Illustration by Tag Hartman-Simkins/Futurism. Source: Getty Images
There’s dirty science, and there’s AI dirty science.
In an ironic twist of fate, embattled AI researchers are warning that the field is being choked by a flood of substandard academic papers written with large language models, making it harder than ever to discover and surface high-quality work.
Part of the problem is that AI research has grown in popularity. The more people jumping on the bandwagon, the more people are trying to rapidly boost academic reputation by turning out dozens – and sometimes hundreds – of papers a year, thereby discrediting the entire endeavor.
in one interview with GuardianHany Farid, a computer science professor at UC Berkeley, called the situation “hysterical.” He says that after so much negligence at the top, he now advises his students not to enter this field.
“A lot of young people want to go into AI,” Farid said. Guardian“It’s just a mess, You can’t persist, you can’t publish, you can’t do good work, you can’t be thoughtful,”
Farid sparked debate on the topic by calling out the output of an AI researcher named Kevin Zhu, who claims to have published 113 papers on AI this year.
“I cannot read 100 technical papers carefully in a year,” Farid wrote in an article. linkedin post Last month, “So imagine my surprise when I learned of an author who claims to participate in the research and writing of over 100 technical papers a year.”
Zhu, who recently earned a graduate degree in computer science at UC Berkeley — the same place where Farid teaches — launched an AI researcher program called Algoverse for high schoolers and college students. Many of its participants are co-authors of Zhu’s papers, Guardian noted. Each student pays $3,325 for a 12-week online course, during which they are expected to present work at AI conferences.
One of those conferences is NeuReps, considered one of the big three conferences in a field that was once obscure but is now the center of attention as AI gains massive investment and societal traction. According to this, it filed less than 10,000 papers in 2020 GuardianThis year, that number has increased to more than 21,500, a trend shared by other major AI conferences, The explosion has been so severe that Neurips is now relying on PhD students to help review its flood of submissions,
This huge amount is thanks to people like Zhu: 89 of his papers spanning more than a century are being presented in NewReps this week.
Farid called Zhu’s documents a “disaster” and said he “could not possibly make a meaningful contribution” to them.
“I’m fairly convinced that the whole thing from top to bottom is just vibe coding,” Farid said, using new slang that has emerged to describe using AI tools to build software faster.
When asked, Zhu did not confirm or deny whether his papers were written with AI. GuardianBut he said his teams “use standard productivity tools like reference managers, spell check, and sometimes language models for copy-editing or improving clarity.”
The role that AI has rapidly developed in academic research has been a subject of controversy since its popularity first rose several years ago. Tools like ChatGPT still have a tendency to produce misleading citations, or inventing sources that do not exist, which often slip through even the peer review process of reputable journals. Other examples, such as when a peer-reviewed paper used an AI-generated diagram of a mouse with impossibly super-sized genitals, make you question whether there were any omissions. The technology has become so ingrained in academia that some clever authors are inserting hidden text into their papers designed to trick “reviewers,” who are themselves AI-powered, into positively evaluating their work.
What’s particularly disappointing to hear now, however, is how AI research has begun to be sabotaged by the technology itself. How long can Search keep its own product alive? And what does it mean for the coming generation of AI scientists, if novel research is being drowned out by their far more prolific peers who are studying with fabricated sources?
Even a seasoned vet like Farid says it has now become impossible to keep track of what is happening in the AI field.
“As an average reader you have no chance of trying to understand what’s going on in the scientific literature,” Farid explained. Guardian“Your signal-to-noise ratio is basically one, I can barely go to these conferences and figure out what the hell is going on,”
More on AI: AI researchers say they have invented spells that are too dangerous to release to the public
