It's all of the above AND the gaslighting. The fact that it's being forced on us everywhere, even though it doesn't work as well as humans. All of a sudden, every website and app I use now either has switched to AI or has an AI option they're trying to force me to use. It's like someone flipped a switch. It wasn't there, now it is. And it doesn't really work that well, but companies don't care. They're just so hell bent on saving money that they're forcing it on us and trying to convince us that this new tool is better. It's not. Not yet, at least. Maybe someday but right now it just sucks.
I agree because it’s impossible to call a company and talk to a human without going through a network of push this and press that and say keywords. I spent 7.5 hours on hold to finally get to a human about a unique tax issue — and no amount of “please hang up and look at our website or talk to our chatbot” could fix this. A day of my life, gone for a 3 minute conversation with a real person. ugh!!!!!
That's not new though. Customer service has been going downhill for years. The push and marketing for AI is really geared toward investors not consumers. Elon Musk is a trillionaire because he sells science fiction to investors. The corpocracy has realized their real customer is not the consumer of their goods and services but Wall Street investors large and small. That's why they are pushing AI as a replacement for workers, they're selling and competing for investors. This is a paradigm shift in marketing strategy where exploiting and abusing customers that have to use their technology is seen as a good thing by investors. It is in fact Wall Street investors who have ruined our world and the broligarchs are exploiters of this paradigm shift in marketing and creating wealth through science fiction and future PE potential. Warren Buffet's world of traditional investment is over and will no longer work in this world. Elon has proven that the product and customer service is less important than the pitch to Wall Street investors.
I don't know who Chase Bank's ad agency is, but they really caught the current zeitgeist with their TV ads that use Al Green's song "Tired of Being Alone." Have y'all seen it? Features a 20-something tired of having to interact with robots everywhere and the ad's message is that at OUR bank, you get to interact with humans. These ads have been airing for months now. They seemed to have hit on the cultural moment.
You have to hand it to them, convincing their customers that they are the good guys while being by far the world's largest financier of fossil fuel extraction and rain forest destruction. But, hey, focus on that smiling face and personal service.
I actually switched from a bank I’d gone to for 20 years because they kept hollowing out the number of people they hired and more importantly the level of expertise. My new credit union has a lot more people and expertise and it’s actually enjoyable to go to the bank again.
Credit unions are the answer. We are longtime CU members and love ours. In this age on online banking it doesn’t matter where your physical bank is if they TALk to you if you need help. We kept our WI CU when we moved to MA five years ago and have NO regrets.
If you bank with Citizens (which funds our concentration camps) definitely move your money out. But anyone can and should pick a CU to move to. They don’t run ads at all. But they do provide services.
I am a Chase customer who prefers to do my banking in person. I can attest that our local branches have been very people-centric for several years. Sometimes they go overboard with being too friendly, but it beats the alternative! Thanks for pointing this out, Lisa Z.
AI is literally being shoved down our throats, whether we like it or not.
And from my own experience, AI does NOT, CANNOT work as well as humans! It’s so frustrating. And infuriating.
So. Several weeks ago, I decided to simply ask AI if it was biased or just plain stupid. This was its response, verbatim -
“I don’t have feelings or human consciousness, so I can't be "stupid" or "biased" in the way a person is. Instead, AI systems process vast amounts of data using complex algorithms. How users and developers perceive this often comes down to the model's limitations and its training data.” (Yep)
“The "Stupid" Factor
I don’t truly "understand" the world; I process language and predict patterns based on billions of examples of human writing. Because I lack common sense and general intelligence, this can lead to: (well, at least AI was smart enough to admit that it lacks commonsense and general intelligence!) …
• Hallucinations: I might confidently state things that are completely factually incorrect.
• Lack of Originality: My responses are based on previously generated human knowledge, meaning I often synthesize existing data rather than creating genuine new insights. (Yep, there ya have it, folks! 😂)
“The Bias Factor
AI systems learn from the internet and historical texts, which inevitably contain human biases, stereotypes, and inaccuracies. This introduces a few types of bias: (Ah, yes, then there’s the BIAS FACTOR!)
• Data Bias: If my training data over-represents certain demographics or cultural perspectives, my answers will skew that way.
• Algorithmic Bias: The rules and preferences built by developers to ensure the AI is helpful and safe CAN inadvertently favor certain types of viewpoints or answers over others.” (And so it goes. The RULES AND PREFERENCES ie the “VIEWPOINTS” [polite way of saying that developers might possibly, inadvertently or unconsciously], mind you, skew the data towards “biased information rather than TRUTH. Another way of saying the ole “alternative facts” version of the truth!
No thanks. I’ll rely on my own commonsense, my own critical thinking skills, and my own ability to discern facts and truth over “alternative facts” and “inadvertent biases and viewpoints”!
And the primary reason I try to avoid APPs when possible.
Remember the movie, HER?
“Set in a near-future Los Angeles, the story centers on Theodore, who makes his living writing personal, touching letters for people who struggle to express themselves. Heartbroken and isolated following the end of his marriage, he purchases a new, artificially intelligent operating system designed to meet his every need. He names her Samantha. As they spend their days together—sharing experiences, thoughts, and emotional vulnerabilities—their friendship deepens into a complex, romantic relationship.)
Yes! I absolutely detest AI. It takes forever, like you said, to get to talk to a human being! In those almost everyday instances, I'm surprised I don't have a stroke! Also, it is no intrusive when I write anything on the computer. IT decides what I should say! We've lost all privacy and agency. I could go on and on....Thanks Sharon.
Have you noticed the company (like AT&T) computers are programmed to state to people on hold something like "Please be respectful of your customer service representative." They know they're driving us crazy and frustrating us to the point of warning us to not be rude. They are doing everything in their power to prevent us having to speak to a real person.
There are. I've been going through some medical issues this past year. The hospital has an AI agent on their main number. It fails 100% of time, and my voice is trained to not have an accent (after years in radio). Everyone hates it. I always ask the rep I eventually get to on any customer service call now for a direct number for my records. Usually they don’t have one, but sometimes they do. If they don't, I make a point of asking them to forward my complaint "up the chain" in hopes of change. (when it comes to the hospital, I always ask for (and get) an answerable number when sent to a department first time) If enough people start doing this,maybe?….
No kidding. The (Palantir based) NHS chatbot at our Doctors' office says the same thing when you're on hold. Don't disrespect the data harvesters I guess.
I've had the same experience with Social Security. My problem involved incorrect data at their end that kept me from signing in on-line. The automation insisted on trying to get me to use their website. Only saying "Agent" every time it paused its script saved me.
Right. But two days ago I phoned a company and an AI bot sounding close to a friendly human answered. "She" accomplished my request perfectly and in half the time a human would. It's real, it's there, and it will win.
With "luck" like that, you should buy a lottery ticket. The bots are programmed on FAQs, if your query is not in "their" menu it is an hour in a queue. Or just a dial tone.
Yes, I was disconnected 3 times in total - “we are transferring you to ….” And disconnect. I had four calls to get to 7.5 hours - and literally doing nothing but sitting on hold thinking any moment I could be disconnect if I let my mind wander and miss someone answering!
I called Best Buy to ask about recycling a printer. The website didn't specifically say that the store location recycled. AI answered the phone and since my question wasn't on the list I had no other options either it hung up or I hung up after about 5 tries
Your conclusion is not at all based on your statements. I've dealt with humans who solved my service problem quickly — they're real, they're there so by your logic wouldn't they win?
It's useful for some use cases. But only in proportion to the predictability of the use case. Sadly though, it may not have been full AI: many businesses used scripts in the past to facilitate a swift process (shipping companies like FedEx do this a lot). Nowadays they replace the phone person with a bot: it's more machine learning (which I'm quite a big fan of) than AI. Mind you that's true of much "AI".
I just went through this after I discovered my parked car was hit on the street hours after it happened. I tried to reach both my insurer (Lemonade) and the insurance company of the guy who hit my car (Progressive) by phone. It was literally impossible. In the meantime I was being barraged by robo text messages and emails asking irrelevant and absurd questions about the accident as if I had actively witnessed it, and receiving animated "questionaires" essentially demanding me to film the scene if I expected to get any service. It took up half my day dealing with this BS when a phone call with a human would have taken less than five minutes. The whole process was utterly inept and I am imagining a business world where person to person interactions will be entirely extinct. If we don't play along they won't pay our claims I guess. How anything of substance will get done I have no idea.
I spent 15 minutes trying to ask a "customer service rep" a question, prefacing it with a narrative of the original problem, which "she" didn't understand. Finally I asked, "are you are real person?" "I'm <company's> AI assistant.
I've encountered the same thing--Many hour on loops trying to get instructions or a for needed to settle an estate where I am the personal representative. At one credit union I got a person right away who quickly told me what I needed to get them to close out her account. It was such a breath of fresh air!
I agree, especially with the way AI is being forced on consumers. Google has all but destroyed their two signature products--Gmail and the Google search/browser--by forcing Google's AI into everything. I can no longer receive an email that does not have an AI summary bolted to the top--even when the email is only a single sentence! I can no longer search for anything without getting an entire first page of AI-generated results that usually have nothing at all to do with what I'm actually looking for.
And AI has even intruded tin Google Maps! Last month, I used Google Maps for what should have been a simple trip driving to Orlando International Airport. Maps spent the entire drive trying to force me onto toll roads, then actually had me get off the highway and make a 360-degree loop to get back on the same highway going the same way. It finally tried to get me back onto a toll road as I was entering the airport, then spent the entire time I was in the airport telling me to turn around because I had missed the Florida's Turnpike.
Forcing AI on consumers would lead to backlash anyway. But forcing extremely shitty AI on us guarantees people will hate it.
I just switched to Duck Duck Go this week, after getting exceedingly annoyed with search results that, once I got beyond the first page, had NOTHING to do with my question at all. Not tangentially, not if I squinted and stood on my head...nothing.
I've also been using DDG for some time, but I'm finding that the same search produces different results from day to day. It's almost as if the tool says "we showed this to you yesterday, and you clicked on it. You don't need to see it again."
Unf. it's not as good, especially on news searches. I use news searches for my work all the time, and sadly Google has always had the best news search engine. But now it sucks. I don't get deep archive stuff, just the current stuff (and no, it's not my settings.)
Agree completely. Just want to add to your last paragraph that forcing it on us everywhere (at every turn in your example 😉 sorry, couldn’t resist) is also a direct line to making us despise it.
That's why I use duckduckgo for search, and Proton email. The implication here is they're accessing the content of what's supposed to be private communications. They were doing that even before AI.
Lisa Z, I completely agree. AI makes so many mistakes, at my health insurer, at my language learning app -- it's scary, wondering how AI is going to kill me inadvertently. How can we rely on something so prone to error? And how do the broligarchs get so disgustingly rich by peddling something that doesn't always work? And why do we put up with it? Perhaps, no more.
Check out the paper "Agents of Chaos" It's about researchers who tricked hundreds of individual AIs into making bad decisions, demonstrating that a poorly designed system of linked AIs can generate chaos. Here's the link -- https://arxiv.org/abs/2602.20021
Google admits AI can't be trusted! In a German court their defense- that users could check the sources themselves via the links and were aware "that AI-generated information should not be trusted blindly" - was rejected.
From all accounts it's improving exponentially. My coding kids who used to scoff at it's abilities are all in on it now...for coding. They're just as concerned about societal effects.
AI coding capabilities are improving, but definitely not exponentially. Most of the improvements come from coding centric agentic frameworks which encode some software development best practices. Improvements in LLMs themself come largely from RLHF (Reinforcement Learning from Human Feedback), which is pretty slow and costly. AI is definitely not improving itself.
With the money currently involved in AI, and more to come it isn’t going anywhere, no matter what. The key is proper understanding and then good regulation, which isn’t going to happen with the GOP-controlled Congress we have now and Trump and his toadies in charge. Full disclosure: I use AI on my home LLM (MM Studio & Gemma 4-mlx) to build agents that manage mundane computer tasks. And I’m almost 77.
I still don't see what kind of regulation can stop AIs, in part because so many of them were trained to emulate human behavior. Criminals must be eagerly learning how to use AIs to "crime" better.
AI as deployed in the US is so obviously a trap: Most of the big models are loosing money big-time. They're trying to force us to find a workflow where AI helps, not hire and train the people who would otherwise do that workflow, then once we are dependent, jack-up the price. How else can they justify trillion dollar IPOs?
I see one alternative playing out in Europe: Move away from the software (Microsoft Office, Windows) that is pushing general AI. Focus AI in the niches where there is a benefit and minimize its cost.
Folks like Will Lockett and Cory Doctorow (the enshittification guy) keep posting stuff the mass media is ignoring: The cost of providing AI is creating a bubble bigger, by far, than any that has been seen in the tech biz, or any biz, for that matter.
There will be a shakeout, and then maybe your hope that niches where it can actually help us will be established by the companies that survive the shakeout.
Check out Ed Zitron as well. He can be a bit verbose and hyperbolic, but he dives deeply into the cold, hard reality that the huge investment numbers simply can't, and don't, add up. The frenzied folks are betting the farm and more on debt that can't be repaid and capital expenditures that will never produce a return on investment. The numbers are truly astonishing, and the limits of power, water, and the fallible technology itself are very real.
There are unanswered questions about the AI models appropriating vast amounts of copyrighted material to train their models. The demand for memory chips has already doubled the prices of everyday storage devices like external SSDs for consumers. Some big, important companies that jumped on the bandwagon of encouraging/forcing employees to maximize their AI use have blown through annual token budgets in a few months without actual productivity gains. China, meanwhile, is developing simpler, cheaper models that don't need orbiting data centers in space to work at scale. The absolute dependence of the whole pipedream on hyperscaled data centers that now face torches-and-pitchforks opposition from every political and regional corner is casting a long shadow over the whole rat race to dominate a future most of us are now openly rebelling against. No thank you, bros.
A very timely post from our favorite teacher today.
Also check out "House of El" (that last letter is a lower case "L"). Her videos run about half an hour. She has a nice set of talks about AI; was initially supportive and recently soured.
European nations are ditching Microsoft and other U.S. software products for other reasons. Congress recently passed a law that requires these companies to turn over any information they have upon request, even if it's stored outside the U.S.. The administration is using this law to grab personal information about European judges and police officials in order to sanction them if they do things DonnyJon doesn't like.
European nations (as well as others) are also divorcing their banking and credit card systems from the U.S. to make it impossible for the U.S. to impose sanctions on their citizens. I worry a little that sometime in the near future, my VISA card will become useless in Italy or Portugal (should we manage to get back there).
As with previous tech bubbles where companies hyped their new tech products that couldn't compete, most of those AI companies could fail miserably and it wouldn't affect the fact that the surviving AI is with us for the foreseeable future. Chinese AIs, which are cheaper to operate, may be the ones to become dominant.
Agree completely, too! You said it so well. It seems this time the new technology isn't just a tool - it's also a way of controlling us. I feel I'm losing control of my life. It's almost frightening. It feels as if it wants to take away the uniqueness of each individual person. AI is a "thing." WE are human beings with amazing, creative minds, with emotions, with compassion... and, we have the ability to make moral and ethical decisions and choices. AI is only the result of what it's been programmed to do - and it's being forced on us so someone else can make money off us. And... jobs will be lost. Maybe there is a place for it in certain situations, but I don't want to have to deal with it multiple times each and every day of my valuable life. I want to be the one in control of my life!
I find comfort in the thought that AIs can effectively replace corporate executives and techbros more readily than they can replace my plumber or my hair cutter.
Given that OpenAI has admitted that AI "hallucinations" can't be fixed, your "someday" may be some time. It will work well for some use cases, as long as you're not paying the bill for it. My own feelings are 1. DeepSeek works ok and is cheap, 2. There may be many half-built data centers full of out of date GPUs in out future
I remember when Cloud computing was all the rage 10-15 years ago. AT&T (for whom I worked at the time) set up several centers. Within two years they had a problem figuring out which servers were still running and at what capacity. I don't know if they ever found a solution, but I still don't put anything on the Cloud.
It seems to me that these huge data centers will have the same problem, only worse. And what does AI do when part of its "brain" quits working? Maybe develop dementia?
It sounds like at the time, AT&T had a room full of computers, not a cloud.
If you look at the technical papers from Google, Facebook/Meta, Microsoft, and Amazon you'll notice robust system management tools are used to run their Clouds.
Your second point is a very realistic concern. GPUs in general have many fewer error detection and correction features. Some people think the probabilistic math in AI system sidesteps this issue, but that is wrong.
Your second point is one most people don't know about. The speculators are buying warehouses full of processor chips for data centers that aren't even under preliminary construction yet. The primary supplier of chips, Nvidia, has an aggressive upgrade cycle that potentially renders last year's chips comparatively obsolete. So, by the time those bazillions of dollars worth of older chips are finally up and running several years later, they may be as useful as a 1980s PC is today.
"Customer service" is completely gone in most of the companies I buy from online. Got a problem with an order? Tell it to the AI bot and good luck with getting any satisfaction. These companies have scrubbed their websites of any and all contact phone numbers, email addresses and in some cases, physical addresses.
And what Paul said about search engines is so true. I keep saying that AI has killed internet searching. Not only are we supposed to use AI to search, but even when we can avoid it with some of the lesser-known search engines, all the search returns are garbage sites written by/with AI.
I don’t hate computer modeling, forecasting, even climate models are pretty useful…what I hate is the massive stealing other’s digitized works and marketing it as intelligent…predictive language models that have been taught by us is interesting…could be a useful editing tool, maybe. I do find the next word predictors a bit strange sometimes…they seriously change the meaning in my writing…the most likely next word is not the best word!
And the digitized world is strange world compared to human communication, to speech….it’s much more racist. The best information source is still WikiPedia…strangely fun group writing.
And they're not just getting rid of the direct human interaction. They're also trying to get rid of the thoughtful human design behind their standard interfaces. Every site/web app is now shoving a chatbot icon in your face. Half of them are even animated. Microsoft forced manufacturers to shoehorn a button for their proprietary AI into every laptop keyboard. Amazon's app pops up a chat box covering half your screen every time you try to look at a product listing. It's insanely intrusive.
Most people outside of IT don't even like using regular command-line interfaces. Why would they think we want to use *non-deterministic* command-line interfaces to do things that we used to be able to do directly with the click of a button or the swipe of a finger?
That is typical of new ventures, irritating for sure. Technology is wonderful in so many ways but as we've learned over time, it isn't always the best option. When we get eliminate personal contact, society suffers. What is exacerbating, I agree, is the way it is forced upon us without, in many incidences, an escape if we choose.
There are so many times these days where I think, "I *know* there's a way to do this with current AI that doesn't suck and actually would be an improvement over what they had before AI, but they're doing *this* instead!"
1. While there are plenty of past examples of defective outputs from earlier AIs, overall they keep improving at a scary pace. Clinging to amusing examples of hallucinations or confusion keeps us from accepting how much better they've become since then.
2. Humans don't all work very well. If you compare the AI output to the work of the most expert or talented humans, they may not measure up, but more and more AIs are winning awards and accolades from human judges when they didn't know it was an AI that produced the work.
3. Training and correcting a human worker doesn't always take, while changing the specification for the AI to follow generally corrects their capabilities over what they screwed up before. That's because a lot of work people are paid to do are not a natural fit to our evolved abilities.
Yes, AIs are electricity hogs, but they don't need sleep or bathrooms or lunch breaks or personal time off.
One upside for me: A lot of what executives and techbros do are some of the easiest jobs to replace with AI. 🙄
Yes, AI is literally being shoved down our throats.
And NO, AI does NOT, CANNOT work as well as humans!
So. Several weeks ago, I decided to simply ask AI if it was biased or just plain stupid. This was its response, verbatim -
“I don’t have feelings or human consciousness, so I can't be "stupid" or "biased" in the way a person is. Instead, AI systems process vast amounts of data using complex algorithms. How users and developers perceive this often comes down to the model's limitations and its training data.” (Yep)
“The "Stupid" Factor
I don’t truly "understand" the world; I process language and predict patterns based on billions of examples of human writing. Because I lack common sense and general intelligence, this can lead to: (well, at least AI was smart enough to admit it!) …
• Hallucinations: I might confidently state things that are completely factually incorrect.
• Lack of Originality: My responses are based on previously generated human knowledge, meaning I often synthesize existing data rather than creating genuine new insights. (Yep, there ya have it, folks! 😂)
“The Bias Factor
AI systems learn from the internet and historical texts, which inevitably contain human biases, stereotypes, and inaccuracies. This introduces a few types of bias: (Ah, yes, then there’s the BIAS FACTOR!)
• Data Bias: If my training data over-represents certain demographics or cultural perspectives, my answers will skew that way.
• Algorithmic Bias: The rules and preferences built by developers to ensure the AI is helpful and safe CAN inadvertently favor certain types of viewpoints or answers over others.” (And so it goes. The RULES AND PREFERENCES ie the “VIEWPOINTS” [polite way of saying that developers might possibly, inadvertently or unconsciously], mind you, skew the data towards “biased information rather than TRUTH. Another way of saying the ole “alternative facts” version of the truth!
No thanks. I’ll rely on my own commonsense, my own critical thinking skills, and my own ability to discern facts and truth over “alternative facts” and “inadvertent biases and viewpoints”!
And the primary reason I try to avoid APPs when possible.
Remember the movie, HER?
“Set in a near-future Los Angeles, the story centers on Theodore, who makes his living writing personal, touching letters for people who struggle to express themselves. Heartbroken and isolated following the end of his marriage, he purchases a new, artificially intelligent operating system designed to meet his every need. He names her Samantha. As they spend their days together—sharing experiences, thoughts, and emotional vulnerabilities—their friendship deepens into a complex, romantic relationship.)
Having people like Musk out front in the development of AI certainly doesn’t help. Gosh, if you believe “empathy is a weakness,” what could possibly go wrong?
And for someone who proclaims to be against government spending, virtually 100% of Musk’s core businesses have heavily benefited from – and in key cases been outright saved by – government support. Meaning taxpayer (our) support.
Over the past two decades, his companies have secured at least $38 billion in publicly-disclosed federal and state contracts, low-interest loans, subsidies, and tax credits.
Musk is the prince of enshitification. He didn't develop tesla concepts but bought them. He rolled out lots of stuff before it was ready, causing injuries, deaths, and lawsuits. His self-driving cars are the latest in a string of failures. He is more obsessed with having a tight structure of loyalists to command than getting the best from his resources.
WSJ has a great expose on Tesla self driving features. They only rely on cameras instead of that an other sensors as well. Waymo is kicking their proverbial butts. Their robo taxis in Texas aren't doing so well.
I did read about a Tesla semi truck that's looking promising. The truckers who are using it like it a lot.
Jay Leno made a test run in one of those semis. He said it was great, except that the windows don't open enough to use a fast food drive through window. They work very well in companies that can use them all day and park them at a depot all night to charge. My immediate thought was that one would be perfect for the local deliveries at grocery stores.
I think it was WSJ that I read about the semis. It was for short haul and the truckers especially liked that the driver's seat was in the center of the cab which gave them greater visibility. I have heard that the cybertruck was Elon's concept. I laugh whenever I see it. How could they call it a truck? You couldn't haul anything in the back, which is what you're supposed to be able to do in a truck. Even the new trucks have some bed in the back, even though they're so small you couldn't put a sheet of plywood in it. Imagine a load of firewood, rocks or manure? LOL Some sort of macho my fat old behind.
No doubt there are fine Tesla products, but Musk bought the company with these ideas baked in. Musk's attempt to add advanced features like self-driving are not working well.
Anyone who thinks empathy is a weakness is a bully. Bullies are cowards. They pick on those they see as weak and vulnerable. When punched in the nose, they run like chickens.
Trump is the archetypical bully. Iran punches back, and Trump loses interest—hides behind blaming, accepts a deal, moves on. That's what cowardice looks like at scale.
His reflecting pool turns out to be the perfect mirror of incompetence. A failed human with one great success: conning American voters.
I think his Colossus Data Center in Memphis is in big trouble and hand waving about national security is probably not enough. Grok is a disaster and implicated in the missile attack that killed the Iranian school children. I daresay a Congressional majority can use investigations to get to the bottom of at least the latter possibility.
An early study by Harvard and BCG found AI a useful tool for more senior employees who used it judiciously. Junior employees treated it as a solution and it reduced work quality. I find that on the team I lead. My programmers and writers can be more efficient with very specific tasks (translating analytic output to a table or graph; programming certain macros). But they always check the results and not infrequently find errors. More junior people are using it broadly and not verifying results, which has INCREASED my workload, reviewing and sending things back.
AI is a tool, not a solution, and it is impeding skill-building and deteriorating work product in my more junior workers.
Due to the way LLMs work under the hood, an LLM will improve the work of marginal employees by bringing them up to the mean. It will grind down the truly brilliant workers until, they too, are close to the mean.
It’s actually the opposite. High skill employees use it as a tool and know how to verify the output. Lower skill employees use it to DO the work without the skills to evaluate the output. As such, they don’t develop the skills and put the burden on their managers to review and send back.
Oh, and what it is doing to a college education! We all know about the rampant use of AI and LLMs to cheat, although for some reason students don’t think of it as cheating.
What infuriates me is how AI wants to tell me how to write. If I and others took every suggestion it puts forth, it wouldn’t be long before none of us will have a distinctive voice because we’ll all be sounding alike.
I shifted to in-class writing of minute papers because students think nothing of using AI to write discussion posts, short papers, and anything else. I had a grad student use AI by submitting other students’ discussion posts to revise as his own. He was suspended in his final class for his degree. I’ll bet he was doing that for all his classes. I now have students write something at the start of class so I can see how they write before they use AI. It’s so much more work for faculty. We are constantly being challenged to devise new assignments that aren’t easily done with AI.
I twice turned down tenure-track academic offers in favor of public health and industry jobs. I have to say, I am so happy now that I made those decisions. I do not envy the academic world you now live in.
I am so glad that I am at retirement age. I love teaching but I ache about the reality that not much learning seems to take place. Students today refuse to read books. They want to be given summaries from me on what books say. I refuse and always have to use power point or something else to tell students what they should know. Instead I ask them what we should question.
I actually think you are closer to what Tracy meant than you know.
AI can give a novice or marginal (ie. close to the incompetent line) a minimal set of basic tools.
The trick is that is nowhere near the mean. It is definitely "toward the mean".
If you increase the capacity of marginal workers, you increase the demand and workload on the skilled workers who have to make things production quality.
Couldn’t agree more. The irony that these are simply tools that can help increase productivity (greatly in some circumstances) if employed carefully and judiciously in a “never trust, always verify” manner while being hawked by the companies as nearly full human replacements is rich.
While they have stretched their capabilities more than I had expected, I will be shocked if this current technology can ever become anything more than useful tools that must be managed by humans.
I dislike AI because it was built through wholesale theft of intellectual property. I dislike it because a few companies are allowed to suck up vast amounts of electricity for data centers that make consumer electric bills go up, create environmental damage, and disrupt life in communities unlucky enough to host them. I dislike AI because individuals like Sam Altman should not be making decisions that will have worldwide repercussions. I also think it has been overhyped. It will have real benefits I'm sure, but where are the cost-benefit analyses that should be done with something this big and important?
I've read numerous articles about AI proving shockingly helpful in very specific ways: undiscovered patterns of protein folding in biology, identifying new objects and patterns using mind-boggling amounts of astronomical data, and even advances in arcane mathematical proofs. All of them very 'niche'.
I hope that an 'AI Moratorium' might make the bubble shrink rather than burst and that intelligent legislation will result in deliberate effective concentration of AI in ways it excels and prevent it from ruining search engines and customer assistance..
It's disrupting the economy/job market and we have datacenters being built near people's homes. All for corporate profit. Can't wait for the bubble to pop if I'm honest.
Our tech culture has dehumanized us already. Social media is doing incredible damage to our kids, not to mention how it is negatively impacting mental health and personal relationships. We're paying a price for technology "advancement."
Couldn't agree more about social media. It's not just kids it's damaging. My middle aged daughter is a high user of social media and has bought into the new religion of mental health. She also engages in a lot of the toxic on-line gender wars. She's become very self sabotaging and erratic while convinced that she's taking care of herself. It goes way beyond the usual emotional problems we all experience from time to time.
It's interesting that so many people who work in Tech stay far away from social media. It's toxic for profit propaganda.
So sorry to learn that, Sharon. Your daughter is not alone. I've seen a lot of people change (not for the better) and relationships fall apart because of social media abuse. I can't imagine AI would help any of this.
I'm more worried about the amount of money they control than the AI they control. Investors have given them an ocean of money based on mostly pure faith and now they can buy up vast swaths of other markets and control the media and politics...
The gravity of big piles of money sets in an no matter their actual success, those people now become entrenched in the system.
I'm an engineer. I have no issues with AI as a technology. But I don't like AI, the product. There are two main issues with the AI products:
1. The technology isn't mature enough, and it has enormous costs.
There is something called environmental sustainability. I don't think it needs much explanation. A project shouldn't be implemented just because it's commercially viable, unless it's environmentally sustainable too. The environmental impacts of a project has a cost too, that's often very hard to estimate. The problem is that this cost is almost never transferred to the project. Instead, it's often the community that pays that hefty price. That means these oligarchs have no incentive to be environmentally responsible.
So, what does AI technology maturity have anything to do with that? There are neural systems in nature that do the same amount of processing at a tiny fraction of their resource usage (power, water, pollution, etc). Our brain is a superb example. It can do with 120W what no datacenter of 400 MW can do. That means that artificial neural nets have a huge room for energy efficiency improvements. These big AI datacenters should be kept to a minimum until the energy efficiency is improved a few orders of magnitude. Instead, these oligarchs are using the immature technology to strip mine nature and wealth while they can.
2. The AI is trained and tuned to serve the oligarch's purposes.
Many AI models are designed and trained to serve the commercial interests of these oligarchs. Which means, they're very agreeable so that you spend more of their time and attention on them. Unfortunately, it leads to serious issues like AI psychosis.
Imagine if you could design and train powerful AI models yourselves? You could design ones that are more honest than agreeable. Ones that don't reinforce oligarchic propaganda. Ones that can be used in novel applications. Ones that are modular. Ones that have better safeguards. Or even a new AI architectures. Possibilities are endless.
But these currently need massive resource hogging datacenters and several PhDs. The entry cost is so high that independent research is nearly impossible. This again circles back to the issue of improving energy efficiency of AI architectures. We need to improve them so that all these can be done by one person on a desktop, rather than by a few dozen PhDs in a gargantuan datacenter.
More accurately, employees are using AI in ways that don't justify the cost.
In some ways the issue is that AI is too flexible.
If you don't know what you are doing, "build me an app to use an AI API to process this dataset into a database" can turn into a black hole sucking up tokens.
If you know what you are doing, scrub the input data, and think about the schema and operations, you can get a better result faster with some AI assistance.
From a business manager perspective it is hard to know which employee is doing which workflow by only counting tokens.
There's a cartoon going around that asks what might have happened if some of history's greatest thinkers used AI. It highlight's Copernicus being told that, no the sun does indeed revolve around the earth, because everyone knows it. Einstein and several others were included as well with similar "results".
Americans also see a political inability to regulate tech. Europeans have privacy protection, and their governments successfully limit tech overreach and hold companies accountable for violations. In America, it's the Wild West in the name of "freedom." No wonder Americans are scared.
The same thing is true about food additives. The European Union requires proof that ingredients are not harmful. The US view is that additives are presumed innocent, so long as similar additives have not been found harmful. And a difference as well in the amount of sugar in cereals and desserts.
The EU generally protects consumers much better than the US across the board - health & beauty products, auto safety, the named tech and food regulations. Better for their population and the environment. Maybe they can’t squeeze out every last penny for c-suite pay and shareholders, but that is both an unsustainable and an unethical way to operate a business/industry/economy. Companies still make plenty of money and their economy as a whole, per Paul’s recent posts, is doing fine.
It's called the precautionary principle. A substance has to be proven safe before it can be released to the public. In the US every substance is assumed to be safe unless it's proven to be harmful.
The other part of why people are turning against AI, and tech overall is that instead of using AI to clean up and stop trolls, fake accounts, and the general spreading of lies, AI is being used to generate more of this. Trolls helped swing the 2016 election. They have been a factor in every election since. People want the the nonsense to stop and AI could help tremendously with that instead of helping to create more disruption.
Enshittification is the issue i've experienced. The bots on websites don't work well and it's almost impossible to get a human to help. Try booking a flight these days. Horrible.
Good piece. The datacenter point feels like the heart of it. People can't see the chips or the financing, but they can see the land, the power lines, and the water being pulled from their towns. That's where the backlash gets real. And it matters for the whole chain. A town that blocks a substation or a water permit can slow a datacenter more than any worry about jobs ever will. So the part of AI people can actually touch — the power and the water — may end up setting the limit on how fast the rest of it can grow.
It's all of the above AND the gaslighting. The fact that it's being forced on us everywhere, even though it doesn't work as well as humans. All of a sudden, every website and app I use now either has switched to AI or has an AI option they're trying to force me to use. It's like someone flipped a switch. It wasn't there, now it is. And it doesn't really work that well, but companies don't care. They're just so hell bent on saving money that they're forcing it on us and trying to convince us that this new tool is better. It's not. Not yet, at least. Maybe someday but right now it just sucks.
I agree because it’s impossible to call a company and talk to a human without going through a network of push this and press that and say keywords. I spent 7.5 hours on hold to finally get to a human about a unique tax issue — and no amount of “please hang up and look at our website or talk to our chatbot” could fix this. A day of my life, gone for a 3 minute conversation with a real person. ugh!!!!!
That's not new though. Customer service has been going downhill for years. The push and marketing for AI is really geared toward investors not consumers. Elon Musk is a trillionaire because he sells science fiction to investors. The corpocracy has realized their real customer is not the consumer of their goods and services but Wall Street investors large and small. That's why they are pushing AI as a replacement for workers, they're selling and competing for investors. This is a paradigm shift in marketing strategy where exploiting and abusing customers that have to use their technology is seen as a good thing by investors. It is in fact Wall Street investors who have ruined our world and the broligarchs are exploiters of this paradigm shift in marketing and creating wealth through science fiction and future PE potential. Warren Buffet's world of traditional investment is over and will no longer work in this world. Elon has proven that the product and customer service is less important than the pitch to Wall Street investors.
I don't know who Chase Bank's ad agency is, but they really caught the current zeitgeist with their TV ads that use Al Green's song "Tired of Being Alone." Have y'all seen it? Features a 20-something tired of having to interact with robots everywhere and the ad's message is that at OUR bank, you get to interact with humans. These ads have been airing for months now. They seemed to have hit on the cultural moment.
You have to hand it to them, convincing their customers that they are the good guys while being by far the world's largest financier of fossil fuel extraction and rain forest destruction. But, hey, focus on that smiling face and personal service.
I actually switched from a bank I’d gone to for 20 years because they kept hollowing out the number of people they hired and more importantly the level of expertise. My new credit union has a lot more people and expertise and it’s actually enjoyable to go to the bank again.
Credit unions are the answer. We are longtime CU members and love ours. In this age on online banking it doesn’t matter where your physical bank is if they TALk to you if you need help. We kept our WI CU when we moved to MA five years ago and have NO regrets.
If you bank with Citizens (which funds our concentration camps) definitely move your money out. But anyone can and should pick a CU to move to. They don’t run ads at all. But they do provide services.
I am a Chase customer who prefers to do my banking in person. I can attest that our local branches have been very people-centric for several years. Sometimes they go overboard with being too friendly, but it beats the alternative! Thanks for pointing this out, Lisa Z.
Bingo. this is thee
PERFECT Summation of Capitalism,
Unhinged, and the Death of 'our little Democracy.'
.
Bravissimo.
AI is literally being shoved down our throats, whether we like it or not.
And from my own experience, AI does NOT, CANNOT work as well as humans! It’s so frustrating. And infuriating.
So. Several weeks ago, I decided to simply ask AI if it was biased or just plain stupid. This was its response, verbatim -
“I don’t have feelings or human consciousness, so I can't be "stupid" or "biased" in the way a person is. Instead, AI systems process vast amounts of data using complex algorithms. How users and developers perceive this often comes down to the model's limitations and its training data.” (Yep)
“The "Stupid" Factor
I don’t truly "understand" the world; I process language and predict patterns based on billions of examples of human writing. Because I lack common sense and general intelligence, this can lead to: (well, at least AI was smart enough to admit that it lacks commonsense and general intelligence!) …
• Hallucinations: I might confidently state things that are completely factually incorrect.
• Lack of Originality: My responses are based on previously generated human knowledge, meaning I often synthesize existing data rather than creating genuine new insights. (Yep, there ya have it, folks! 😂)
“The Bias Factor
AI systems learn from the internet and historical texts, which inevitably contain human biases, stereotypes, and inaccuracies. This introduces a few types of bias: (Ah, yes, then there’s the BIAS FACTOR!)
• Data Bias: If my training data over-represents certain demographics or cultural perspectives, my answers will skew that way.
• Algorithmic Bias: The rules and preferences built by developers to ensure the AI is helpful and safe CAN inadvertently favor certain types of viewpoints or answers over others.” (And so it goes. The RULES AND PREFERENCES ie the “VIEWPOINTS” [polite way of saying that developers might possibly, inadvertently or unconsciously], mind you, skew the data towards “biased information rather than TRUTH. Another way of saying the ole “alternative facts” version of the truth!
No thanks. I’ll rely on my own commonsense, my own critical thinking skills, and my own ability to discern facts and truth over “alternative facts” and “inadvertent biases and viewpoints”!
And the primary reason I try to avoid APPs when possible.
Remember the movie, HER?
“Set in a near-future Los Angeles, the story centers on Theodore, who makes his living writing personal, touching letters for people who struggle to express themselves. Heartbroken and isolated following the end of his marriage, he purchases a new, artificially intelligent operating system designed to meet his every need. He names her Samantha. As they spend their days together—sharing experiences, thoughts, and emotional vulnerabilities—their friendship deepens into a complex, romantic relationship.)
Is this where we’re headed? To our own demise?
…bigly sigh
"AI is literally being shoved down our throats"
At least AI knows what "literally" means.
Yes! I absolutely detest AI. It takes forever, like you said, to get to talk to a human being! In those almost everyday instances, I'm surprised I don't have a stroke! Also, it is no intrusive when I write anything on the computer. IT decides what I should say! We've lost all privacy and agency. I could go on and on....Thanks Sharon.
Have you noticed the company (like AT&T) computers are programmed to state to people on hold something like "Please be respectful of your customer service representative." They know they're driving us crazy and frustrating us to the point of warning us to not be rude. They are doing everything in their power to prevent us having to speak to a real person.
I’ve noticed that too. There must be a lot frustrated and angry people out there.
My wife has learned to ignore the cursing and screaming coming from my home office.
Same. I didn't have such a potty mouth until automated answering "people" became the standard. There I sit, arguing with a f'n computer...
Congratulations; mine hasn't, goldern it!
There are. I've been going through some medical issues this past year. The hospital has an AI agent on their main number. It fails 100% of time, and my voice is trained to not have an accent (after years in radio). Everyone hates it. I always ask the rep I eventually get to on any customer service call now for a direct number for my records. Usually they don’t have one, but sometimes they do. If they don't, I make a point of asking them to forward my complaint "up the chain" in hopes of change. (when it comes to the hospital, I always ask for (and get) an answerable number when sent to a department first time) If enough people start doing this,maybe?….
So frustrating,
No kidding. The (Palantir based) NHS chatbot at our Doctors' office says the same thing when you're on hold. Don't disrespect the data harvesters I guess.
The worst case is flight attendants who can't be replaced by a robotic drinks cart and who get physically assaulted pretty often.
Bank tellers have worked behind bank counters for generations, now work only at drive-up windows during extended Friday hours.
There's lots of physical evidence that customers are dangerous and have been consistently dangerous in the past.
The trick is to call sales. Then ask to have the issue fixed. Sales will then transfer you to support. It usually works.
Nice! I must remember that one.
Or transfer to the billing department. They ALWAYS have a person answering THAT phone.
I strongly agree — especially if you want Microsoft technical support, ha ha.
It actually worked!
I even got a case number… which, of course, didn’t work.
Therefore, I love AI and Copilot. At least there’s a chance I can solve the problem myself.
Ditto.
Sounds like the plot of the Terminator movies, except Arnold is too old to save us.
I found for the IRS just keep shouting "Agent" or " Refund" over and over again. It has worked for other site too.
I've had the same experience with Social Security. My problem involved incorrect data at their end that kept me from signing in on-line. The automation insisted on trying to get me to use their website. Only saying "Agent" every time it paused its script saved me.
I find myself yelingl "agent!" or "representarive!" alot
Right. But two days ago I phoned a company and an AI bot sounding close to a friendly human answered. "She" accomplished my request perfectly and in half the time a human would. It's real, it's there, and it will win.
With "luck" like that, you should buy a lottery ticket. The bots are programmed on FAQs, if your query is not in "their" menu it is an hour in a queue. Or just a dial tone.
Yes, I was disconnected 3 times in total - “we are transferring you to ….” And disconnect. I had four calls to get to 7.5 hours - and literally doing nothing but sitting on hold thinking any moment I could be disconnect if I let my mind wander and miss someone answering!
I called Best Buy to ask about recycling a printer. The website didn't specifically say that the store location recycled. AI answered the phone and since my question wasn't on the list I had no other options either it hung up or I hung up after about 5 tries
Your conclusion is not at all based on your statements. I've dealt with humans who solved my service problem quickly — they're real, they're there so by your logic wouldn't they win?
n of 1.
It's useful for some use cases. But only in proportion to the predictability of the use case. Sadly though, it may not have been full AI: many businesses used scripts in the past to facilitate a swift process (shipping companies like FedEx do this a lot). Nowadays they replace the phone person with a bot: it's more machine learning (which I'm quite a big fan of) than AI. Mind you that's true of much "AI".
I've only had that happen when I want to give them money to pay a bill.
Found the same thing .
Marry you robot and live happy!
I also had this happen once recently.
I just went through this after I discovered my parked car was hit on the street hours after it happened. I tried to reach both my insurer (Lemonade) and the insurance company of the guy who hit my car (Progressive) by phone. It was literally impossible. In the meantime I was being barraged by robo text messages and emails asking irrelevant and absurd questions about the accident as if I had actively witnessed it, and receiving animated "questionaires" essentially demanding me to film the scene if I expected to get any service. It took up half my day dealing with this BS when a phone call with a human would have taken less than five minutes. The whole process was utterly inept and I am imagining a business world where person to person interactions will be entirely extinct. If we don't play along they won't pay our claims I guess. How anything of substance will get done I have no idea.
💯agree. They want you to figure out your problem without having to talk to a human because it costs them money. Infuriating.
THIS. A chirpy AI voice is what I get whenever I call "customer service".
I spent 15 minutes trying to ask a "customer service rep" a question, prefacing it with a narrative of the original problem, which "she" didn't understand. Finally I asked, "are you are real person?" "I'm <company's> AI assistant.
I've encountered the same thing--Many hour on loops trying to get instructions or a for needed to settle an estate where I am the personal representative. At one credit union I got a person right away who quickly told me what I needed to get them to close out her account. It was such a breath of fresh air!
Someone pointed out recently that cursing at the AI will get a human. What a world!
I agree, especially with the way AI is being forced on consumers. Google has all but destroyed their two signature products--Gmail and the Google search/browser--by forcing Google's AI into everything. I can no longer receive an email that does not have an AI summary bolted to the top--even when the email is only a single sentence! I can no longer search for anything without getting an entire first page of AI-generated results that usually have nothing at all to do with what I'm actually looking for.
And AI has even intruded tin Google Maps! Last month, I used Google Maps for what should have been a simple trip driving to Orlando International Airport. Maps spent the entire drive trying to force me onto toll roads, then actually had me get off the highway and make a 360-degree loop to get back on the same highway going the same way. It finally tried to get me back onto a toll road as I was entering the airport, then spent the entire time I was in the airport telling me to turn around because I had missed the Florida's Turnpike.
Forcing AI on consumers would lead to backlash anyway. But forcing extremely shitty AI on us guarantees people will hate it.
You should consider getting off Google and switching to Duck Duck Go or other alternatives.
I just switched to Duck Duck Go this week, after getting exceedingly annoyed with search results that, once I got beyond the first page, had NOTHING to do with my question at all. Not tangentially, not if I squinted and stood on my head...nothing.
I've been using duckduckgo for many years now. I'll never go back to anything Google. And Chrome phones home. I'll stick with Firefox.
I've also been using DDG for some time, but I'm finding that the same search produces different results from day to day. It's almost as if the tool says "we showed this to you yesterday, and you clicked on it. You don't need to see it again."
I use it too
Unf. it's not as good, especially on news searches. I use news searches for my work all the time, and sadly Google has always had the best news search engine. But now it sucks. I don't get deep archive stuff, just the current stuff (and no, it's not my settings.)
Use “-ai” to turn off ai.
That didn’t work for me
From Google? That’s odd, I just tested it and it worked.
I switched many years ago. It was recommended by ACLU.
Agree completely. Just want to add to your last paragraph that forcing it on us everywhere (at every turn in your example 😉 sorry, couldn’t resist) is also a direct line to making us despise it.
That's why I use duckduckgo for search, and Proton email. The implication here is they're accessing the content of what's supposed to be private communications. They were doing that even before AI.
It's all part of the modern surveillance state.
why I use apple maps now....used Waze before but they are fully magafied and I think now owned by google
Use “-ai” in Google search to turn it off.
Oooh! I gotta try that!
That is interesting. It now explains why Google Maps has not been working properly lately for me and my husband.
Yeah, Google Maps always prefers toll roads. I think they get paid to do that. Try MapQuest out. I used to use that years ago.
Lisa Z, I completely agree. AI makes so many mistakes, at my health insurer, at my language learning app -- it's scary, wondering how AI is going to kill me inadvertently. How can we rely on something so prone to error? And how do the broligarchs get so disgustingly rich by peddling something that doesn't always work? And why do we put up with it? Perhaps, no more.
Check out the paper "Agents of Chaos" It's about researchers who tricked hundreds of individual AIs into making bad decisions, demonstrating that a poorly designed system of linked AIs can generate chaos. Here's the link -- https://arxiv.org/abs/2602.20021
Very interesting, will send this to friends
😱 scary
Evil in, evil out.
Google admits AI can't be trusted! In a German court their defense- that users could check the sources themselves via the links and were aware "that AI-generated information should not be trusted blindly" - was rejected.
There's a real problem when something works 99% of the time. You trust it and don't check for that 1% as thoroughly as you should.
From all accounts it's improving exponentially. My coding kids who used to scoff at it's abilities are all in on it now...for coding. They're just as concerned about societal effects.
AI coding capabilities are improving, but definitely not exponentially. Most of the improvements come from coding centric agentic frameworks which encode some software development best practices. Improvements in LLMs themself come largely from RLHF (Reinforcement Learning from Human Feedback), which is pretty slow and costly. AI is definitely not improving itself.
With the money currently involved in AI, and more to come it isn’t going anywhere, no matter what. The key is proper understanding and then good regulation, which isn’t going to happen with the GOP-controlled Congress we have now and Trump and his toadies in charge. Full disclosure: I use AI on my home LLM (MM Studio & Gemma 4-mlx) to build agents that manage mundane computer tasks. And I’m almost 77.
I still don't see what kind of regulation can stop AIs, in part because so many of them were trained to emulate human behavior. Criminals must be eagerly learning how to use AIs to "crime" better.
Correction: →Current versions of AI← make so many mistakes.
Also, do you know every time you are encountering AI?
It's like saying you can always tell if someone is wearing a toupée or a wig, when you don't notice the well-fitted ones.
I'm sure if you think hard enough, you'll remember all the ways that human employees have screwed up or lied to you before AI was a thing.
AI as deployed in the US is so obviously a trap: Most of the big models are loosing money big-time. They're trying to force us to find a workflow where AI helps, not hire and train the people who would otherwise do that workflow, then once we are dependent, jack-up the price. How else can they justify trillion dollar IPOs?
I see one alternative playing out in Europe: Move away from the software (Microsoft Office, Windows) that is pushing general AI. Focus AI in the niches where there is a benefit and minimize its cost.
Folks like Will Lockett and Cory Doctorow (the enshittification guy) keep posting stuff the mass media is ignoring: The cost of providing AI is creating a bubble bigger, by far, than any that has been seen in the tech biz, or any biz, for that matter.
There will be a shakeout, and then maybe your hope that niches where it can actually help us will be established by the companies that survive the shakeout.
Check out Ed Zitron as well. He can be a bit verbose and hyperbolic, but he dives deeply into the cold, hard reality that the huge investment numbers simply can't, and don't, add up. The frenzied folks are betting the farm and more on debt that can't be repaid and capital expenditures that will never produce a return on investment. The numbers are truly astonishing, and the limits of power, water, and the fallible technology itself are very real.
There are unanswered questions about the AI models appropriating vast amounts of copyrighted material to train their models. The demand for memory chips has already doubled the prices of everyday storage devices like external SSDs for consumers. Some big, important companies that jumped on the bandwagon of encouraging/forcing employees to maximize their AI use have blown through annual token budgets in a few months without actual productivity gains. China, meanwhile, is developing simpler, cheaper models that don't need orbiting data centers in space to work at scale. The absolute dependence of the whole pipedream on hyperscaled data centers that now face torches-and-pitchforks opposition from every political and regional corner is casting a long shadow over the whole rat race to dominate a future most of us are now openly rebelling against. No thank you, bros.
A very timely post from our favorite teacher today.
Also check out "House of El" (that last letter is a lower case "L"). Her videos run about half an hour. She has a nice set of talks about AI; was initially supportive and recently soured.
European nations are ditching Microsoft and other U.S. software products for other reasons. Congress recently passed a law that requires these companies to turn over any information they have upon request, even if it's stored outside the U.S.. The administration is using this law to grab personal information about European judges and police officials in order to sanction them if they do things DonnyJon doesn't like.
European nations (as well as others) are also divorcing their banking and credit card systems from the U.S. to make it impossible for the U.S. to impose sanctions on their citizens. I worry a little that sometime in the near future, my VISA card will become useless in Italy or Portugal (should we manage to get back there).
As with previous tech bubbles where companies hyped their new tech products that couldn't compete, most of those AI companies could fail miserably and it wouldn't affect the fact that the surviving AI is with us for the foreseeable future. Chinese AIs, which are cheaper to operate, may be the ones to become dominant.
Agree completely, too! You said it so well. It seems this time the new technology isn't just a tool - it's also a way of controlling us. I feel I'm losing control of my life. It's almost frightening. It feels as if it wants to take away the uniqueness of each individual person. AI is a "thing." WE are human beings with amazing, creative minds, with emotions, with compassion... and, we have the ability to make moral and ethical decisions and choices. AI is only the result of what it's been programmed to do - and it's being forced on us so someone else can make money off us. And... jobs will be lost. Maybe there is a place for it in certain situations, but I don't want to have to deal with it multiple times each and every day of my valuable life. I want to be the one in control of my life!
I find comfort in the thought that AIs can effectively replace corporate executives and techbros more readily than they can replace my plumber or my hair cutter.
The irony is they are NOT saving money. They are just spending our tax dollars and squandering our resources building the Data Centers.
... for the benefit of the kakistocrats and broligarchy.....
Given that OpenAI has admitted that AI "hallucinations" can't be fixed, your "someday" may be some time. It will work well for some use cases, as long as you're not paying the bill for it. My own feelings are 1. DeepSeek works ok and is cheap, 2. There may be many half-built data centers full of out of date GPUs in out future
I remember when Cloud computing was all the rage 10-15 years ago. AT&T (for whom I worked at the time) set up several centers. Within two years they had a problem figuring out which servers were still running and at what capacity. I don't know if they ever found a solution, but I still don't put anything on the Cloud.
It seems to me that these huge data centers will have the same problem, only worse. And what does AI do when part of its "brain" quits working? Maybe develop dementia?
"My mind is going, Dave...."
Monitoring servers is job one in a Cloud system.
It sounds like at the time, AT&T had a room full of computers, not a cloud.
If you look at the technical papers from Google, Facebook/Meta, Microsoft, and Amazon you'll notice robust system management tools are used to run their Clouds.
Your second point is a very realistic concern. GPUs in general have many fewer error detection and correction features. Some people think the probabilistic math in AI system sidesteps this issue, but that is wrong.
Your second point is one most people don't know about. The speculators are buying warehouses full of processor chips for data centers that aren't even under preliminary construction yet. The primary supplier of chips, Nvidia, has an aggressive upgrade cycle that potentially renders last year's chips comparatively obsolete. So, by the time those bazillions of dollars worth of older chips are finally up and running several years later, they may be as useful as a 1980s PC is today.
"Customer service" is completely gone in most of the companies I buy from online. Got a problem with an order? Tell it to the AI bot and good luck with getting any satisfaction. These companies have scrubbed their websites of any and all contact phone numbers, email addresses and in some cases, physical addresses.
And what Paul said about search engines is so true. I keep saying that AI has killed internet searching. Not only are we supposed to use AI to search, but even when we can avoid it with some of the lesser-known search engines, all the search returns are garbage sites written by/with AI.
A pox on all their houses.
The consumer is the test bed.
The consumer always has been in my former field of telecom.
Yup. You can't build a test system as big as the real world.
I don’t hate computer modeling, forecasting, even climate models are pretty useful…what I hate is the massive stealing other’s digitized works and marketing it as intelligent…predictive language models that have been taught by us is interesting…could be a useful editing tool, maybe. I do find the next word predictors a bit strange sometimes…they seriously change the meaning in my writing…the most likely next word is not the best word!
And the digitized world is strange world compared to human communication, to speech….it’s much more racist. The best information source is still WikiPedia…strangely fun group writing.
AI is neither! Coding and stealing…
Like sharks smelling blood in the water, rapacious capitalists are circling the scent of quick fortune.
And they're not just getting rid of the direct human interaction. They're also trying to get rid of the thoughtful human design behind their standard interfaces. Every site/web app is now shoving a chatbot icon in your face. Half of them are even animated. Microsoft forced manufacturers to shoehorn a button for their proprietary AI into every laptop keyboard. Amazon's app pops up a chat box covering half your screen every time you try to look at a product listing. It's insanely intrusive.
Most people outside of IT don't even like using regular command-line interfaces. Why would they think we want to use *non-deterministic* command-line interfaces to do things that we used to be able to do directly with the click of a button or the swipe of a finger?
That is typical of new ventures, irritating for sure. Technology is wonderful in so many ways but as we've learned over time, it isn't always the best option. When we get eliminate personal contact, society suffers. What is exacerbating, I agree, is the way it is forced upon us without, in many incidences, an escape if we choose.
There are so many times these days where I think, "I *know* there's a way to do this with current AI that doesn't suck and actually would be an improvement over what they had before AI, but they're doing *this* instead!"
"even though it doesn't work as well as humans"
----
That's an attitude I have to push back against:
1. While there are plenty of past examples of defective outputs from earlier AIs, overall they keep improving at a scary pace. Clinging to amusing examples of hallucinations or confusion keeps us from accepting how much better they've become since then.
2. Humans don't all work very well. If you compare the AI output to the work of the most expert or talented humans, they may not measure up, but more and more AIs are winning awards and accolades from human judges when they didn't know it was an AI that produced the work.
3. Training and correcting a human worker doesn't always take, while changing the specification for the AI to follow generally corrects their capabilities over what they screwed up before. That's because a lot of work people are paid to do are not a natural fit to our evolved abilities.
Yes, AIs are electricity hogs, but they don't need sleep or bathrooms or lunch breaks or personal time off.
One upside for me: A lot of what executives and techbros do are some of the easiest jobs to replace with AI. 🙄
Yes, AI is literally being shoved down our throats.
And NO, AI does NOT, CANNOT work as well as humans!
So. Several weeks ago, I decided to simply ask AI if it was biased or just plain stupid. This was its response, verbatim -
“I don’t have feelings or human consciousness, so I can't be "stupid" or "biased" in the way a person is. Instead, AI systems process vast amounts of data using complex algorithms. How users and developers perceive this often comes down to the model's limitations and its training data.” (Yep)
“The "Stupid" Factor
I don’t truly "understand" the world; I process language and predict patterns based on billions of examples of human writing. Because I lack common sense and general intelligence, this can lead to: (well, at least AI was smart enough to admit it!) …
• Hallucinations: I might confidently state things that are completely factually incorrect.
• Lack of Originality: My responses are based on previously generated human knowledge, meaning I often synthesize existing data rather than creating genuine new insights. (Yep, there ya have it, folks! 😂)
“The Bias Factor
AI systems learn from the internet and historical texts, which inevitably contain human biases, stereotypes, and inaccuracies. This introduces a few types of bias: (Ah, yes, then there’s the BIAS FACTOR!)
• Data Bias: If my training data over-represents certain demographics or cultural perspectives, my answers will skew that way.
• Algorithmic Bias: The rules and preferences built by developers to ensure the AI is helpful and safe CAN inadvertently favor certain types of viewpoints or answers over others.” (And so it goes. The RULES AND PREFERENCES ie the “VIEWPOINTS” [polite way of saying that developers might possibly, inadvertently or unconsciously], mind you, skew the data towards “biased information rather than TRUTH. Another way of saying the ole “alternative facts” version of the truth!
No thanks. I’ll rely on my own commonsense, my own critical thinking skills, and my own ability to discern facts and truth over “alternative facts” and “inadvertent biases and viewpoints”!
And the primary reason I try to avoid APPs when possible.
Remember the movie, HER?
“Set in a near-future Los Angeles, the story centers on Theodore, who makes his living writing personal, touching letters for people who struggle to express themselves. Heartbroken and isolated following the end of his marriage, he purchases a new, artificially intelligent operating system designed to meet his every need. He names her Samantha. As they spend their days together—sharing experiences, thoughts, and emotional vulnerabilities—their friendship deepens into a complex, romantic relationship.)
Is this where we’re headed? To our own demise?
…bigly sigh
Having people like Musk out front in the development of AI certainly doesn’t help. Gosh, if you believe “empathy is a weakness,” what could possibly go wrong?
It could produce the current Republican party. What could possibly go wrong?
It could produce the next Trump. Things really would go wrong.
It already has Trump characteristics. It will resort to lying instead of just saying, “I don’t know.”
And for someone who proclaims to be against government spending, virtually 100% of Musk’s core businesses have heavily benefited from – and in key cases been outright saved by – government support. Meaning taxpayer (our) support.
Over the past two decades, his companies have secured at least $38 billion in publicly-disclosed federal and state contracts, low-interest loans, subsidies, and tax credits.
So, where's the wood chipper now?
YES!! Good point!
Musk is the prince of enshitification. He didn't develop tesla concepts but bought them. He rolled out lots of stuff before it was ready, causing injuries, deaths, and lawsuits. His self-driving cars are the latest in a string of failures. He is more obsessed with having a tight structure of loyalists to command than getting the best from his resources.
WSJ has a great expose on Tesla self driving features. They only rely on cameras instead of that an other sensors as well. Waymo is kicking their proverbial butts. Their robo taxis in Texas aren't doing so well.
I did read about a Tesla semi truck that's looking promising. The truckers who are using it like it a lot.
Jay Leno made a test run in one of those semis. He said it was great, except that the windows don't open enough to use a fast food drive through window. They work very well in companies that can use them all day and park them at a depot all night to charge. My immediate thought was that one would be perfect for the local deliveries at grocery stores.
I think it was WSJ that I read about the semis. It was for short haul and the truckers especially liked that the driver's seat was in the center of the cab which gave them greater visibility. I have heard that the cybertruck was Elon's concept. I laugh whenever I see it. How could they call it a truck? You couldn't haul anything in the back, which is what you're supposed to be able to do in a truck. Even the new trucks have some bed in the back, even though they're so small you couldn't put a sheet of plywood in it. Imagine a load of firewood, rocks or manure? LOL Some sort of macho my fat old behind.
No doubt there are fine Tesla products, but Musk bought the company with these ideas baked in. Musk's attempt to add advanced features like self-driving are not working well.
Anyone who thinks empathy is a weakness is a bully. Bullies are cowards. They pick on those they see as weak and vulnerable. When punched in the nose, they run like chickens.
Trump is the archetypical bully. Iran punches back, and Trump loses interest—hides behind blaming, accepts a deal, moves on. That's what cowardice looks like at scale.
His reflecting pool turns out to be the perfect mirror of incompetence. A failed human with one great success: conning American voters.
For the record, Musk's AI system is particularly bad.
He is a self appointed front person for the hype, but he is nowhere near the front of development.
I think his Colossus Data Center in Memphis is in big trouble and hand waving about national security is probably not enough. Grok is a disaster and implicated in the missile attack that killed the Iranian school children. I daresay a Congressional majority can use investigations to get to the bottom of at least the latter possibility.
An early study by Harvard and BCG found AI a useful tool for more senior employees who used it judiciously. Junior employees treated it as a solution and it reduced work quality. I find that on the team I lead. My programmers and writers can be more efficient with very specific tasks (translating analytic output to a table or graph; programming certain macros). But they always check the results and not infrequently find errors. More junior people are using it broadly and not verifying results, which has INCREASED my workload, reviewing and sending things back.
AI is a tool, not a solution, and it is impeding skill-building and deteriorating work product in my more junior workers.
Due to the way LLMs work under the hood, an LLM will improve the work of marginal employees by bringing them up to the mean. It will grind down the truly brilliant workers until, they too, are close to the mean.
I'm not sure this is really what companies want.
It’s actually the opposite. High skill employees use it as a tool and know how to verify the output. Lower skill employees use it to DO the work without the skills to evaluate the output. As such, they don’t develop the skills and put the burden on their managers to review and send back.
Oh, and what it is doing to a college education! We all know about the rampant use of AI and LLMs to cheat, although for some reason students don’t think of it as cheating.
What infuriates me is how AI wants to tell me how to write. If I and others took every suggestion it puts forth, it wouldn’t be long before none of us will have a distinctive voice because we’ll all be sounding alike.
I love that some professors are requiring that writing be done in class under direct observation.
The only way to learn writing skills is to write!
I shifted to in-class writing of minute papers because students think nothing of using AI to write discussion posts, short papers, and anything else. I had a grad student use AI by submitting other students’ discussion posts to revise as his own. He was suspended in his final class for his degree. I’ll bet he was doing that for all his classes. I now have students write something at the start of class so I can see how they write before they use AI. It’s so much more work for faculty. We are constantly being challenged to devise new assignments that aren’t easily done with AI.
I twice turned down tenure-track academic offers in favor of public health and industry jobs. I have to say, I am so happy now that I made those decisions. I do not envy the academic world you now live in.
I am so glad that I am at retirement age. I love teaching but I ache about the reality that not much learning seems to take place. Students today refuse to read books. They want to be given summaries from me on what books say. I refuse and always have to use power point or something else to tell students what they should know. Instead I ask them what we should question.
I actually think you are closer to what Tracy meant than you know.
AI can give a novice or marginal (ie. close to the incompetent line) a minimal set of basic tools.
The trick is that is nowhere near the mean. It is definitely "toward the mean".
If you increase the capacity of marginal workers, you increase the demand and workload on the skilled workers who have to make things production quality.
Bingo.
Couldn’t agree more. The irony that these are simply tools that can help increase productivity (greatly in some circumstances) if employed carefully and judiciously in a “never trust, always verify” manner while being hawked by the companies as nearly full human replacements is rich.
While they have stretched their capabilities more than I had expected, I will be shocked if this current technology can ever become anything more than useful tools that must be managed by humans.
Exactly.
A"I" is a "tool". No "tool" is better or worse than the hands that wield it. Too many of the hands are guided by rabid egos and grandiose delusions.
Martian colonies anyone?
I'd settle for the Moon, if Elon Musk migrated. Come to think about it, Mars is the way to go. He couldn't come back from Mars.
I thought the moon was going to be the 51st state…?
😆
I dislike AI because it was built through wholesale theft of intellectual property. I dislike it because a few companies are allowed to suck up vast amounts of electricity for data centers that make consumer electric bills go up, create environmental damage, and disrupt life in communities unlucky enough to host them. I dislike AI because individuals like Sam Altman should not be making decisions that will have worldwide repercussions. I also think it has been overhyped. It will have real benefits I'm sure, but where are the cost-benefit analyses that should be done with something this big and important?
I am with you!
I've read numerous articles about AI proving shockingly helpful in very specific ways: undiscovered patterns of protein folding in biology, identifying new objects and patterns using mind-boggling amounts of astronomical data, and even advances in arcane mathematical proofs. All of them very 'niche'.
I hope that an 'AI Moratorium' might make the bubble shrink rather than burst and that intelligent legislation will result in deliberate effective concentration of AI in ways it excels and prevent it from ruining search engines and customer assistance..
"make the bubble shrink rather than burst" is a nice turn of phrase.
Thank you!
Agree. I wish! But do you think “intelligent legislation” in our current dysfunctional situation has a snowball’s chance in hell?!
Not a snowball's chance - not any “intelligent legislation” regarding *anything*.
That's from different more application specific AI models though.
I agree with you.! Well written. Can we follow each other
I’m with you too.
It's disrupting the economy/job market and we have datacenters being built near people's homes. All for corporate profit. Can't wait for the bubble to pop if I'm honest.
Agree! Datacenters shouldn't be forced on wildlife either. Who's paying attention to how the sound, lights and pollution are affecting wildlife?
Who? Not this Administration.
The noise is driving away birds, for one thing. And human beings can't stand it either.
Burning fossil fuels to power AI datacenters is just inexcusable malfeasance with the knowledge we possess about climate change.
Datacenters should be required to fully run on renewables. Period.
The other big issue being water consumption of course.
Don't mention climate change. They'll come and take you away.
Datacenter water use is easy to fix with regulation or pricing. It is just a minor cooling optimization.
Power is a real issue.
Ditto Robert! I can't wait either for the bubble to pop! Thanks.
Our tech culture has dehumanized us already. Social media is doing incredible damage to our kids, not to mention how it is negatively impacting mental health and personal relationships. We're paying a price for technology "advancement."
Not to mention the damage done to our political environment.
Couldn't agree more about social media. It's not just kids it's damaging. My middle aged daughter is a high user of social media and has bought into the new religion of mental health. She also engages in a lot of the toxic on-line gender wars. She's become very self sabotaging and erratic while convinced that she's taking care of herself. It goes way beyond the usual emotional problems we all experience from time to time.
It's interesting that so many people who work in Tech stay far away from social media. It's toxic for profit propaganda.
So sorry to learn that, Sharon. Your daughter is not alone. I've seen a lot of people change (not for the better) and relationships fall apart because of social media abuse. I can't imagine AI would help any of this.
I am not worried about AI as much as I am about the people controlling it and our government that can use it.
That's the issue with AI - none of us have a clue what's going on behind the screen so we can't control it.
"Pay no attention to the man behind the curtain."
I'm more worried about the amount of money they control than the AI they control. Investors have given them an ocean of money based on mostly pure faith and now they can buy up vast swaths of other markets and control the media and politics...
The gravity of big piles of money sets in an no matter their actual success, those people now become entrenched in the system.
I'm an engineer. I have no issues with AI as a technology. But I don't like AI, the product. There are two main issues with the AI products:
1. The technology isn't mature enough, and it has enormous costs.
There is something called environmental sustainability. I don't think it needs much explanation. A project shouldn't be implemented just because it's commercially viable, unless it's environmentally sustainable too. The environmental impacts of a project has a cost too, that's often very hard to estimate. The problem is that this cost is almost never transferred to the project. Instead, it's often the community that pays that hefty price. That means these oligarchs have no incentive to be environmentally responsible.
So, what does AI technology maturity have anything to do with that? There are neural systems in nature that do the same amount of processing at a tiny fraction of their resource usage (power, water, pollution, etc). Our brain is a superb example. It can do with 120W what no datacenter of 400 MW can do. That means that artificial neural nets have a huge room for energy efficiency improvements. These big AI datacenters should be kept to a minimum until the energy efficiency is improved a few orders of magnitude. Instead, these oligarchs are using the immature technology to strip mine nature and wealth while they can.
2. The AI is trained and tuned to serve the oligarch's purposes.
Many AI models are designed and trained to serve the commercial interests of these oligarchs. Which means, they're very agreeable so that you spend more of their time and attention on them. Unfortunately, it leads to serious issues like AI psychosis.
Imagine if you could design and train powerful AI models yourselves? You could design ones that are more honest than agreeable. Ones that don't reinforce oligarchic propaganda. Ones that can be used in novel applications. Ones that are modular. Ones that have better safeguards. Or even a new AI architectures. Possibilities are endless.
But these currently need massive resource hogging datacenters and several PhDs. The entry cost is so high that independent research is nearly impossible. This again circles back to the issue of improving energy efficiency of AI architectures. We need to improve them so that all these can be done by one person on a desktop, rather than by a few dozen PhDs in a gargantuan datacenter.
Lately I have been reading that some companies are starting to limit employees’ usage of AI because it’s driving up their costs.
Uber blew through its projected annual AI token budget by April.
More accurately, employees are using AI in ways that don't justify the cost.
In some ways the issue is that AI is too flexible.
If you don't know what you are doing, "build me an app to use an AI API to process this dataset into a database" can turn into a black hole sucking up tokens.
If you know what you are doing, scrub the input data, and think about the schema and operations, you can get a better result faster with some AI assistance.
From a business manager perspective it is hard to know which employee is doing which workflow by only counting tokens.
There's a cartoon going around that asks what might have happened if some of history's greatest thinkers used AI. It highlight's Copernicus being told that, no the sun does indeed revolve around the earth, because everyone knows it. Einstein and several others were included as well with similar "results".
Americans also see a political inability to regulate tech. Europeans have privacy protection, and their governments successfully limit tech overreach and hold companies accountable for violations. In America, it's the Wild West in the name of "freedom." No wonder Americans are scared.
The same thing is true about food additives. The European Union requires proof that ingredients are not harmful. The US view is that additives are presumed innocent, so long as similar additives have not been found harmful. And a difference as well in the amount of sugar in cereals and desserts.
The EU generally protects consumers much better than the US across the board - health & beauty products, auto safety, the named tech and food regulations. Better for their population and the environment. Maybe they can’t squeeze out every last penny for c-suite pay and shareholders, but that is both an unsustainable and an unethical way to operate a business/industry/economy. Companies still make plenty of money and their economy as a whole, per Paul’s recent posts, is doing fine.
The MAGA crowd tells me that the EU is a complete disaster.
It's called the precautionary principle. A substance has to be proven safe before it can be released to the public. In the US every substance is assumed to be safe unless it's proven to be harmful.
The other part of why people are turning against AI, and tech overall is that instead of using AI to clean up and stop trolls, fake accounts, and the general spreading of lies, AI is being used to generate more of this. Trolls helped swing the 2016 election. They have been a factor in every election since. People want the the nonsense to stop and AI could help tremendously with that instead of helping to create more disruption.
Enshittification is the issue i've experienced. The bots on websites don't work well and it's almost impossible to get a human to help. Try booking a flight these days. Horrible.
How about the sense government no longer provides significant oversight, and the game is rigged for big business, thank you republicans.
Thank you Fleetwooz--lack of government oversight and regulations. The game sure is rigged for big business as you said.
Good piece. The datacenter point feels like the heart of it. People can't see the chips or the financing, but they can see the land, the power lines, and the water being pulled from their towns. That's where the backlash gets real. And it matters for the whole chain. A town that blocks a substation or a water permit can slow a datacenter more than any worry about jobs ever will. So the part of AI people can actually touch — the power and the water — may end up setting the limit on how fast the rest of it can grow.
I will continue to use my mind to learn what I need to know.
Let AI solve diseases and things that i dont have any say about.
However there should be severe restrictions on data centers and things like grok producing unseemly images.
Yes! “Nudification” is not an AI high point.
What the essay fails to convey about the CEO of Microsoft is that he was angry about his peers saying the quiet part out loud.
Krugman should have elaborated here because it adds to his overall assertion about why people hate in AI.
Satya was essentially complaining about his Confederacy of villains spilling the beans on their agenda.
Exactly. He’s part of the reason.
He certainly is!