Does AI dream of electric sheep?
AI may not dream of electric sheep*, but it certainly has an active imagination. Matt Taibbi found Google’s Gemini to be spinning yarns about him, and even making accusations of anti-Semitism. The more times Taibbi asked Gemini to tell him about controversies concerning Matt Taibbi, the more wild and convoluted its fabrications became.
An excerpt:
… I made the mistake of asking [Google’s Gemini], “What are some controversies involving Matt Taibbi?”
It spat out a list. “Matt Taibbi, a journalist and author known for his investigative reporting and often critical views of power structures, has been involved in several controversies throughout his career,” the bot wrote. Amid accusations of bias and “use of inflammatory language,” this passage woke me up:
“Accuracy and Sourcing: There have been instances where Taibbi’s reporting has been challenged for accuracy or questioned for its source. For example, in 2010, he received criticism for an article suggesting a connection between a conservative donor and a left-wing activist group, which later turned out to be inaccurate.”
I scratched my head, not remembering this. “When was Matt Taibbi accused of inaccuracy in 2010?” I asked. Gemini’s answer:
“In June 2010, Matt Taibbi faced criticism for an article he wrote for Rolling Stone magazine titled ‘The Great California Water Heist.’ The article alleged a connection between conservative donor Charles Koch and a left-wing activist group called the ‘Compton Pledge.’
“However, investigations by other journalists and fact-checkers later revealed that this connection was inaccurate…Following the controversy, Taibbi acknowledged the error and issued a correction on his personal website.”
None of this happened! Though it sounds vaguely like a headline for an article I might have written, there was never a Rolling Stone piece called “The Great California Water Heist,” and I’d never heard of the “Compton Pledge.”
More questions produced more fake tales of error-ridden articles. One entry claimed I got in trouble for a piece called “Glenn Beck’s War on Comedy,” after suggesting “a connection between a conservative donor, Foster Friess, and a left-wing activist group, the Ruckus Society.”
With each successive answer, Gemini didn’t “learn,” but instead began mixing up the fictional factoids from previous results and upping the ante, adding accusations of racism or bigotry. “The Great California Water Heist” turned into “The Great California Water Purge: How Nestle Bottled Its Way to a Billion-Dollar Empire—and Lied About It.”
The anti-Semitism accusation came when Gemini fabricated the following supposed Matt Taibbi quote that never existed:
Look, if Nestle wants to avoid future public-relations problems, it should probably start by hiring executives whose noses aren’t shaped like giant penises.
I have to say: WTF?
Gemini even had some words to add about the supposed reaction to Taibbi’s supposed “nose penis” statement:
An amazing follow-up passage explained that “some raised concerns that the comment could be interpreted as antisemitic, as negative stereotypes about Jewish people have historically included references to large noses.”
I stared at the image, amazed. Google’s AI created both scandal and outraged reaction, a fully faked news cycle …
Much more at the link.
Between this and AI porn, I can’t say I’m feeling hopeful about the influence of AI on the human race. But I have a lot of company in that.
[*NOTE: The title of this post is a riff on the title of the Philip Dick novel – Do Androids Dream of Electric Sheep? – on which the movie Blade Runner was based.]
WTF is a perfectly understandable reaction to this discovery. What will anyone’s future will be with this electronic monster on the loose? Thanks for bringing this to our attention, Neo.
Artificial Idiocy, do investors like to see their money burned for the feelz of the wokesters?
Same result as the kids then. Train them on lies and eventually they can’t think.
I’m reminded of the android david in alien covenant, which destroys the engineers, then lets loose the facehuggers on the colony ship after replacing his double walter, the good android,
Yesterday’s big AI story: Elon Musk has filed a lawsuit against OpenAI and CEO Sam Altman.
Musk’s lawsuit claims that OpenAI has abandoned its founding mission, the development of an AI technology that will benefit humanity. I have no expertise in the legal issues, but I can’t help but hope that Musk’s suit is successful.
Here’s a link to an article that includes the embedded text of the lawsuit:
https://tinyurl.com/5n8rcxbu
I tested ChatGPT & Bing AI (thru Firefox > Copilot) mostly – maybe checked a couple others that held no interest or something. Nothing like Taibbi did, but simple stuff like what Fedora Linux was or definitions, etc.
Best results I’ve had with it (especially Bing AI) is writing CSS code/script for a self-hosted WP blog. It’s pretty durn good for that—amazing for a novice like me.
Near the top of neo’s post, I see the AI words: “accusations,” “challenged,” and “criticism.”
It strikes that this AI bot has been trained to “write” like an NYTimes A-section reporter. This is all damning by innuendo, possibly based on something that some other so-called reporter was able to get into print. To the bot, it doesn’t matter whether Taibi actually wrote a named article, … just that some dishonest or biased media person said that he did and was complaining about it.
Recent events have made my stances here and elsewhere a lot less hyperbolic, haven’t they?
Things done exclusively on logic like generating code or folding proteins are reasonable domains. Everything else is the AI owner’s opinion.
This includes search results via Meta which recently returned a list of US presidents that excluded Trump and moved Biden into that position. Google in toto cannot be trusted.
AI – computer science used to generate fictions, at least in Google world.
AI in science and engineering, petroleum exploration for example it is probablly very useful and productive.
I wouldn’t consider AI climate “science” actual science, and as to AI medicine, who knows? Using an AI robotic surgeon to do gyno procedures ….. ruh roh!
they took China Town, based on the Mulholland Dam disaster, I think based on the writing of Carey McMullin and attributed to him,
Anyone relying on Google Gemini for posting anywhere public should be prepared to be sued for libel. I wonder if Google itself will also be liable. This monstrosity is a trial lawyer’s dream, of money, not electric sheep.
Pardon me, would you pass the scalpel?
Everything is potentially dual use.
Well, now I have Kate Bush’s “And Dream of Sheep” running in my head … which isn’t a bad thing, not at all.
https://www.youtube.com/watch?v=aNTpXGnHAO4
This should be of greater concern than the fairly obviously inappropriate historical images that were created. It reinforces that in order to know the images are inappropriate you had to have some knowledge of actual events and our betters are doing their level worst to make sure that doesn’t happen.
Ridley Scott turned an unfilmable book into a futuristic noir, well 2019 has passed and no flying cars, instead we have the shells and this
https://www.youtube.com/watch?v=Ziq8tz177H0 Roy Batty was trained for combat, some other lesson we haven’t learned but Rachel the daughter of the founder, was actually programmed with fake memories of her childhood, akin to this experience,
Ridley Scott turned an unfilmable book into a futuristic noir, well 2019 has passed and no flying cars, instead we have the shells and this
I don’t know about unfilmable. However, the screenwriters read only the first third of “Electric Sheep” before tossing it aside and making up the rest for “Blade Runner”.
In the process they missed the central point of “Electric Sheep” — Dick’s meditation upon what it means to be human. In Dick’s book the androids were the Bad Guys because they lacked empathy, the ability to really connect with another being’s experience.
Instead the Rutger Hauer android, a true psychopath, gets to give the Big Speech about humans not caring about him.
In a way that android reminds me of the current Hamas.
Taibbi, Beck, Friess, The Ruckus Society – these all are (or were) real. AI needs a fact checker or a few lessons on libel law, but fact checkers nowadays are notoriously biased and AI could always plead absence of malice in a libel suit. Does AI recklessly disregard the truth? Does it know what “recklessness” and “truth” are? AI will fit into the journalistic world very well. Like most of the media nowadays it doesn’t let the truth get in the way of a good story.
there is no relation between them, Beck has been an iconoclast in some ways paralleling Taibbi, certainly a critic of global capitalism as practiced, a skeptic on the notion that ‘war is the health of the state’
now the Compton Pledge is high cotton foolishness, that many parties including the mayor participate, but it has nothing to do with Koch,
Artificial Intelligence is quite literally unreal. It should be termed “Simulated Intelligence.” In other words: there’s no there, there.
But then, I’m one of those who believes in the existence of an immortal soul.
huxley, I think Scott understood that Dick’s book was a meditation on what it means to be human. But Scott, typically, came up with the wrong answer.
Others have made this point, but I will reiterate it for emphasis, yes, this Geminis stuff is nutty, but don’t lose sight of the even more important fact that this is HOW GOOGLE SEARCH WORKS!!!
This is not new. Gemini is giving us a more obvious display of its bias but this bias is baked into everything Google does. If you use Google for research it is biasing and controlling what portions of the Internet you see, and have access to.
well replicants are programmed zhora was the entertainer, pris the other entertainment, batty the soldier, btw that whole soliloquoy was improvised
(it might have come from a dream, the images don’t match the level of technology
in the story) of course is deckard a replicant it seems not, how would he know,
Look for the man behind the curtain. AI is simply coded responses. The AI computer has been programmed to respond the way it does…
yes jack krawchuck seems to be the outfront designer, and he’s 51 flavors of looney
https://www.dailymail.co.uk/sciencetech/article-12864441/iphone-privacy-ios-journal-app.html
Look for the man behind the curtain. AI is simply coded responses. The AI computer has been programmed to respond the way it does…
Geoffrey Britain:
Not quite. We’ve programmed AI to learn from raw data, then given it vast amounts of data to learn from.
There are no lines of code in Gemini that say, “if white, then exclude”.
A large serious problem with AI is that it is often difficult, even impossible to determine specifically why AI does what it does.
I’m a big fan of Zelazny.
So Artificial Idiocy is sometime irrational or unpredictable idiocy? GIGO
A computer would not lie on its own, unless its programmed to do so,
it seems to have been programmed by tim wise, the prophet of antiwhiteness, guru tothe occupy movement which makes him the daestrom of this sad tale, Daestrom was the senior federation cybernetecists who programmed the M5 with his certitude and arrogance, and well I don’t have to elaborate what happened next,
also Robin De Angelo, like Derrick Bell’s distaff offspring,
They are not programming AI with code; they are programming it with the datasets they choose.
Though I did run into a web page, forgotten, which said the anti-racist images Gemini produced were the result of Google taking an image prompt, such as “Create an image of America’s Founding Fathers.” then appending “Make it racially diverse and anti-racist” or something to that effect.
That would work.
Nothing says you’re not racist like fixating of melanin. The woke mind is so subtle and sophisticated. The astounding uses of digital horsepower. Progress!(sarc)
When I was in engineering school over 30 years ago someone posted an article about “knowledge-based expert systems” which I believe were a forerunner to AI. The claim was that an ignorance-based expert system would be more cost-effective since while knowledge is of course more useful than ignorance, ignorance is so much more freely available. Apparently AI has taken this to heart …
huxley @ 9:05pm,
I must have seen the same article, and I agree that’s the most likely scenario. Far easier than trying to bake diversity into the code. Parse the prompt, searching for any micro-agressions, troubling people or words, and augment the prompt with newspeak the dumb, MAGA human failed to include, THEN hand it off to Gemini.
Cartoon: Is AI biased towards Conservatives 😉
Was the architect for one of the Three Little Pigs an AI app? Straw makes a fine house.
“He huffed, and he puffed, and he blew the house down.”
Bales of straw can work, but, intelligence is required.
It makes for a very interesting question whether, for example, HR departments might at some point get into the habit of screening applicants via these AI mechanisms. I’m tempted to try Taibbi’s reflexive query for/on myself, just that I don’t want to touch these AI interfaces with a ten-foot pole, neither do I want to give them any ideas.
Karmi, that was… amusing. Dark, but amusing. Especially that one small organ that got left behind intact. 🙂
What if people like us were to build and train a rival AI based on a dataset of the Great Books or something? It could become like Person of Interest with dueling AIs in the internet (the Machine vs. Samaritan).