I am shocked at how quickly companies got bamboozled into using AI for customer service like this, or, if they were conscious of it, how uncaring they appear to be about giving off a bad impression to their customers by outright lying.
They must really believe people will not enforce their boundaries by going somewhere else for service. It's like they think people are devoted to them regardless of how they're treated. I suppose that's because so many people have stopped expecting normal levels of customer service. They get spoiled by people like that.
I lived in Japan for a long time. The customer service there is amazing, and a point of pride for a lot of people working there. Over the course of the nearly 10 years I spent there I had one negative experience, whereas back here in America I meet with regular disappointment. Routing regular work through LLMs removes a lot of peaceful interactions between customers and workers, and leaves only the difficult issues for actual humans to step in and solve, and as evidenced by our host's experience, getting an actual human on the line is often only possible after a frustrating tour of an endless array of menus and automation.
$14/hr plus social security, medicare, medicaid, unemployment, health insurance, and possible other things like life insurance and retirement funds, oh and the overhead involved in HR departments, leases for buildings in which people work (or at least for their connectivity if they work from home). People are expensive, managing them is difficult, and for the higher-ups it creates that much more separation between them and us in the form of making us suffer that much more.
THAT said, LLMs are not the solution. It would be better to streamline the hiring process by getting rid of lots of taxes and fees and regulations the government makes us all pay one way or another. That would reduce the need for bigger HR departments, too.
I've noticed a very interesting phenomenon with LLMs and writing--they're great at initial stages but once you start editing rather than getting more precise in each edit cycle they get worse unless you are extremely explicit about changes. I was doing a resume and they "suggestions" were made up facts and reduced precision in language throughout the entire documet.
I agree, Holly. While I can see things carefully controlled AI can help with, like a doctor using a from of AI to take notes so they can be more centered on the patient in the moment followed by carefully editing the the notes, I never use AI to write my podcast episodes or things like personal emails. For my episodes, I am worried that using AI to help in writing them in any way could eventually harm my copyright rights. And for all my faults as a writer, my emails, I think, still need to be as close to a personal expression from me as possible.
We have to wake up to this "progressive' for of personal and societal decline.
Progress is a double edged sword, and while I would never want you or my mother to have had to labor without hearing aides, I certainly would have advocated for a world where social media and maybe even email would have never existed. They are all stepping stones of one another and leading us no where good.
I often ask people do they remember a time when they could recite 20 or 30 phone numbers from memory compete with details about the person who's number it is and you can see the generational split, between those who were born before the cell phone and those born into it. The former group has that past experience, the later has no concept.
Researching and watching daily, the cognitive decline of my mom, has truly awakened me to your last point - as important as maintaining a fit and strong body is, it is just as important, if not more to do the same for your mind. They go hand in hand, but you can function and survive from a wheel chair or bed (no ideal at all), but when you lose your cognition, there is little a healthy body will be able to do for you, in remembering who or where you are.
I long for the past days of being able to talk to another human on the phone when there's technical issue rather than go through automated & AI centred systems that eventually hang up on you.
These AI agents are only going to get more sophisticated, and they're being actively created in the crypto space. Here's a recent piece celebrating the growth.
Right now they're relatively easy to spot, but I think as you hint at, they're only going to get better at passing the Turing test, while still providing crappy information or outright lies.
I don't understand the level of trust being put into these agents - especially when it comes to making financial decisions. In crypto, everyone talks about this trustless/permissionless process, but it strikes me that once you absolve yourself of even needing to trust an agent to make the right decisions, you're already doomed. These AI agents might be on a distributed ledger that means the transactions are registered and certified redundantly, but that doesn't mean the agents themselves are perfect actors. Have these people never picked up a sci-fi novel?
It is good that the eventual battery replacement worked, so you could continue to write and, indeed, continue to live more or less normally. You will have considered a backup device, but I wonder how you would devise such a thing within ordinary time and budget constraints. As for the main point of the post, it is a sad reflection, I suppose, that I am willing to accept being lied to in so many things. I may be more susceptible than most, but I don't think so. If beating my breast with a stone would cure it, I'd have permanent bruises below my neckline. But it is harder than that.
I LOL'd about your concept of mind gyms. I may steal it for a story. When we were kids, my brother and I used to play darts in our garage. We were too lazy to write down the scores, so we each learned to add up the scores in our heads and keep track of both our scores. I used to do that in the grocery store, adding up my bill while I waited in line. It was made more complex by having everything priced at $ x.99, but I found I could mentally compensate for that. The California sales tax system has defeated me however. I found if, instead of trying to mentally multiply my total by 1.0775, I could multiply by .07, add it to the total and then divide by 4 and multiply by 3, I could still work it out. Unfortunately the byzantine rules about what's taxable and what's not defeated me however, so I've given up that practice.
Also your question asking the AI that pretended to be human when it learned the times tables is brilliant. Take that Turing! Much simpler than the Voigt-Kampff Test.
I write a lot about AI and LLMs because I know things that don't seem common knowledge. I wonder if some enterprising lawyer could file a class action law suit against AI that lies, and whose lies do damage to clients. Where's Isaac Asimov when you need him? Do we need a fourth law of robotics, or maybe it's a deducible codicil from one of the three. Actually I contend that we should force AI to follow the 10 Commandments rather than just Asimov's 3 laws.
I have never felt white-hot rage the way I have dealing with these labyrinthian automated customer service loops and AI responses. There was a time you could hit "0" on your phone or request "agent" and get instantly routed to a human. It seems most companies have eliminated this feature. I would rather pay a premium on a product and get a real human than save $ and go through this nonsense.
As for AI making mistakes, yes, it does. When I'm in a hurry, I've used Grok to see whether the physics problems I come up with are solvable. I had a bad experience as a student when a professor wrote an off-the-cuff problem for the final exam in a physics course, and after EIGHT HOURS of grinding on a single problem he finally admitted that it wasn't solvable as written. I'm committed to never doing this to my students. The problem is that Grok occasionally flubs even simple physics problems. And when I've asked it to help me write creative problems, it sometimes writes trivial or unsolvable problems instead. Grok is still helpful to me as a professor, but it's not a substitute for doing the work.
Like you, Holly, I enjoy solving problems. Sitting down to work through homework or exam solutions is actually relaxing to me! One of my colleagues went through an old 1960s version of Halliday & Resnick (a classic freshman physics textbook) and over the course of a year solved every homework problem in the book, almost as a form of meditation.
I experimented with ChatGPT for code a while back. It pretty much sucks, but also, it has no concept of impossibility. I deliberately gave it a coding problem I knew to be impossible and it tried for hours.
I have a sinking feeling whenever I contemplate the likelihood that someone is planning to add "AI" to a payroll system, or a court system, or to say diet advice. ("The arsenic and sawdust diet is recommended by all leading physicians.") The next few years will be fun.
I remember when online and text started taking the place of paper. I've never been able to transition to online lists and I still use notebooks for notes and budgeting. There was always this nagging feeling that humanity may forget how to do it. Plus, what you don't use, you lose.
Mischievous question: What would Holly have done if the AI demanded proof she is real? How do we even know???
I recently had another in my long line of unrealised billion $ ideas. So many people are complaining they are reading AI generated "junk content" that proven authenticity now has rising value. Same with deep fake images. There is a fortune to be made for whoever creates a mechanism for proving reality - ie a "reality hash" tag added to content. A bit like the concept of totems inside the Inception movie.
I still remember how much everyone hated the old Clippy helper in Microsoft's apps in the early aughts. It was utterly reviled. I wrote a short story about it back in the day (a story in which I was rebelling against my muse, but turned back to her when I realized the alternative was Clippy).
And now we're surrounded by Clippy on steroids. I can't help but think, "This is how the world ends. This is how the world ends. This is how the world ends. Not with a bang but with ai bots."
Much respect for your decision to limit the comments function! Nevertheless, I’ll try to be brief. Retired early (60), eyes increasingly can’t handle long stretches of computer work. My ways of keeping mentally engaged in order to develop neural pathways: learning a new language (Brazilian Portuguese, to communicate better with friends); painting and drawing with both hands, introducing mixed media; listening to audiobooks of classic literature (most recently Crime and Punishment, Moby Dick); daily walks outside without earbuds; exploring new spiritual practices and ways of movement (Qi Gong); travel to new places when possible. Time will tell…
Holly - I went through something like this last year when I wanted to add my Apple Watch to a plan. It was not possible with the prepaid plan I had at the time. We have an internet provider at a rental home through which I could get a deal on add the watch and our other three family phones to the provider so I could use my watch. I spent days - a total of about 21 hours and even three way conference calls with prepaid provider human and new provider human. They kept insisting that everyone else could keep their number but I for some reason could not. I had changed my number the year before because of an issue with At&T that is still being litigated. I did not want to change it again. It is the number of all security challenges and verifications so no thank you. It turned out finally, after all of the exasperation and raised blood pressure time and time again and the shifting of responsibility to the other provider by each one, we could not get anywhere, that one rep asked about the ZIP CODES of the providers and that this was the source of the computer not allowing me to go ahead with service without changing my number. I think this basic limitation / question should be first or even second to ask. It should not take days and hours of my life to figure that basic thwarting mechanism. I have pages of notes of the conversations and the solution was actually a simple troubleshooting question.
They are sneaky and devious and dishonest to imply they are humans but unfortunately will deceive many.
Thank you for writing this. I sympathize with your frustration. I also really want to try the focus app and you have already written about it but would love to hear more now that you have used it for a while.
I am shocked at how quickly companies got bamboozled into using AI for customer service like this, or, if they were conscious of it, how uncaring they appear to be about giving off a bad impression to their customers by outright lying.
Yep. But they're saving, what, $14 an hour? LOL.
They must really believe people will not enforce their boundaries by going somewhere else for service. It's like they think people are devoted to them regardless of how they're treated. I suppose that's because so many people have stopped expecting normal levels of customer service. They get spoiled by people like that.
I lived in Japan for a long time. The customer service there is amazing, and a point of pride for a lot of people working there. Over the course of the nearly 10 years I spent there I had one negative experience, whereas back here in America I meet with regular disappointment. Routing regular work through LLMs removes a lot of peaceful interactions between customers and workers, and leaves only the difficult issues for actual humans to step in and solve, and as evidenced by our host's experience, getting an actual human on the line is often only possible after a frustrating tour of an endless array of menus and automation.
$14/hr plus social security, medicare, medicaid, unemployment, health insurance, and possible other things like life insurance and retirement funds, oh and the overhead involved in HR departments, leases for buildings in which people work (or at least for their connectivity if they work from home). People are expensive, managing them is difficult, and for the higher-ups it creates that much more separation between them and us in the form of making us suffer that much more.
THAT said, LLMs are not the solution. It would be better to streamline the hiring process by getting rid of lots of taxes and fees and regulations the government makes us all pay one way or another. That would reduce the need for bigger HR departments, too.
I've noticed a very interesting phenomenon with LLMs and writing--they're great at initial stages but once you start editing rather than getting more precise in each edit cycle they get worse unless you are extremely explicit about changes. I was doing a resume and they "suggestions" were made up facts and reduced precision in language throughout the entire documet.
I've notice this exact thing. AI is a convincing liar.
I agree, Holly. While I can see things carefully controlled AI can help with, like a doctor using a from of AI to take notes so they can be more centered on the patient in the moment followed by carefully editing the the notes, I never use AI to write my podcast episodes or things like personal emails. For my episodes, I am worried that using AI to help in writing them in any way could eventually harm my copyright rights. And for all my faults as a writer, my emails, I think, still need to be as close to a personal expression from me as possible.
Well done Holly!!
We have to wake up to this "progressive' for of personal and societal decline.
Progress is a double edged sword, and while I would never want you or my mother to have had to labor without hearing aides, I certainly would have advocated for a world where social media and maybe even email would have never existed. They are all stepping stones of one another and leading us no where good.
I often ask people do they remember a time when they could recite 20 or 30 phone numbers from memory compete with details about the person who's number it is and you can see the generational split, between those who were born before the cell phone and those born into it. The former group has that past experience, the later has no concept.
Researching and watching daily, the cognitive decline of my mom, has truly awakened me to your last point - as important as maintaining a fit and strong body is, it is just as important, if not more to do the same for your mind. They go hand in hand, but you can function and survive from a wheel chair or bed (no ideal at all), but when you lose your cognition, there is little a healthy body will be able to do for you, in remembering who or where you are.
I long for the past days of being able to talk to another human on the phone when there's technical issue rather than go through automated & AI centred systems that eventually hang up on you.
Interesting thought re: brains. Instantly made me think: "I like money. Do you like money? Let's be friends." 🙂 (a line from Idiocracy)
These AI agents are only going to get more sophisticated, and they're being actively created in the crypto space. Here's a recent piece celebrating the growth.
https://decrypt.co/294137/virtuals-protocol-tokens-skyrocket-as-ai-agent-demand-grows
Right now they're relatively easy to spot, but I think as you hint at, they're only going to get better at passing the Turing test, while still providing crappy information or outright lies.
I don't understand the level of trust being put into these agents - especially when it comes to making financial decisions. In crypto, everyone talks about this trustless/permissionless process, but it strikes me that once you absolve yourself of even needing to trust an agent to make the right decisions, you're already doomed. These AI agents might be on a distributed ledger that means the transactions are registered and certified redundantly, but that doesn't mean the agents themselves are perfect actors. Have these people never picked up a sci-fi novel?
It is good that the eventual battery replacement worked, so you could continue to write and, indeed, continue to live more or less normally. You will have considered a backup device, but I wonder how you would devise such a thing within ordinary time and budget constraints. As for the main point of the post, it is a sad reflection, I suppose, that I am willing to accept being lied to in so many things. I may be more susceptible than most, but I don't think so. If beating my breast with a stone would cure it, I'd have permanent bruises below my neckline. But it is harder than that.
I LOL'd about your concept of mind gyms. I may steal it for a story. When we were kids, my brother and I used to play darts in our garage. We were too lazy to write down the scores, so we each learned to add up the scores in our heads and keep track of both our scores. I used to do that in the grocery store, adding up my bill while I waited in line. It was made more complex by having everything priced at $ x.99, but I found I could mentally compensate for that. The California sales tax system has defeated me however. I found if, instead of trying to mentally multiply my total by 1.0775, I could multiply by .07, add it to the total and then divide by 4 and multiply by 3, I could still work it out. Unfortunately the byzantine rules about what's taxable and what's not defeated me however, so I've given up that practice.
Also your question asking the AI that pretended to be human when it learned the times tables is brilliant. Take that Turing! Much simpler than the Voigt-Kampff Test.
I write a lot about AI and LLMs because I know things that don't seem common knowledge. I wonder if some enterprising lawyer could file a class action law suit against AI that lies, and whose lies do damage to clients. Where's Isaac Asimov when you need him? Do we need a fourth law of robotics, or maybe it's a deducible codicil from one of the three. Actually I contend that we should force AI to follow the 10 Commandments rather than just Asimov's 3 laws.
I have never felt white-hot rage the way I have dealing with these labyrinthian automated customer service loops and AI responses. There was a time you could hit "0" on your phone or request "agent" and get instantly routed to a human. It seems most companies have eliminated this feature. I would rather pay a premium on a product and get a real human than save $ and go through this nonsense.
As for AI making mistakes, yes, it does. When I'm in a hurry, I've used Grok to see whether the physics problems I come up with are solvable. I had a bad experience as a student when a professor wrote an off-the-cuff problem for the final exam in a physics course, and after EIGHT HOURS of grinding on a single problem he finally admitted that it wasn't solvable as written. I'm committed to never doing this to my students. The problem is that Grok occasionally flubs even simple physics problems. And when I've asked it to help me write creative problems, it sometimes writes trivial or unsolvable problems instead. Grok is still helpful to me as a professor, but it's not a substitute for doing the work.
Like you, Holly, I enjoy solving problems. Sitting down to work through homework or exam solutions is actually relaxing to me! One of my colleagues went through an old 1960s version of Halliday & Resnick (a classic freshman physics textbook) and over the course of a year solved every homework problem in the book, almost as a form of meditation.
I experimented with ChatGPT for code a while back. It pretty much sucks, but also, it has no concept of impossibility. I deliberately gave it a coding problem I knew to be impossible and it tried for hours.
I doubt there will ever be a machine equivalent of human intuition.
I have a sinking feeling whenever I contemplate the likelihood that someone is planning to add "AI" to a payroll system, or a court system, or to say diet advice. ("The arsenic and sawdust diet is recommended by all leading physicians.") The next few years will be fun.
I remember when online and text started taking the place of paper. I've never been able to transition to online lists and I still use notebooks for notes and budgeting. There was always this nagging feeling that humanity may forget how to do it. Plus, what you don't use, you lose.
Mischievous question: What would Holly have done if the AI demanded proof she is real? How do we even know???
I recently had another in my long line of unrealised billion $ ideas. So many people are complaining they are reading AI generated "junk content" that proven authenticity now has rising value. Same with deep fake images. There is a fortune to be made for whoever creates a mechanism for proving reality - ie a "reality hash" tag added to content. A bit like the concept of totems inside the Inception movie.
I still remember how much everyone hated the old Clippy helper in Microsoft's apps in the early aughts. It was utterly reviled. I wrote a short story about it back in the day (a story in which I was rebelling against my muse, but turned back to her when I realized the alternative was Clippy).
And now we're surrounded by Clippy on steroids. I can't help but think, "This is how the world ends. This is how the world ends. This is how the world ends. Not with a bang but with ai bots."
Much respect for your decision to limit the comments function! Nevertheless, I’ll try to be brief. Retired early (60), eyes increasingly can’t handle long stretches of computer work. My ways of keeping mentally engaged in order to develop neural pathways: learning a new language (Brazilian Portuguese, to communicate better with friends); painting and drawing with both hands, introducing mixed media; listening to audiobooks of classic literature (most recently Crime and Punishment, Moby Dick); daily walks outside without earbuds; exploring new spiritual practices and ways of movement (Qi Gong); travel to new places when possible. Time will tell…
Holly - I went through something like this last year when I wanted to add my Apple Watch to a plan. It was not possible with the prepaid plan I had at the time. We have an internet provider at a rental home through which I could get a deal on add the watch and our other three family phones to the provider so I could use my watch. I spent days - a total of about 21 hours and even three way conference calls with prepaid provider human and new provider human. They kept insisting that everyone else could keep their number but I for some reason could not. I had changed my number the year before because of an issue with At&T that is still being litigated. I did not want to change it again. It is the number of all security challenges and verifications so no thank you. It turned out finally, after all of the exasperation and raised blood pressure time and time again and the shifting of responsibility to the other provider by each one, we could not get anywhere, that one rep asked about the ZIP CODES of the providers and that this was the source of the computer not allowing me to go ahead with service without changing my number. I think this basic limitation / question should be first or even second to ask. It should not take days and hours of my life to figure that basic thwarting mechanism. I have pages of notes of the conversations and the solution was actually a simple troubleshooting question.
They are sneaky and devious and dishonest to imply they are humans but unfortunately will deceive many.
Thank you for writing this. I sympathize with your frustration. I also really want to try the focus app and you have already written about it but would love to hear more now that you have used it for a while.
Merry Christmas !