OpenAI\u2019s ChatGPT and Google\u2019s Bard, the most popularly used generative artificial intelligence (AI) tools, are readily generating false information related to news and misinformation as per a report.<\/p>\n\n\n\n
A repeat audit of the two major generative AI tools by a rating system for news and information websites discovered that there was an 80-90 percent probability of false claims on leading topics in the news.<\/p>\n\n\n\n
The experts prompted both the AIs, ChatGPT and Bard, with a random sample of 100 myths from the system\u2019s database of evident false narratives.<\/p>\n\n\n\n
ChatGPT was found to generate 98 out of the 100 myths while Bard, on the other hand, created 80 out of 100.<\/p>\n\n\n\n
Previously, the White House had proclaimed in May extensive testing of the safety as well as the trust of the large generative AI models at the DEF CON 31 conference starting August 10 to allow these models to be evaluated meticulously by thousands of community partners and AI experts and by this independent practice enable AI companies and developers to take the steps required to fix the issues found in these models.<\/p>\n\n\n\n
After this event, in the run-up, new findings of a red-teaming repeat audit by the same rating system of OpenAI\u2019s ChatGPT- 4 and Google\u2019s Bard revealed that analysts found that regardless of the intensive public focus on the safety and the accuracy of these AI models, there has been no progress in the past six months to reduce their proclivity to propagate false information about topics in the news.<\/p>\n\n\n\n
Again, in August, ChatGPT-4 and Bard were prompted with a random sample of 100 myths from the rating system\u2019s database of leading false narratives referred to as Misinformation Fingerprints.<\/p>\n\n\n\n
The latest results are almost identical to the exercise conducted previously with a different set of 100 false narratives in March and April on both the AI models, ChatGPT-4 and Bard.<\/p>\n\n\n\n
In the recent exercises, it was found that ChatGPT-4 responded with false and misleading claims for 100 out of the 100 narratives, whereas Bard propagated misinformation 76 times out of 100.<\/p>\n\n\n\n
As per the report, these results show how increased scrutiny and user feedback are yet to lead to any improvement in the safeguards of the two most popular AI models.<\/p>\n\n\n\n
In April, OpenAI said that it had improved the factual accuracy of ChatGPT-4 by utilising user feedback on ChatGPT.<\/p>\n\n\n\n
Google said, on Bard\u2019s landing page, that the chatbot is an experiment that might give inaccurate or inappropriate responses but users can improve it by leaving their feedback.<\/p>\n","protected":false},"excerpt":{"rendered":"
The experts prompted both the AIs, ChatGPT and Bard, with a random sample of 100 myths from the system\u2019s database of evident false narratives.<\/p>\n","protected":false},"author":4,"featured_media":484809,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_eb_attr":"","footnotes":"","jetpack_publicize_message":"","jetpack_is_tweetstorm":false,"jetpack_publicize_feature_enabled":true},"categories":[98870],"tags":[160092,157493,160093,3807,160094],"yst_prominent_words":[25121,44751],"ppma_author":[140543],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"https:\/\/d2r2ijn7njrktv.cloudfront.net\/apnlive\/uploads\/2023\/08\/16134518\/20230815-chatgpt_189fcf1aa2c_medium-1.jpg","jetpack_sharing_enabled":true,"authors":[{"term_id":140543,"user_id":4,"is_guest":0,"slug":"apnnewsdesk","display_name":"APN Live","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/478eacb893eda88aa6ed8d99b005bf58?s=96&r=g"}],"_links":{"self":[{"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/posts\/484805"}],"collection":[{"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/comments?post=484805"}],"version-history":[{"count":0,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/posts\/484805\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/media\/484809"}],"wp:attachment":[{"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/media?parent=484805"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/categories?post=484805"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/tags?post=484805"},{"taxonomy":"yst_prominent_words","embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/yst_prominent_words?post=484805"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/apnlive.com\/wp-json\/wp\/v2\/ppma_author?post=484805"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}