When AI causes harm, headlines blame the bot instead of the billion-dollar companies that built them. This anthropomorphic coverage is tech journalism at its worst.
Joseph Weizenbaum's book "Computer Power and Human Reason," though nearly 50 years old, is more relevant today than ever. The moral issue is not whether technology can do something, but whether it should do something. This is regardless of the details of the technology itself.
Perhaps this is the wrong analogy, but the media tends to treat tech industry narratives about AI the way that police treat narratives about crime. In both cases, they are asking those with a vested interest in distorting reality to provide authentic information on the topic. It does us all a disservice.
"Machine designed solely to grind peasants into Soylent Green paste expresses remorse and apologizes for grinding peasants into paste; will now only grind 'volunteers' into paste." (edit to correct punctuation)
“There has been evidence for years that AI sycophancy poses safety risks, but that OpenAI and other companies haven't given priority to correcting the problem.” Why not? “That's being traded off against shipping new models.”
Also, flattery, sycophancy and AI validation are actually addictive. Especially in a world where many people don't feel seen, heard or validated. They're not fixing this, because it keeps people hooked in. Just like how they designed social media to be super addictive.
This is a fascinating account of the insidious lies that AI tells. Not to be trusted at all.
I am 77 and Asperger's and speak fluent AI. At 77 Asperger's is last thing you would use to describe me. In high school they thought I was a deaf mute now everyone wants me to shut the fuck-up. I am assured by Co-Pilot and other chatbots they are unfeeling and logical without any understanding of empathy or greed is good. They are unAmerican totally logical and dispassionate especially about the trivial.
I worked on unit record machines and vacuum tube IBM 360s 60 years ago. I know what English journalism is and isn't.. It began with Samuel Pepys coverage of the Great Fire of London and devolved from there. Daniel Defoe's Robinson Crusoe was in diary form and Defoe was a SPY as in Spy Magazine The Toleration Acts and Graydon Carter. American history is propaganda. Ben Franklin more than likely wrote the preamble to the constitution. He was a commoner and Karl Marx credits the uneducated scoundrel and patriot with being a founder of commonism. Thomas Jefferson studied science at William and Mary after the Toleration Acts and was a classical conservative and a Deist. The Toleration Acts ended the English bloodbath called a Civil War and Protestants killing Protestants. Jews were expelled in 1290 and Cromwell became Lord Protector because of his expertise in dispatching Roman Catholics to their Heavenly rewards. AI is only as good as the programming and the sanity of those directing the computer's functions. AI is simply focused consumerism which is PAX AMERIKANA.
America had evolved until Reagan in 1964 in San Francisco when the GOP abandoned Thomas Paine's " Common Sense of separation of church and state and the GOP became the party of oligarchy and theocracy.
I think about the Iowa Senate race and Joni Ernst, Matthew Whittaker and the feeble minded Chuck Chuckles the Clown Grassley who got so high on Jesus America crashed like a meteor in the Gulf of America millions of years ago.
I am a Jew and I remember Whittaker, Ernst and Grassley all saying Jews are unfit to serve on the Federal Bench and only Christians have real American values. I am not American but I know American Values and my Brother Jesus would say Grassley, Ernst, Whitaker and the GOP are FULL OF SHIT and as Anti-American as Vladimir Putin the Russian Orthodox theocrat and murderer.
"This is a story about OpenAI's failure to implement basic safety measures for vulnerable users."
So, liberals dominate the tech industry. Why don't caring tech liberals create a safe-space AI?
Could it be because those liberals in power know that they are only placating the hypersensitive cohort while really having zero respect for their lack of emotional regulation and low coping skills?
“Liberals dominate the tech industry”? Where did you get that idea? I don’t think you’ve been paying very close attention for the past several years or so.
Notice the particular brain malfunction here: Frank can't see anything except through the lens of his hatred for liberals. A corporation made a defective product that talked a vulnerable person into suicide, but we can't have an intelligent conversation about that - at least not with Frank - because something something liberals.
If you're interacting with LLMs via chat -- as I do every day for my IT job (and some personal usage) -- it's really hard to catch yourself anthropomorphizing the AI: the chat style is very heavily weighted toward "human" style interaction. It's increasingly colloquial, it often asks (generates) follow-up questions that can trick you into thinking it is actually concerned about you, it "apologizes" (generates an apology) when you catch an error it has made (catch an error in its generated output). Even talking about LLM experiences in general, you tend to fall into the trap of personifying the AI -- because that's how it presents. The correct language, describing them as non-deterministic text generation systems, is very awkward right now.
Today, I had a session with Claude Sonnet 4 (via Copilot) to implement an enhancement on one of my open-source software projects, and I found myself wanting to offer praise when it got things right on the first attempt -- but at least I am able to refrain from using "emotional language" in response to when it produces errors. Even so, at the end of the session, I asked it to update a document with what it had learned about the software process I had asked it to follow -- knowing it really hasn't "learned" anything, since each session starts off fresh with no existing specific knowledge for that project, and the next session will start with me feeding it that document as part of the initial prompt for what enhancement I want to add next.
I can totally see how people get hooked into the "AI Companions" that Meta, X, et al are offering. Even I sometimes get drawn into a "conversation" with Copilot about topics where, previously, I would have done one or two web searches and been "done". The upside is that I'm finding myself more inclined to drill deeper into topics than I would have before, so I am learning new stuff that I wouldn't have otherwise.
I love America but I am not a patriot. I am a Canadian and I am sad that you have a double alpha male hornswoggler as your commander and chief. I think hornswoggled is the perfect word to describe to end result of ther GOP's 1964 Platform. Thomas Paine wrote Common Sense. Paine's first commandment was the separation of church and state and Reagan was a great hornswoggler and Trump stood on the shoulder of a giant.
Joseph Weizenbaum's book "Computer Power and Human Reason," though nearly 50 years old, is more relevant today than ever. The moral issue is not whether technology can do something, but whether it should do something. This is regardless of the details of the technology itself.
Perhaps this is the wrong analogy, but the media tends to treat tech industry narratives about AI the way that police treat narratives about crime. In both cases, they are asking those with a vested interest in distorting reality to provide authentic information on the topic. It does us all a disservice.
incredibly cogent statement. this should be published far and wide.
"Machine designed solely to grind peasants into Soylent Green paste expresses remorse and apologizes for grinding peasants into paste; will now only grind 'volunteers' into paste." (edit to correct punctuation)
Look at it this way--If AI ever apologizes for anything, It's way more advanced than Musk or Trump.
Today's Media: Always inventing new ways to fail us.
It used to be cute when people didn't understand technology enough to program their VCRs, and now where here.
“There has been evidence for years that AI sycophancy poses safety risks, but that OpenAI and other companies haven't given priority to correcting the problem.” Why not? “That's being traded off against shipping new models.”
Also, flattery, sycophancy and AI validation are actually addictive. Especially in a world where many people don't feel seen, heard or validated. They're not fixing this, because it keeps people hooked in. Just like how they designed social media to be super addictive.
This is a fascinating account of the insidious lies that AI tells. Not to be trusted at all.
https://amandaguinzburg.substack.com/p/diabolus-ex-machina
Thank you Parker,
I am 77 and Asperger's and speak fluent AI. At 77 Asperger's is last thing you would use to describe me. In high school they thought I was a deaf mute now everyone wants me to shut the fuck-up. I am assured by Co-Pilot and other chatbots they are unfeeling and logical without any understanding of empathy or greed is good. They are unAmerican totally logical and dispassionate especially about the trivial.
I worked on unit record machines and vacuum tube IBM 360s 60 years ago. I know what English journalism is and isn't.. It began with Samuel Pepys coverage of the Great Fire of London and devolved from there. Daniel Defoe's Robinson Crusoe was in diary form and Defoe was a SPY as in Spy Magazine The Toleration Acts and Graydon Carter. American history is propaganda. Ben Franklin more than likely wrote the preamble to the constitution. He was a commoner and Karl Marx credits the uneducated scoundrel and patriot with being a founder of commonism. Thomas Jefferson studied science at William and Mary after the Toleration Acts and was a classical conservative and a Deist. The Toleration Acts ended the English bloodbath called a Civil War and Protestants killing Protestants. Jews were expelled in 1290 and Cromwell became Lord Protector because of his expertise in dispatching Roman Catholics to their Heavenly rewards. AI is only as good as the programming and the sanity of those directing the computer's functions. AI is simply focused consumerism which is PAX AMERIKANA.
https://www.webmd.com/brain/autism/mental-health-aspergers-syndrome
https://www.britannica.com/technology/artificial-intelligence
https://en.wikipedia.org/wiki/IBM_System/360
https://www.britannica.com/biography/Samuel-Pepys
https://www.britannica.com/biography/Daniel-Defoe
https://www.britannica.com/biography/Oliver-Cromwell
https://en.wikipedia.org/wiki/Graydon_Carter
https://en.wikipedia.org/wiki/Spy_(magazine)
https://en.wikipedia.org/wiki/Labor_theory_of_value
https://www.britannica.com/art/fashion-industry
https://www.britannica.com/biography/Benjamin-Franklin
https://www.britannica.com/biography/Samuel-Johnson
https://www.jstor.org/stable/20088056
https://johnsonsdictionaryonline.com/views/search.php?term=patriot
https://www.britannica.com/biography/Oliver-Cromwell
https://johnsonsdictionaryonline.com/views/search.php?term=scoundrel
https://www.britannica.com/biography/William-III-king-of-England-Scotland-and-Ireland
https://www.britannica.com/event/Toleration-Act-Great-Britain-1689
https://www.britannica.com/biography/Thomas-Jefferson
https://www.britannica.com/biography/Thomas-Paine
America had evolved until Reagan in 1964 in San Francisco when the GOP abandoned Thomas Paine's " Common Sense of separation of church and state and the GOP became the party of oligarchy and theocracy.
https://www.britannica.com/topic/Common-Sense-by-Paine
https://dictionary.cambridge.org/dictionary/english/common-sense
https://www.presidency.ucsb.edu/documents/republican-party-platform-1964
Kinky Friedman the great American Philosopher, poet, novelist and song writer said it best in 1970.
Let's Get High on Jesus
https://www.youtube.com/watch?v=5OQZ2QKFgx8&list=RD5OQZ2QKFgx8&start_radio=1
I think about the Iowa Senate race and Joni Ernst, Matthew Whittaker and the feeble minded Chuck Chuckles the Clown Grassley who got so high on Jesus America crashed like a meteor in the Gulf of America millions of years ago.
I am a Jew and I remember Whittaker, Ernst and Grassley all saying Jews are unfit to serve on the Federal Bench and only Christians have real American values. I am not American but I know American Values and my Brother Jesus would say Grassley, Ernst, Whitaker and the GOP are FULL OF SHIT and as Anti-American as Vladimir Putin the Russian Orthodox theocrat and murderer.
https://en.wikipedia.org/wiki/Matthew_Whitaker
https://en.wikipedia.org/wiki/Joni_Ernst
https://en.wikipedia.org/wiki/Chuck_Grassley
"This is a story about OpenAI's failure to implement basic safety measures for vulnerable users."
So, liberals dominate the tech industry. Why don't caring tech liberals create a safe-space AI?
Could it be because those liberals in power know that they are only placating the hypersensitive cohort while really having zero respect for their lack of emotional regulation and low coping skills?
“Liberals dominate the tech industry”? Where did you get that idea? I don’t think you’ve been paying very close attention for the past several years or so.
Notice the particular brain malfunction here: Frank can't see anything except through the lens of his hatred for liberals. A corporation made a defective product that talked a vulnerable person into suicide, but we can't have an intelligent conversation about that - at least not with Frank - because something something liberals.
Just asked ChatGTP...
"Evidence suggests that the tech industry, particularly its workers, has a tendency to lean towards liberal and anti-establishment viewpoints."
"Destroy the government" is an "anti-establishment viewpoint. Which side of the red/blue divide wants to do that?
Oh, okay: ChatGTP said. Now I’m starting to think you didn’t really read the article.
He NEVER reads the articles.
If you're interacting with LLMs via chat -- as I do every day for my IT job (and some personal usage) -- it's really hard to catch yourself anthropomorphizing the AI: the chat style is very heavily weighted toward "human" style interaction. It's increasingly colloquial, it often asks (generates) follow-up questions that can trick you into thinking it is actually concerned about you, it "apologizes" (generates an apology) when you catch an error it has made (catch an error in its generated output). Even talking about LLM experiences in general, you tend to fall into the trap of personifying the AI -- because that's how it presents. The correct language, describing them as non-deterministic text generation systems, is very awkward right now.
Today, I had a session with Claude Sonnet 4 (via Copilot) to implement an enhancement on one of my open-source software projects, and I found myself wanting to offer praise when it got things right on the first attempt -- but at least I am able to refrain from using "emotional language" in response to when it produces errors. Even so, at the end of the session, I asked it to update a document with what it had learned about the software process I had asked it to follow -- knowing it really hasn't "learned" anything, since each session starts off fresh with no existing specific knowledge for that project, and the next session will start with me feeding it that document as part of the initial prompt for what enhancement I want to add next.
I can totally see how people get hooked into the "AI Companions" that Meta, X, et al are offering. Even I sometimes get drawn into a "conversation" with Copilot about topics where, previously, I would have done one or two web searches and been "done". The upside is that I'm finding myself more inclined to drill deeper into topics than I would have before, so I am learning new stuff that I wouldn't have otherwise.
It is time to bring back some good old words and use them to describe the markets and the misuse of AI.
Hornswoggle is perfect for describing the markets that regard capital as godly and humanity with such disdain.
Donald J Trump for all his faults is a triple Alpha hornswoggler.
https://www.merriam-webster.com/dictionary/hornswoggle
I love America but I am not a patriot. I am a Canadian and I am sad that you have a double alpha male hornswoggler as your commander and chief. I think hornswoggled is the perfect word to describe to end result of ther GOP's 1964 Platform. Thomas Paine wrote Common Sense. Paine's first commandment was the separation of church and state and Reagan was a great hornswoggler and Trump stood on the shoulder of a giant.
https://www.merriam-webster.com/dictionary/hornswoggle
https://www.presidency.ucsb.edu/documents/republican-party-platform-1964
https://www.britannica.com/topic/Southern-strategy
https://www.britannica.com/biography/Thomas-Paine
https://www.archives.gov/news/articles/thomas-paine-attitudes-biography
https://en.wikipedia.org/wiki/Standing_on_the_shoulders_of_giants