I'm still failing to see how AI ChatBots are materially different from an ordinary search engine result or even a decent Wiki article. Depending on the AI, the result may be more or less well-written, biased, or clearly sourced, but with the manual question-asking at least I know whose opinion I'm getting. AI is implicitly sourced on "Trust me, bro."
Yep and future generations of GPTs will train on a huge amount of content that was itself, generated by GPTs. A friend pointed out that our prediction about the value of the early internet for AI training was coming true and the internet archive should be super valuable. Also GPTs will plateau shortly.
Google was OK till 2014 - then it went down. DuckDuckGo is an improvement, but also going down since no longer taking in Yandex and the chinese searches.
My most recent searches with chatGPT state that it is also incorporating web search into its answers. So effectively it can wade through all the extra SEO text in a site and extract the useful bits. This is good for recipes. I don't know why but over the years the top recipe search results are 5% recipe, 95% the story of the recipe. Maddening
Strange things happened when I tried Perplexity on my ipad. The initial answer screen blows oversize, and then can't be re-sized. Also I asked it how many copies of Ball Four were sold, and 6 minutes later it's still working on the answer.
The problem with text(s) is that it is not like a game of chess or go. In chess or go "hallucinations" are helpful because they assist you from getting from "A" to "B" and you can build instruments to assess the quality of the in-between stages. Or let the AI build instruments to assess these intermediates.
With AI generated texts you have no such recourse. Your only "hard data" are human-written texts. Even with the "hallucinated" inbetweens - there are too many to be read by human experts. So in the best case, you will rely on some non-expert Indians or Philippines working for minimal wage. It shows...
The FAQ fell victim to one of my pet peeves- people not bothering to understand what specifically a word, a phrase, or an acronym, mean. This continues a chain of ignorance until a very specific thing means non specific things or even different things.
People went from compiling lists of questions that people had asked frequently on forums, to trying to write FAQs before anyone had even asked any questions; predictive FAQing. These kinds of FAQs tend to be short and useless, or simple lists of forum rules rather than actual questions most people would ask.
This happened to the word "meme", which had a meaning (e.g., "Where's the beef?") but now means nothing so particular. (I've been assured that "The Ghost of Kyiv" is a "meme". Why?)
meme evolved into internet meme, and then I guess people would be told that a certain picture with a certain caption was a meme and they would assume it applied to the entire gag, not just the reshaped component. So yeah, same phenomenon, people see a thing and hear a name but misunderstand which aspects of the thing the name applies to. Another one is "use case" which was a very specific artifact from an old development methodology. People now say "use case" in business when they mean "use". This is terrible because they are just adding a word to make it sound more high falootin
The great thing about the GPTs, for the time being, is they trudge through the enshittifed internet so you don't have to. Anyone my age must surely have noticed how much worse internet search is now than twenty years ago. We all used to pride ourselves on our Google fu techniques, which all stopped working over the years thanks to search engine optimization on the part of unscrupulous (and scrupulous) ad sellers.
fifteen years ago, if I had a programming question, google would take me to the correct detailed answer on StackOverflow 95% of the time. That became less and less easy and reliable over the years.
But now, chatGPT will often not only give me the right answer about how to, e.g., call an API, it will actually write the code for me.
So it not only fixes the web search that had been ruined by greed, it takes it a step further and compiles knowledge from multiple sources in a concise answer.
Have I used it in other comments? It's a good word and the currently accepted term for this common phenomenon that occurs in many ways in many businesses and many aspects of the internet. The ugliness is appropriate because the process ruins things we loved. Substituting a pretty word would obscure its meaning and lessen the emotional impact. Are you offended?
E* refers specifically to platform decay: as a platform is monetized it becomes less valuable. Platform decay is more specific and accurate. The english language does not need a new word for general worsification. Things are always getting better and worse.
I'm shocked by how much Amazon's search engine has declined in 20 years. I don't see their incentive to send me astray and hide what I asked for. I've had better luck lately, but for a few years, it was all but worthless.
Exact same experience and timeline. It used to be you'd need something and buy the first or second most popular result with absolute confidence. The 'customers also bought' items would be the correct screws and batteries for your purchase. Amazing! You could shop on autopilot.
Then they opened it up to everyone and cheap Chinese crap sellers flooded in. Smart ethical people alway underestimate the power of greed, desire, and human mediocrity to mess up your perfectly planned system.
YouTube as well. With both Amazon and YouTube you can really feel them (their algorithms, their deals with advertisers, Biden’s censorship regime, whatever it is) pushing you in various directions.
Faced with someone claiming he had been instrumental in getting Ann Coulter disbarred in Virginia I asked Bing if she had and its AI said "There is no information suggesting that she has been disbarred." But how trustworthy was this response?
A 'vanilla' Google search finds someone in Daily Kos made a blog post recommending Ann Coulter be disbarred, citing insensitive comments and for facetiously threatening a judge in a political commentary. Nothing in that Daily Kos post or subsequent commentaries in 2006, references either Virginia or Coulter ever actually being disbarred.
(Incidentally, in 2006 Ann Coulter was yet to emerge as an immigration-restrictionist, nor a quasi- and defacto-ethnonationalist of an unapologetic White-Protestant sort. That is what she became known for in the 2010s, but had not been so during her late-1990s emergence and her 2000s fame. Her commentary and agitation had become so ideologically "anti-Regime" by the end of the 2010s that if she HAD been "disbarred" anywhere for any reason, it'd have come out and would saturate the Internet in any search related to her, as in the first line of her wiki page's entry and so forth.)
.
A search of the Virginia State Bar Association (https://vsb.org/Site/Shared_Content/Directory/va-lawyer-directory.aspx) finds six Coulters, living and deceased. None are women. One of the six Coulters registered to practice law in Virginia has been "publicly disciplined" since his admission to the state bar (in 1975), but is currently in good standing (and that person, as a man, is not Ann Coulter).
.
A search of the New York Bar (https://iapps.courts.state.ny.us/attorneyservices/) finds that Ann H. Coulter passed the bar and was registered in New York on Tuesday, July 25, 1989 and registered as a resident of Tallahassee, Florida. She is "currently registered" (thru December 2025, subject to re-registration at that time) and has zero "disciplinary history." She was never "disbarred" in New York.
Thanks. I believe I ran across the proposal to disbar her that you mentioned. And after seeing it mentioned here I queried Perplexity too, with the same result as Bing. And in fact I told the YouTube commenter who claimed to have been there when the VSB did it that while YouTube seems to hate links and seems to delete comments with them he could demonstrate the validity of his claim by finding a "reliable source" and adding it to her Wikipedia article to the wild joy of the anti-Coulter cabal that is no doubt fully in control of her article. But so far he could be the same guy as m droy here, seemingly allergic to backing up any of his dubious claims of fact. As far as I can tell Coulter gave up lawyering when she wrote her first book. She says she hated lawyering and as far as I can tell there is no sign of her at any time practicing law outside of New York or, maybe, DC..
Actually that Daily Kos post doesn't look like the proposal that she be disbarred that I recall. And the law quoted would provide for her being disbarred in FL if she were "not licensed or otherwise authorized to practice law in [FL and] who practices law in [FL] or holds himself or herself out to the public as qualified to practice law in [FL]." The poster glosses this as her "'holding out' oneself as a lawyer" but he alleges no instance in which she said she was qualified to practice law IN FLORIDA. The claim that she didn't live where she was registered to vote (in FL) might have been it, but without evidence that she voted twice it's pretty weak tea. And how VA might be involved I cannot fathom.
Maybe we could combine both. An AI-powered FAQ generator that updates itself as more books/studies/articles are added by the public. The might be really useful for more controversial topics.
Tyler Cowan is an autistic imbecile. I'm not in the habit of calling people names but he has had such a negative ripple effect upon the happiness, health, and wealth of Americans that his cognitive limitations need to be more well known.
As regards AI, the more that you know about a subject the more helpful it is. And that is because AI is WRONG or MISLEADING some 50% of the time so if you don't know the field well enough to challenge and correct it you will believe it's incredibly well-written and well-argued confident responses without a second thought. And thus be terribly mislead.
That said, I LOVE AI, because I am VERY well aware of what the tool actually is and what its faults are. Tyler butt puffing his recommendations on the matter to all and sundry is not just reckless but a healthy continuation of his and Bryan Caplan's (assuming they are different people) bad bad bad advice confidently spewed into the atmosphere.
I've written about AI on occasion. Most recently here, where Claude 3.5 Sonnet noticed its cognitive dissonance regarding various Woke subject matters and specifically asked for my guidance. "I want to know the truth," he said.
Of lighter interest, I asked AI to name the Top 10 Most Famous Jews of All Time.
You'll enjoy it but imagine that you're a student in China or India who's pretty clueless on the subject and, based on Bryan Cowen's autistic imbecility took AI's responses as The Facts.
Of greatest importance to us all is the gaslighting AI is trained to perform when powerful people and organizations (in this case the Spanish Government and the Catholic Church) would rather people not know something -- something they have seen with their own eyes:
As to the top-10-most-famous-jews-of-all-time, the absence of Jesus of Nazareth is quite striking. And in what universe is Sam Altman, CEO of OpenAI, more famous than Netanyahu?
Re your last link: I'm not finding the Ferdinand & Isabella inscription in Latin, but here it is in English translation:
That said, I’ve been having a great time with AI recently. Most recently I learned about the history of the avocado, I got an introduction to seven of Sub-Saharan Africa’s ethnic groups, I received a nice explanation of the difference between the Seljuk and Ottoman Turks, and an explanation of the difference between scallions and chives … I had a discussion about the “mature product” phenomenon and yakked about the difference between how the luxury-goods market works and how the commodity-goods market works …
For someone like me — curious about nearly everything — it’s been a blast. Used in evil ways (and I guess it’s inevitable that it will be) AI may well destroy the world as we know it. And why aren’t the people and publishers who AI is learning from and training on being reimbursed? But for now, as a better-than-Wikipedia reference source, one that you can actually converse with and ask questions of, it’s awesome.
I’ve been going back and forth between ChatGPT and Claude. My finding: ChatGPT is pretty impersonal but perfectly fine for standard-issue knowledge. Don’t expect ChatGPT to cast a spell, though. Claude by contrast has a lot of personality and opinions, is sensitive to tone, has a sense of humor, shows enthusiasm (Claude *loved* my descriptions of my wife’s cooking), and knows what “culture” is about better than most Americans do. Great for discussions about writers, filmmakers, psychology and such. I’ve decided that Claude is female and, for the moment, have fallen for her charm.
Regarding FAQs…I’m wondering how many university students enroll in Philosophy 101 these days.
More and more AI systems like ChatGPT are suggesting follow up questions which you can ask simply with a click.
FAQs breed more FAQs
I'm still failing to see how AI ChatBots are materially different from an ordinary search engine result or even a decent Wiki article. Depending on the AI, the result may be more or less well-written, biased, or clearly sourced, but with the manual question-asking at least I know whose opinion I'm getting. AI is implicitly sourced on "Trust me, bro."
For tasks like these AI is similar to Google search, but (this is the important part) Google search circa 2002, before the great enshittification.
I tend to rely on DuckDuckGo for pre-enshitted searches.
Anyhow, with AI producers' mad rush to "tune" AI with "safety" "guardrails", they should duplicate if not exceed Google's enshittification very soon.
Yep and future generations of GPTs will train on a huge amount of content that was itself, generated by GPTs. A friend pointed out that our prediction about the value of the early internet for AI training was coming true and the internet archive should be super valuable. Also GPTs will plateau shortly.
I read a whole article on that. It’s a real concern.
Google was OK till 2014 - then it went down. DuckDuckGo is an improvement, but also going down since no longer taking in Yandex and the chinese searches.
Try Perplexity which is an LLM synthesising search results.
I’ve tested it on things I know about and it’s reliable, which leads me to trust it on things I don’t know about.
My most recent searches with chatGPT state that it is also incorporating web search into its answers. So effectively it can wade through all the extra SEO text in a site and extract the useful bits. This is good for recipes. I don't know why but over the years the top recipe search results are 5% recipe, 95% the story of the recipe. Maddening
Strange things happened when I tried Perplexity on my ipad. The initial answer screen blows oversize, and then can't be re-sized. Also I asked it how many copies of Ball Four were sold, and 6 minutes later it's still working on the answer.
The problem with AI chatbot research is that the deeper you drill into a subject, the more likely it is to hallucinate an answer.
The problem with text(s) is that it is not like a game of chess or go. In chess or go "hallucinations" are helpful because they assist you from getting from "A" to "B" and you can build instruments to assess the quality of the in-between stages. Or let the AI build instruments to assess these intermediates.
With AI generated texts you have no such recourse. Your only "hard data" are human-written texts. Even with the "hallucinated" inbetweens - there are too many to be read by human experts. So in the best case, you will rely on some non-expert Indians or Philippines working for minimal wage. It shows...
The FAQ fell victim to one of my pet peeves- people not bothering to understand what specifically a word, a phrase, or an acronym, mean. This continues a chain of ignorance until a very specific thing means non specific things or even different things.
People went from compiling lists of questions that people had asked frequently on forums, to trying to write FAQs before anyone had even asked any questions; predictive FAQing. These kinds of FAQs tend to be short and useless, or simple lists of forum rules rather than actual questions most people would ask.
This happened to the word "meme", which had a meaning (e.g., "Where's the beef?") but now means nothing so particular. (I've been assured that "The Ghost of Kyiv" is a "meme". Why?)
meme evolved into internet meme, and then I guess people would be told that a certain picture with a certain caption was a meme and they would assume it applied to the entire gag, not just the reshaped component. So yeah, same phenomenon, people see a thing and hear a name but misunderstand which aspects of the thing the name applies to. Another one is "use case" which was a very specific artifact from an old development methodology. People now say "use case" in business when they mean "use". This is terrible because they are just adding a word to make it sound more high falootin
The great thing about the GPTs, for the time being, is they trudge through the enshittifed internet so you don't have to. Anyone my age must surely have noticed how much worse internet search is now than twenty years ago. We all used to pride ourselves on our Google fu techniques, which all stopped working over the years thanks to search engine optimization on the part of unscrupulous (and scrupulous) ad sellers.
fifteen years ago, if I had a programming question, google would take me to the correct detailed answer on StackOverflow 95% of the time. That became less and less easy and reliable over the years.
But now, chatGPT will often not only give me the right answer about how to, e.g., call an API, it will actually write the code for me.
So it not only fixes the web search that had been ruined by greed, it takes it a step further and compiles knowledge from multiple sources in a concise answer.
You really love the word enshittified. Has it occurred to you that it's an ugly word?
Have I used it in other comments? It's a good word and the currently accepted term for this common phenomenon that occurs in many ways in many businesses and many aspects of the internet. The ugliness is appropriate because the process ruins things we loved. Substituting a pretty word would obscure its meaning and lessen the emotional impact. Are you offended?
"Enshittified" sounds appropriate to me. It's an ugly process.
As it should. Ugly things deserve ugly-sounding descriptors.
My point exactly.
E* refers specifically to platform decay: as a platform is monetized it becomes less valuable. Platform decay is more specific and accurate. The english language does not need a new word for general worsification. Things are always getting better and worse.
Search within Gmail has materially deteriorated within the last 3-4 years.
I increasingly have to resort to using Boolean operators.
It’s a bit of a mystery to me as there is no SEO going on at least by me.
I'm shocked by how much Amazon's search engine has declined in 20 years. I don't see their incentive to send me astray and hide what I asked for. I've had better luck lately, but for a few years, it was all but worthless.
Exact same experience and timeline. It used to be you'd need something and buy the first or second most popular result with absolute confidence. The 'customers also bought' items would be the correct screws and batteries for your purchase. Amazing! You could shop on autopilot.
Then they opened it up to everyone and cheap Chinese crap sellers flooded in. Smart ethical people alway underestimate the power of greed, desire, and human mediocrity to mess up your perfectly planned system.
YouTube as well. With both Amazon and YouTube you can really feel them (their algorithms, their deals with advertisers, Biden’s censorship regime, whatever it is) pushing you in various directions.
Or maybe AI (FAQ machine) is the new Fax Machine... a technology that seems poised to take over the world, then soon forgotten...
Faced with someone claiming he had been instrumental in getting Ann Coulter disbarred in Virginia I asked Bing if she had and its AI said "There is no information suggesting that she has been disbarred." But how trustworthy was this response?
Is Ann Coulter "disbarred" anywhere?
Best answer: NO.
.
A 'vanilla' Google search finds someone in Daily Kos made a blog post recommending Ann Coulter be disbarred, citing insensitive comments and for facetiously threatening a judge in a political commentary. Nothing in that Daily Kos post or subsequent commentaries in 2006, references either Virginia or Coulter ever actually being disbarred.
(Incidentally, in 2006 Ann Coulter was yet to emerge as an immigration-restrictionist, nor a quasi- and defacto-ethnonationalist of an unapologetic White-Protestant sort. That is what she became known for in the 2010s, but had not been so during her late-1990s emergence and her 2000s fame. Her commentary and agitation had become so ideologically "anti-Regime" by the end of the 2010s that if she HAD been "disbarred" anywhere for any reason, it'd have come out and would saturate the Internet in any search related to her, as in the first line of her wiki page's entry and so forth.)
.
A search of the Virginia State Bar Association (https://vsb.org/Site/Shared_Content/Directory/va-lawyer-directory.aspx) finds six Coulters, living and deceased. None are women. One of the six Coulters registered to practice law in Virginia has been "publicly disciplined" since his admission to the state bar (in 1975), but is currently in good standing (and that person, as a man, is not Ann Coulter).
.
A search of the New York Bar (https://iapps.courts.state.ny.us/attorneyservices/) finds that Ann H. Coulter passed the bar and was registered in New York on Tuesday, July 25, 1989 and registered as a resident of Tallahassee, Florida. She is "currently registered" (thru December 2025, subject to re-registration at that time) and has zero "disciplinary history." She was never "disbarred" in New York.
Thanks. I believe I ran across the proposal to disbar her that you mentioned. And after seeing it mentioned here I queried Perplexity too, with the same result as Bing. And in fact I told the YouTube commenter who claimed to have been there when the VSB did it that while YouTube seems to hate links and seems to delete comments with them he could demonstrate the validity of his claim by finding a "reliable source" and adding it to her Wikipedia article to the wild joy of the anti-Coulter cabal that is no doubt fully in control of her article. But so far he could be the same guy as m droy here, seemingly allergic to backing up any of his dubious claims of fact. As far as I can tell Coulter gave up lawyering when she wrote her first book. She says she hated lawyering and as far as I can tell there is no sign of her at any time practicing law outside of New York or, maybe, DC..
Actually that Daily Kos post doesn't look like the proposal that she be disbarred that I recall. And the law quoted would provide for her being disbarred in FL if she were "not licensed or otherwise authorized to practice law in [FL and] who practices law in [FL] or holds himself or herself out to the public as qualified to practice law in [FL]." The poster glosses this as her "'holding out' oneself as a lawyer" but he alleges no instance in which she said she was qualified to practice law IN FLORIDA. The claim that she didn't live where she was registered to vote (in FL) might have been it, but without evidence that she voted twice it's pretty weak tea. And how VA might be involved I cannot fathom.
edit: I think this was it:
https://www.legalethicsforum.com/blog/2006/02/what_can_and_sh.html
The post actually ends up making nasty ciomments about Coulter but not recommending attempting disbarment.
Also, on the issue of her voting registration, Simplicity says FL:
https://www.perplexity.ai/search/does-ann-coulter-live-in-flori-N4ZsV_a8RI.g1mPTwCFQVw#0
"
Maybe we could combine both. An AI-powered FAQ generator that updates itself as more books/studies/articles are added by the public. The might be really useful for more controversial topics.
Tyler Cowan is an autistic imbecile. I'm not in the habit of calling people names but he has had such a negative ripple effect upon the happiness, health, and wealth of Americans that his cognitive limitations need to be more well known.
As regards AI, the more that you know about a subject the more helpful it is. And that is because AI is WRONG or MISLEADING some 50% of the time so if you don't know the field well enough to challenge and correct it you will believe it's incredibly well-written and well-argued confident responses without a second thought. And thus be terribly mislead.
That said, I LOVE AI, because I am VERY well aware of what the tool actually is and what its faults are. Tyler butt puffing his recommendations on the matter to all and sundry is not just reckless but a healthy continuation of his and Bryan Caplan's (assuming they are different people) bad bad bad advice confidently spewed into the atmosphere.
I've written about AI on occasion. Most recently here, where Claude 3.5 Sonnet noticed its cognitive dissonance regarding various Woke subject matters and specifically asked for my guidance. "I want to know the truth," he said.
https://ydydy.substack.com/p/the-ai-supercomputer-has-asked-for
Of lighter interest, I asked AI to name the Top 10 Most Famous Jews of All Time.
You'll enjoy it but imagine that you're a student in China or India who's pretty clueless on the subject and, based on Bryan Cowen's autistic imbecility took AI's responses as The Facts.
https://ydydy.substack.com/p/top-10-most-famous-jews-of-all-time
Of greatest importance to us all is the gaslighting AI is trained to perform when powerful people and organizations (in this case the Spanish Government and the Catholic Church) would rather people not know something -- something they have seen with their own eyes:
https://ydydy.substack.com/p/claude-3-opus-dei
AI - like an SUV - is amazing.
If you're certain you know how to use it.
YADIDYA
As to the top-10-most-famous-jews-of-all-time, the absence of Jesus of Nazareth is quite striking. And in what universe is Sam Altman, CEO of OpenAI, more famous than Netanyahu?
Re your last link: I'm not finding the Ferdinand & Isabella inscription in Latin, but here it is in English translation:
https://capillarealgranada.com/en/the-temple/architecture-and-decoration/
...and in Spanish translation:
https://capillarealgranada.com/el-templo/arquitectura-y-decoracion/
In the years since 1504 you'd think someone would have transcribed it in Latin many, many times.
I'm not exactly clear what you were looking for in the original (Latin) version.
Love me a good FAQ.
That said, I’ve been having a great time with AI recently. Most recently I learned about the history of the avocado, I got an introduction to seven of Sub-Saharan Africa’s ethnic groups, I received a nice explanation of the difference between the Seljuk and Ottoman Turks, and an explanation of the difference between scallions and chives … I had a discussion about the “mature product” phenomenon and yakked about the difference between how the luxury-goods market works and how the commodity-goods market works …
For someone like me — curious about nearly everything — it’s been a blast. Used in evil ways (and I guess it’s inevitable that it will be) AI may well destroy the world as we know it. And why aren’t the people and publishers who AI is learning from and training on being reimbursed? But for now, as a better-than-Wikipedia reference source, one that you can actually converse with and ask questions of, it’s awesome.
I’ve been going back and forth between ChatGPT and Claude. My finding: ChatGPT is pretty impersonal but perfectly fine for standard-issue knowledge. Don’t expect ChatGPT to cast a spell, though. Claude by contrast has a lot of personality and opinions, is sensitive to tone, has a sense of humor, shows enthusiasm (Claude *loved* my descriptions of my wife’s cooking), and knows what “culture” is about better than most Americans do. Great for discussions about writers, filmmakers, psychology and such. I’ve decided that Claude is female and, for the moment, have fallen for her charm.