meta paused artificial intelligence (AI) bot last week, just two days after it was published. The bot, called Galactica, was “trained on his 106 billion tokens of open-access scientific text and data, including papers, textbooks, scientific websites, encyclopedias, reference materials, knowledge bases, and more.” will be available,” the company said. Daily Beast.
It was supposed to help scholars and researchers find papers and studies quickly and concisely, but was overwhelmed by the sheer amount of misinformation that miscited reputable scientists.
Karl Bergstrom, a professor of biology at the University of Washington, said a scientist’s reputation could be at stake if they were misquoted in a text. CNET The problem with Galactica is that it was touted as a way to get facts and information. But instead, he said it acted like “a flashy version of the game that starts with half a sentence and autocomplete fills in the rest of the story.”
Within hours of Galactica being published, users began reporting racist and inaccurate articles. post an image In response to requests regarding linguistic bias. The bot’s response falsely claimed that black people “do not have their own language” and that immigrants “do not speak a language different from the language of the country to which they are immigrating.”
Other information generated included bogus research on the benefits of eating crushed glass and bogus information that researchers at Stanford University were creating “gaydar” software. find gays on facebook.
A Meta AI spokesperson told CNET: [machine learning] A system for learning and summarizing information. He added that Galactica is “essentially short-term exploratory research without a product plan.” Yann LeCun, Meta AI’s chief scientist, told the outlet that the bot was removed because the team was “so distraught by the violence on Twitter.”
Two days after Galactica’s launch, the Meta AI team suspended the bot and Meta’s chief AI scientist Yann LeCun tweeted“The Galactica demo is currently offline. Casually exploiting and enjoying it is no longer possible. Happy?”
Meta AI (previously called Facebook Artificial Intelligence Research) developed Galactica as a method. “Organizing science” and condensing it The overwhelming amount of scientific information that can be found online.The idea was to have an AI Solve math problems, write scientific code, and create research summaries.
meta Note on Galactica website AI models have limitations that can cause “hallucinations”.This site advises users to check the information that pops up and follows “There is no guarantee of truthful or reliable output from language models, even for large-scale models based on high-quality data like Galactica,” he added, adding that the generated text was “very reliable and highly reliable.” It may appear to be “trusted to”, but it could still be wrong, he added.
Vincent Conitzer, professor of computer science at Carnegie Mellon University in Pittsburgh, told The Daily Beast.
“My impression is that Meta put a lot more effort into this public release, doing more serious user research first, drawing attention to the desired usage of the system, and being open and forthright about the undesired usage. would have done better.”