High-profile examples of made-up sources, including from the Make America Healthy Again Commission, highlight the perils of using generative AI without doing your own fact checks.
When we use AI, we should slow down and check every source it provides.Journalists and the general public play an important role in helping to uncover fake citations. In May, the U.S. Make America Healthy Again Commission released a report on children’s health that included a number of fake citations, including some related to medications and to the advertising of medications.
gone awry, the U.S. Make America Healthy Again Commission released a report it described as a “clear, evidence-based foundation” for progress on children’s health. Journalists, however, quickly foundwith the sources the report used to support their “evidence,” including links that did not work and summaries of studies that misrepresented the underlying work. Perhaps most troubling, there were, generative AI is far from perfect, and “hallucinations” – information that is inaccurate – can occur. And we might think AI is getting more accurate as it matures, but the more recent sophisticated models actuallythat the errors were what one would expect if the authors of the report on children’s health used generative AI . Oransky stated, “We’ve seen this particular movie before, and it’s unfortunately much more common in scientific literature than people would like or than really it should be.” AI hallucinations are, of course, not limited to science. They occur across disciplines and contexts. Indeed, such errors are a growing problem in court where legal documents increasingly include references to non-exist legal cases. For example, in less than four weeks in May,at least 23 made-up legal citations. Moreover, it appears that professionals – lawyers and trained legal staff – are largely responsible for these fake references. Lawyers may be overly relying on AI, yet not checking what AI produces. Reporters noted that the actual number of fake citations is almost certainly higher given that judges may not always catch when it happens. Why? Reporters speculated that lawyers and their staff may not “have the resources or know-how for sophisticated legal research” or may be “technologically challenged.” In fact, technology challenges abound in the professional workspace. But an overreliance on AI can have far more serious ramifications than other tech glitches. In the legal arena, judges and juries may make poor decisions based on inaccurate information, and lawyers responsible for errors have been sanctioned. In the scientific and medical arena, information based on false citations may lead to dangerous policy decisions.expert Marion Nestle told reporters,"The speed suggests that it could not have been vetted carefully and must have been whisked through standard clearance procedures. The citation problem suggests a reliance on AI." If true, this process belies theof the U.S. Health Secretary, Robert F. Kennedy, Jr., who oversees MAHA and “has repeatedly said he would bring ‘radical transparency’ and ‘gold-standard’ science to the public health agencies.” Beyond the citation problems, Kennedy has even refused to share information about the team who wrote the report. Such cases highlight the need for rigorous fact-checking on the part of the authors of any scientific report or legal brief, but also highlight the need for journalists and the general public to do their own fact-checking. When a topic is important to you, click on the links. Do they lead you to a study that actually exists? Do the conclusions of the researchers fit with what the authors actually wrote about it?itself has strict rules on the use of AI, but we authors are allowed to use it to help generate ideas, check our grammar, and seek research assistance. Instructions tell us that “LLMs are great sources of data, however, all statements derived from gen-AI need to be fact-checked with a non-AI source, e.g. peer-reviewed literature.” This is an excellent standard and one that we hope will be widely adopted. Until then, when doing your own citation check, even one fake article or misrepresented finding should make you question the overall report. It doesn’t mean that the conclusions in the report are necessarily wrong, but you can no longer be confident in the authors’ process and should maintain a healthy skepticism. Let’s all work to make scientific thinking great again.is the author of eight books, the host of an NBA podcast, and an editor of textbooks on statistics and psychology.Self Tests are all about you. Are you outgoing or introverted? Are you a narcissist? Does perfectionism hold you back? Find out the answers to these questions and more with Psychology Today.
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Here’s How Adobe Is Navigating the Maze of Generative AI and Copyright ConcernsAdobe says Firefly is driven by Adobe’s roots in the creative community and its respect for creators.
Read more »
These biotech stocks will benefit as generative AI speeds up drug discovery, Jefferies saysThe adoption of generative AI in drug discovery will save companies years and billions of dollars in getting new drugs to market, the firm said.
Read more »
15 Amazing Generative AI Tools For Lawyers And Legal TasksDiscover the top AI tools transforming the legal world for professionals and consumers alike, from contract automation to court claims.
Read more »
Traditional diagnostic decision support systems outperform generative AI for diagnosing diseaseResearchers compared their long-standing diagnostic decision support systems AI tool, DXplain, with modern large language models like ChatGPT and Gemini, finding DXplain performed slightly better. They say their findings suggest that combining DXplain with LLMs could enhance clinical diagnosis and improve both technologies.
Read more »
FDA deploys generative AI tool Elsa to transform agency operations in historic moveThe FDA launched Elsa, its first generative AI tool, to streamline reviews and modernize agency workflow across all centers.
Read more »
71pc of CMOs plan to invest more than $10M in AI through 2028: BCGConfidence in generative artificial intelligence among high-level advertising professionals is rising to record levels.
Read more »