Physicist, electrical engineer and critic of postmodernism Alan Sokal caused the Sokal Affair in 1996 when he got published a grand-sounding but completely nonsensical paper titled "Transgressing the Boundaries: Toward a Transformative Hermeneutics of Quantum Gravity" which "flattered the editors' ideological preconceptions". He later responded to leftist and postmodernist criticism of the deception by asserting that he was himself a leftist, and that his motivation was to "defend the Left from a trendy segment of itself". — Source: Wikipedia.

Raghavan and Murthy turned up on my doorstep together one evening, carrying a bottle of scotch. “What’s up?” I asked as I led them into the drawing room.

“Ask him,” replied Raghavan, pointing at Murthy. “He’s got a new client.”

Murthy’s cellphone dinged with an incoming message. He looked at it and said, “Ah! My master’s been experimenting with AI!”

“What do you mean?” Raghavan asked.

“Politicians now use AI,” he said. “They think AI can replace party spokespersons. Someone I know who’s been accused of corruption put out a statement in response, but it’s AI slop.” He handed the phone over to Raghavan. “Every sentence in here is ok, but when you put it all together it means nothing at all.”

Raghavan looked at it for a moment and passed it on to me before replying, “It’s not fair to call this AI slop,” he said. “It’s human slop.”

“What’s slop?” I asked.

“Stuff that looks as if it means something but doesn’t,” said Raghavan.

The statement said something about the allegations being baseless and wondering who benefitted from them, and something about the people deserving good governance and not fake controversies. There were lots of long words but it was all hot air. “Why call it AI slop?” I asked.

“An AI agent generated it,” said Murthy. “That makes it AI slop.”

“No,” said Raghavan. “It’s slop generated by an AI agent obeying a human too lazy to write it out themselves. So it’s human slop. Besides, it’s humans who wrote the programs that run the AI system, so, I repeat, it’s human slop.”

“At the end of the day, it’s the AI system that produced the slop, so it’s AI slop,” said Murthy with the air of someone clinching an argument.

“If you use an earthmover instead of a spade to dig a hole in the wrong place,” said Raghavan, “who's responsible? The earthmover or the operator?”

“The operator,” replied Murthy.

“Well, there you are, then,” said Raghavan.

“You can’t compare them!” replied Murthy. “AI stands for Artificial Intelligence. It’s supposed to think for itself, unlike an earthmover.”

Prita joined us then, and after a samosa break, Raghavan continued. “It only appears to think. That’s why you get a little warning each time you use AI, telling you that it can make mistakes. The bigger point is that slop has been around as long as humanity. AI made it plentiful, that’s all… There’s a lot of good work being done by people using AI. Of eight of last year’s Pulitzer Prize winners, five used AI to help them with drafting. But there’s lots more slop coming from people who don’t think as hard as Pulitzer winners, so it’s human slop, not AI slop.”

Murthy was unconvinced. “Without AI agents the guy who produced this statement wouldn’t have been able to produce it. So it is AI slop. And AI at least generates higher quality slop than humans do.”

Raghavan burst into laughter. “Nonsense!” he said. “The best ever piece of slop came not from AI but from a human. Whoever gets AI to generate slop of that quality has to be a genius. What to hear about it?”

“Sure,” said Prita and Murthy and I together.

“Back in 1996 an American physicist called Alan Sokal wrote a paper on sociology and sent it off to an academic journal on sociology called Social Text. The title of the paper? ‘Transgressing the Boundaries: Toward a Transformative Hermeneutics of Quantum Gravity.’

“The editors of this journal went through the paper, found it okay, and placed it in the next issue. After it was published, Sokal wrote an article in another journal called Lingua Franca saying that the Social Text paper was a test to see if a cultural studies journal would ‘publish an article liberally salted with nonsense’.”

“What was the point of all that?” asked Murthy.

“There was criticism of science in cultural studies saying that there’s no objective reality, that scientific facts don’t exist,” replied Raghavan. “Sokal had taken pains putting together a paper that used the right buzzwords and was complex enough to hide that it meant nothing. To show that cultural studies were off the mark.”

“That’s nonsense about science!” Prita said. “The next time those guys take a flight somebody should tell them that the aircraft isn’t real and might crash any time!”

“Exactly,” said Raghavan. “You might question the directions science takes, but to deny that science has meaning…”

I couldn’t understand much of what was going on but there was an edge of venom in the air, so I asked, “What happened after?”

“The editors of Social Text said they knew it was a poor paper but published it anyway because they thought it was a ‘symptomatic document’, whatever that is, that they’d been deceived because they weren’t geared to catch malicious deception. But Social Text had gotten caught because their slop check wasn’t good enough to catch Sokal.”

“So what?” asked Murthy.

“So nothing,” replied Raghavan. “Public memory is short. After the initial ruckus, you could say that slop won.”

“That’s sad,” said Prita. “Just think of the power and water and computing capacity being wasted on AI now!”

“Exactly,” said Raghavan.

“Aren’t there AI slop detectors?” asked Prita.

“Sure,” replied Raghavan. “Publishers use them.”

“You mean there are people making slop with AI and others catching slop with AI?” asked Prita. “Double the waste! What a shame!”

“Yes,” said Raghavan.

“Why don’t we have AI systems that figure out that they’re being used to generate slop and stop it before it’s generated?” asked Murthy.

“Maybe people can’t live without slop!” said Prita. “And God help humanity if that’s true!”

Tags: