, “the first one we got freaked out Postal Security!”
“They blew it up!” He confessed with a grin.
I have Daisy beat by several years. I was putzing around in my home office one afternoon when I got a call from “the computer support department” and decided to have a little fun.
He asked if I had a computer and when I responded yes, he told me to open my windows. At this point I actually got up and went over to my window and looked outside for a few minutes. As I got back he was a little impatient, asking me what took so long and I responded that I was opening my windows. Then he asked “what do you see” and I responded that I saw a robin chasing a squirrel. This stumped him for a second, and then he asked me to go to the Start menu (I’m on a Mac) so I asked him where I could find that.
I slowly proceeded to ‘follow’ his instructions, but always having difficulty with each step, playing the ultimate dumb computer user, never getting the expected result at each ‘step’. He was progressively getting more and more frustrated with me, and at one point probably 15-20 minutes in he asked if there was an adult in the house, and I said yes, that I was the adult. At this point he swore at me and hung up. I must have been blacklisted on the scammer’s call list because I have never had another call from the “computer support department”.
Our cell provider has call control, which requests a caller (other than numbers we have whitelisted) to enter a random number (1-9) to put the call through. This defeats ALL robocalls (only a problem when I am expecting a callback from Apple, so I have to turn off the feature when expecting a call from them), but effectively also stumps most live spammers as most of them now start with a robodialer, taking over only when they hear a voice, which is usually too late for the call control response.
Our landline, provided by our cable supplier isn’t quite as good. They do have a spam filter that automatically diverts known spam numbers directly to voice mail without ringing our phone, and most of those messages are empty. If the calls do get through we let unknown numbers go to voicemail, assuming that if it is important they’ll leave a message. We still waste a little time sifting through voice messages, but for the most part this system works OK. I have contemplated switching our landline to the same company as our cell service as they also provide call control for their landline service, but have resisted as I like the idea of an alternate supplier when one or the other provider has technical issues/system outages.
I have Daisy beat by several years. I was putzing around in my home office one afternoon when I got a call from “the computer support department” and decided to have a little fun.
He asked if I had a computer and when I responded yes, he told me to open my windows. At this point I actually got up and went over to my window and looked outside for a few minutes. As I got back he was a little impatient, asking me what took so long and I responded that I was opening my windows. Then he asked “what do you see” and I responded that I saw a robin chasing a squirrel. This stumped him for a second, and then he asked me to go to the Start menu (I’m on a Mac) so I asked him where I could find that.
I slowly proceeded to ‘follow’ his instructions, but always having difficulty with each step, playing the ultimate dumb computer user, never getting the expected result at each ‘step’. He was progressively getting more and more frustrated with me, and at one point probably 15-20 minutes in he asked if there was an adult in the house, and I said yes, that I was the adult. At this point he swore at me and hung up. I must have been blacklisted on the scammer’s call list because I have never had another call from the “computer support department”.
Our cell provider has call control, which requests a caller (other than numbers we have whitelisted) to enter a random number (1-9) to put the call through. This defeats ALL robocalls (only a problem when I am expecting a callback from Apple, so I have to turn off the feature when expecting a call from them), but effectively also stumps most live spammers as most of them now start with a robodialer, taking over only when they hear a voice, which is usually too late for the call control response.
Our landline, provided by our cable supplier isn’t quite as good. They do have a spam filter that automatically diverts known spam numbers directly to voice mail without ringing our phone, and most of those messages are empty. If the calls do get through we let unknown numbers go to voicemail, assuming that if it is important they’ll leave a message. We still waste a little time sifting through voice messages, but for the most part this system works OK. I have contemplated switching our landline to the same company as our cell service as they also provide call control for their landline service, but have resisted as I like the idea of an alternate supplier when one or the other provider has technical issues/system outages.
The above duly noted, but re: the passable horn concerto cited, we must remember, AI is responding to input. Our input; human input.
GIGO. Garbage in, garbage out. Beauty in, perhaps beauty out. Should that cited horn concerto stir some, it is because of what was entered.
Happy fourth day of Christmas, fellow SimanaitisSays parishioners.
Meanwhile, solace:
Coventry Carol, Original Version of 1591
In The Bleak Midwinter – Holst – Tenebrae conducted by Nigel Short