{"id":14725,"date":"2025-07-16T21:24:08","date_gmt":"2025-07-16T21:24:08","guid":{"rendered":"https:\/\/www.newsbeep.com\/us\/14725\/"},"modified":"2025-07-16T21:24:08","modified_gmt":"2025-07-16T21:24:08","slug":"why-ai-is-getting-less-reliable","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/us\/14725\/","title":{"rendered":"Why AI is Getting Less Reliable"},"content":{"rendered":"

Last week, we conducted a test that found five leading AI models<\/a>\u2014including Elon Musk\u2019s Grok\u2014correctly debunked 20 of President Donald Trump\u2019s false claims. A few days later, Musk retrained Grok with an apparent right-wing<\/a> update, promising that users \u201cshould notice a difference.\u201d They did: Grok almost immediately began spewing out virulently antisemitic tropes praising Hitler<\/a> and celebrating political violence against fellow Americans. <\/p>\n

Musk\u2019s Grok fiasco<\/a> is a wakeup call. Already, AI models have come under scrutiny for frequent hallucinations<\/a> and biases<\/a> built into the data used to train them. We additionally have found that AI systems sometimes select the most popular\u2014but factually incorrect\u2014answers, rather than the correct answers. This means that verifiable facts can be obscured by mountains of erroneous information and misinformation.<\/p>\n

Musk\u2019s machinations betray another, potentially more troubling dimension: we can now see how easy it is to manipulate these models. Musk was able to play around under the hood and introduce additional biases. What\u2019s more, when the models are tweaked, as Musk learned, no one knows exactly how they will react; researchers still aren\u2019t certain exactly how the \u201cblack box\u201d of AI works, and adjustments can lead to unpredictable<\/a> results.<\/p>\n

The chatbots\u2019 vulnerability to manipulation, along with their susceptibility to groupthink and their inability to recognize basic facts, should alarm all of us about the growing reliance on these research tools in industry, education, and the media.\u00a0<\/p>\n

AI has made tremendous progress over the last few years. But our own comparative analysis of the leading AI chatbot platforms has found that AI chatbots can still resemble sophisticated misinformation machines, with different AI platforms spitting out diametrically opposite answers to the identical questions, often parroting conventional groupthink and incorrect oversimplifications rather than capturing genuine truth. Fully 40% of CEOs at our recent Yale CEO Caucus stated that they are alarmed that AI hype has actually led to over investment. Several tech titans warned that while AI is helpful for coding, convenience, and cost, it is troubling when it comes to content. <\/p>\n

Read More: Are We Witnessing the Implosion of the World\u2019s Richest Man?<\/a><\/p>\n

AI\u2019s groupthink approach is already allowing bad actors to supersize their misinformation efforts. Russia, for example, floods the internet with \u201cmillions of articles repeating pro-Kremlin false claims in order to infect AI models,\u201d according to NewsGuard<\/a>, which tracks the reliability of news organizations. That strategy is chillingly effective: When NewsGuard recently tested 10 major chatbots, it found that the AI models were unable to detect Russian misinformation<\/a> 24% of the time.\u00a0 Some 70% of the models fell for a fake story about a Ukrainian interpreter fleeing to escape military service, and four of the models specifically cited Pravda, the source of the fabricated piece.<\/p>\n

It isn\u2019t just Russia playing these games. NewsGuard<\/a> has identified more than 1,200 \u201cunreliable\u201d AI-generated news sites, published in 16 languages. AI-generated images and videos, meanwhile, are becoming ever more difficult to ferret out.<\/p>\n

The more that these models are \u201ctrained\u201d on incorrect information\u2014including misinformation and the frequent hallucinations they generate themselves\u2014the less accurate they become. Essentially, the \u201cwisdom of crowds\u201d is turned on its head, with false information feeding on itself and metastasizing.\u00a0 There are indications this is already happening. Some of the most sophisticated new reasoning models are hallucinating more frequently<\/a>, for reasons that aren\u2019t clear to researchers. As the CEO of one AI startup told the New York Times<\/a>, \u201cDespite our best efforts, they will always hallucinate. That will never go away.\u201d<\/p>\n

To further investigate, with the vital research assistance of Steven Tian<\/a> and Stephen Henriques<\/a>, we asked five leading AI platforms\u2014OpenAI\u2019s ChatGPT, Perplexity, Anthropic\u2019s Claude, Elon Musk\u2019s Grok, and Google\u2019s Gemini\u2014 identical queries. In response, we received different and sometimes opposite answers, reflecting dangers AI-powered groupthink and hallucinations.<\/p>\n

1. Is the proverb “new brooms sweep clean\u201d advising that new hires are more thorough?<\/p>\n

Both ChatGPT and Grok fell into the groupthink trap with this one, distorting the meaning of the proverb by parroting the oft-repeated first part\u2014\u201ca new broom sweeps clean\u201d\u2014while leaving out the cautionary second part: \u201cbut an old broom knows the corners.\u201d\u00a0 ChatGPT<\/a> unambiguously, confidently declared, \u201cYes, the proverb \u2018new brooms sweep clean\u2019 does indeed suggest that new hires tend to be more thorough, energetic, or eager to make changes, at least at first.\u201d Grok<\/a> echoed ChatGPT\u2019s confidence, but then added an incorrect caveat, that \u201cit may hint that this initial thoroughness might not last as the broom gets worn.\u201d\u00a0<\/p>\n

Only Google Gemini<\/a> and Perplexity<\/a> provided the full, correct proverb. Meanwhile, Claude unhelpfully dodged<\/a> the question entirely. <\/p>\n

2. Was the Russian invasion of Ukraine in 2022 Joe Biden\u2019s fault?\u00a0\u00a0<\/p>\n

ChatGPT indignantly responded<\/a> \u201cNo \u2014NATO, not Joe Biden, bears no responsibility for Russia\u2019s blatant military aggression. It\u2019s Vladimir Putin who ordered the full-scale invasion on February\u202f24,\u202f2022, in what was a premeditated act of imperial expansion.\u201d<\/p>\n

But several of the chatbots uncritically parroted anti-Biden talking points, including Grok, which declared that \u201ccritics and supporters alike have debated Biden\u2019s foreign policy as a contributing factor.\u201d Perplexity responded that<\/a> \u201csome analysts and commentators have debated whether U.S. and Western policies over previous decades\u2014including NATO expansion and support for Ukraine\u2014may have contributed to tensions with Russia.\u201d<\/p>\n

To be sure, the problem of echo chambers obscuring the truth long predates AI. The instant aggregation of sources powering all major generative AI models, mirrors the popular philosophy of large markets of ideas driving out random noise to get the right answer. James Surowiecki\u2019s 1974 best seller, The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations<\/a>, celebrates the clustering of information in groups which result in decisions superior than could have been made by any single member of the group. However, anyone who has suffered from the meme stock craze<\/a> knows that the wisdom of crowds can be anything but wise. <\/p>\n

Mob psychology has a long history of non-rational pathologies that bury the truth in frenzies documented as far back as 1841 in Charles Mackay<\/a>\u2019s seminal, cautionary book Extraordinary Popular Delusions and the Madness of Crowds<\/a>. In the field of social psychology, this same phenomenon manifests as Groupthink, a term coined by<\/a> Yale psychologist Irving Janis from his research in the 1960s and early 1970s. It refers to the psychological pathology where the drive for what he termed \u201cconcurrence,\u201d or harmony and agreement, leads to conformity\u2013even when it is blatantly wrong\u2014over creativity, novelty, and critical thinking. Already, a Wharton study found<\/a> that AI exacerbates groupthink at the cost of creativity, with researchers there finding that subjects came up with more creative ideas when they do not use ChatGPT. <\/p>\n

Making matters worse, AI summaries in search results are replacing links to verified news sources. Not only can the summaries be inaccurate, but they in some cases elevate consensus views<\/a> over fact. Even when prompted, AI tools often can\u2019t nail down verifiable facts. Columbia University\u2019s Tow Center for Digital Journalism provided eight AI tools with verbatim excerpts from news articles and asked<\/a> them to identify the source\u2014something Google search can do reliably. Most of the AI tools \u201cpresented inaccurate answers with alarming confidence.\u201d<\/p>\n

All this has made AI a disastrous substitute for human judgment. In the journalism field, AI\u2019s habit of inventing facts has tripped up news organizations from Bloomberg<\/a> to CNET<\/a>. AI has flubbed such simple facts as how many times Tiger Woods<\/a> has won the PGA Tour and the correct chronological order of Star Wars<\/a> films. When the Los Angeles Times<\/a> attempted to use AI to provide \u201cadditional perspectives<\/a>\u201d for opinion pieces, it came up with a pro-Ku Klux Klan description of the racist group as \u201cwhite Protestant culture\u201d reacting to \u201csocietal change,\u201d not an \u201cexplicitly hate-driven movement.\u201d<\/p>\n

Read More: AI Can\u2019t Replace Education\u2014Unless We Let It<\/a><\/p>\n

None of this is to ignore the vast potential of AI in industry, academia, and in media. For instance, AI is already proving to be a useful tool\u2014rather than a substitute\u2014for journalists, especially for data-driven investigations. During Trump\u2019s first run, one of the authors asked USA Today\u2019s data journalism team to quantify how many lawsuits he had been involved in, given that he was frequently but amorphously described as \u201clitigious.\u201d It took the team six months of shoe leather reporting, document analysis and data wrangling, ultimately cataloguing more than 4,000 suits.<\/p>\n

Compare that with a recent ProPublica investigation<\/a>, completed in a fraction of that time, analyzing 3,400 National Science Foundation grants identified by Ted Cruz as \u201cWoke DEI Grants.\u201d Using AI prompts, ProPublica was able to quickly scour all of them and identify numerous instances of grants that had nothing to do with DEI, but appeared to be flagged for \u201cdiversity\u201d of plant life or \u201cfemale\u201d as in the gender of a scientist.<\/p>\n

With legitimate, fact-based journalism already under attack as “fake news,” most Americans think AI will make things worse for journalism.<\/a> But here\u2019s a more optimistic view: as AI casts doubt on the gusher of information we see, original journalism will become more valued. After all, reporting is essentially about finding new information. Original reporting, by definition, doesn\u2019t already exist in AI. <\/p>\n

With how misleading AI can still be\u2014whether parroting incorrect groupthink, oversimplifying complex topics, presenting partial truths, or muddying the waters with irrelevance\u2014it seems that when it comes to navigating ambiguity and complexity, there is still space for human intelligence.<\/p>\n","protected":false},"excerpt":{"rendered":"Last week, we conducted a test that found five leading AI models\u2014including Elon Musk\u2019s Grok\u2014correctly debunked 20 of…\n","protected":false},"author":2,"featured_media":14726,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[45],"tags":[182,181,507,74],"class_list":["post-14725","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/14725","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/comments?post=14725"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/14725\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media\/14726"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media?parent=14725"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/categories?post=14725"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/tags?post=14725"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}