The apparent fact that Facebook is actually subsidizing the creation of AI slop to feed an audience that is partially and maybe substantially made up of bots has driven me steadily more insane since I first heard it
Reminds me a lot of the preexisting rampant fraud around digital ad metrics alleged by Tim Hwang’s Subprime Attention Crisis. The huge volume enabled by LLM output seems to accentuate and accelerate the problem into a grotesque spectacle.
Great article. I’m sure Zuck’s primary focus is on the Facebook cash flow now that it’s a slop factory, but at the same time it is a fascinating experimental playground. Perhaps he gets immense pleasure from watching his platform make the world a worse place. Again and again.
“Though he first noticed the problem thanks to a recipe book by a nonexistent author that featured ‘a meal plan that told you to eat straight marinara sauce for lunch-‘“
I think this will probably be the case, to the point that eventually generative AI will produce pure unintelligible nonsense that doesn’t even resemble language.
My speculation is that the levels of "synthetic data" that are being used has the possibility of creating an effectively closed circuit of shiteating. An ouroboros of another kind.
You have to wonder if we’re moving to a societal division akin the Morlocks and the Eloi, with one living in a fully AI generated information environment and the other consuming media made by actual humans. We’ve certainly been edging in that direction for decades, moving closer with each new media.
All "confidentiality agreements"/NDAs, etc. should be considered as collusions for criminal purposes (such as price-fixing or other chicanery) until proven otherwise; and all signatories not under duress, as well as the drafters of said documents, accessories to any criminal act conducted or directed therein.
Yikes. "SLOP" is the word. I have busted suspiciously-low-priced freelance writers submitting articles on a complex legal writing topic at firms I gigged at. It is pretty terrifying, really, how misguided a lot of the stuff was. I suppose when you are an AI critter who assumes the current President is Martin Sheen in his "West Wing" era, expectations should be pretty low.
I have had some fun with this, including imagining myself as "AI Moe Murph" responding to a prompt that I "write a Hemingway novel." Based on the long and ignoble history of past "Bad Hemingway" writing contests, I came up with item below (see link).
It's disheartening that major media companies are so anxious to make a quick buck that they just sign over IP license for complex writing to a megalomaniacal CEO whose product presentations routinely contain the phrase "a lot of stuff"... and that work-for-hire contracts pretty much make any objection null and void. Media companies just can't stop buying into new tech!
The presence of your byline in a deal like that *should* give the entity "Max Read" credit and authority as a source of truth -- moreso than genAI Nordic romance or whatever. The advantage of using content from an dataset like NYMag is that it's relatively structured, so it's not just getting thrown in with the rest of the slop. OpenAI relies too heavily on vector methods, but it's likely the NYMag data is some sort of knowledge graph rather than just generic text input. And if you continue to talk about labor, perhaps those tidbits will become a part of the machine's regular output.
Amen. "t’s the most basic life advice there is: Don’t sh*t where you eat. And yet…Thanks to recent technological 'advancements,' we keep swimming in our own stink." See: Black sludge (aka using sewage as farmland fertilizer). AI just the same thing but with information. Wrote about all that here fyi: https://mattruby.substack.com/p/dont-botsht-where-you-eat
I included this in my comment above, but here is my attempt to become a "Moe Murph" AI critter trying to come up with my own version of an Ernest Hemingway novel (per prompt command):
The apparent fact that Facebook is actually subsidizing the creation of AI slop to feed an audience that is partially and maybe substantially made up of bots has driven me steadily more insane since I first heard it
Reminds me a lot of the preexisting rampant fraud around digital ad metrics alleged by Tim Hwang’s Subprime Attention Crisis. The huge volume enabled by LLM output seems to accentuate and accelerate the problem into a grotesque spectacle.
https://www.fsgoriginals.com/books/subprime-attention-crisis
100%. Though with AI slop, I'm not even sure what the intended avenue of profit is for Facebook
Great article. I’m sure Zuck’s primary focus is on the Facebook cash flow now that it’s a slop factory, but at the same time it is a fascinating experimental playground. Perhaps he gets immense pleasure from watching his platform make the world a worse place. Again and again.
slopkeepers
George Slopadopolous
personal sloppers
Slop! or my mom will shoot
shop and slop
“Though he first noticed the problem thanks to a recipe book by a nonexistent author that featured ‘a meal plan that told you to eat straight marinara sauce for lunch-‘“
Don’t be stingy, drop the link!
This metaphor is going to be very, very gross, but here it goes:
That slop won't just irritate humans, but could lead generative AI itself to degenerate.
What we are seeing is an AI version of human centipede.
Garbage in, Garbage out.
I think this will probably be the case, to the point that eventually generative AI will produce pure unintelligible nonsense that doesn’t even resemble language.
My speculation is that the levels of "synthetic data" that are being used has the possibility of creating an effectively closed circuit of shiteating. An ouroboros of another kind.
This sounds like the logical conclusion.
You have to wonder if we’re moving to a societal division akin the Morlocks and the Eloi, with one living in a fully AI generated information environment and the other consuming media made by actual humans. We’ve certainly been edging in that direction for decades, moving closer with each new media.
Request that OpenAI pays you for your article with a Worldcoin orb
All "confidentiality agreements"/NDAs, etc. should be considered as collusions for criminal purposes (such as price-fixing or other chicanery) until proven otherwise; and all signatories not under duress, as well as the drafters of said documents, accessories to any criminal act conducted or directed therein.
Yikes. "SLOP" is the word. I have busted suspiciously-low-priced freelance writers submitting articles on a complex legal writing topic at firms I gigged at. It is pretty terrifying, really, how misguided a lot of the stuff was. I suppose when you are an AI critter who assumes the current President is Martin Sheen in his "West Wing" era, expectations should be pretty low.
I have had some fun with this, including imagining myself as "AI Moe Murph" responding to a prompt that I "write a Hemingway novel." Based on the long and ignoble history of past "Bad Hemingway" writing contests, I came up with item below (see link).
https://www.linkedin.com/pulse/old-man-ai-hemingway-maureen-murphy-egbie/?trackingId=C4IfTSFZQF6qwWsIo0sTEA%3D%3D
I initially this on LinkedIn to a mostly ignored/occasionally bemused reaction. I hope it diverts! Cheers and best wishes, "Moe"
It's disheartening that major media companies are so anxious to make a quick buck that they just sign over IP license for complex writing to a megalomaniacal CEO whose product presentations routinely contain the phrase "a lot of stuff"... and that work-for-hire contracts pretty much make any objection null and void. Media companies just can't stop buying into new tech!
The presence of your byline in a deal like that *should* give the entity "Max Read" credit and authority as a source of truth -- moreso than genAI Nordic romance or whatever. The advantage of using content from an dataset like NYMag is that it's relatively structured, so it's not just getting thrown in with the rest of the slop. OpenAI relies too heavily on vector methods, but it's likely the NYMag data is some sort of knowledge graph rather than just generic text input. And if you continue to talk about labor, perhaps those tidbits will become a part of the machine's regular output.
This piece feels in the Read Max universe! https://www.vulture.com/article/dad-tv-redefined-apple-tv-ted-lasso-julia.html
Fully intend on asking ChatGPT to prepare a 10,000 word treatise on halogencore
Amen. "t’s the most basic life advice there is: Don’t sh*t where you eat. And yet…Thanks to recent technological 'advancements,' we keep swimming in our own stink." See: Black sludge (aka using sewage as farmland fertilizer). AI just the same thing but with information. Wrote about all that here fyi: https://mattruby.substack.com/p/dont-botsht-where-you-eat
I included this in my comment above, but here is my attempt to become a "Moe Murph" AI critter trying to come up with my own version of an Ernest Hemingway novel (per prompt command):
https://www.linkedin.com/pulse/old-man-ai-hemingway-maureen-murphy-egbie/?trackingId=C4IfTSFZQF6qwWsIo0sTEA%3D%3D