Gemini 3 and other chatbots scrutinized as unreliable news gatherers

Straight Arrow News

Watch Gemini 3 and other chatbots scrutinized as unreliable news gatherers - Straight Arrow News on Dailymotion

Transcript

00:00AI search engines and chatbots keep rolling out new updates,

00:04with companies claiming they're more helpful and more powerful than ever.

00:09Google released its Gemini 3 Pro this week,

00:12the latest update to its artificial intelligence system.

00:15Microsoft's CoPilot chat is expanding its integration into Outlook,

00:20Word, and other Microsoft programs.

00:23And OpenAI recently launched its own search engine, Atlas,

00:27powered by its AI engine, ChatGPT.

00:31With so many ways to search the web, what source do you turn to for news?

00:36If you're relying on an AI-powered chatbot or search engine, then this story is for you.

00:45Jeffrey Blevins is a professor at University of Cincinnati

00:48and a faculty fellow of the Center for Cyber Strategy and Policy.

00:53He spoke with Straight Arrow News about using artificial intelligence sources

00:57or artificial algorithms, as he likes to call it.

01:01I've never been comfortable with the term intelligence, right?

01:05Artificial, I'm good with.

01:07And to me, you know, algorithm is just a much better, you know, moniker for that.

01:15Blevins says AI can be used for a variety of tasks,

01:18but when it comes to news consumption and reliable, verifiable information,

01:24it still has a long way to go.

01:25Well, there seems to be all this praise about it and all this hype.

01:30And I think there's a real disconnect there.

01:33It's like, yeah, I've seen, you know, Sora 2 in terms of, you know,

01:37how well it can make deepfake videos, and they are really impressive.

01:42But that's a very different thing than when we're talking about relying on,

01:48you know, information integrity.

01:50Blevins says AI can be a first step when researching news, politics, or current events.

01:57But it should never be the only source.

02:00Because as advanced as AI technology has become, it is still flawed.

02:04When we asked ChatGPT who the president of the United States was,

02:08it responded Donald J. Trump, but showed a picture of former President Joe Biden.

02:14When we asked who was in the photo it provided, it doubled down, still calling it Trump.

02:20Even though it was wrong, ChatGPT, like other AI systems,

02:24carries a disclaimer that it can make mistakes.

02:27Google's Gemini-powered AI search results display a similar reminder that they may include mistakes.

02:34And we should absolutely not be relying on it.

02:37You know, again, at best, it's, you know, a first source, you know.

02:42And then, actually, I would take issue with the word source.

02:45You know, it might be a first step.

02:47What can it, you know, gather quickly?

02:49You know, I need to be willing to, you know, to take the next action steps

02:52that go to different sources to try and verify that.

02:56Blevins also referred to a high-profile AI blunder

02:59in the Chicago Sun-Times that was then shared in the Philadelphia Inquirer.

03:04A journalist was tasked with compiling a summer reading list.

03:09And the reporter had AI do the assignment.

03:12It mixed up names, you know, authors with titles, you know, had those mismatched.

03:18You know, sometimes the titles weren't perfect.

03:20And sometimes they were books that, you know, that didn't exist at all,

03:24or authors that didn't exist at all.

03:27So that's a pretty big miss.

03:29Blevins says using AI as a starting point wouldn't have been fine,

03:33but it should never have been the final product.

03:36So if we go back, for instance, to, you know, the Chicago Sun-Times freelancer,

03:41okay, could have put that in and then went and then fact-checked other sources,

03:47you know, to get that information.

03:48What concerns me is that it's likely to develop more bad habits.

03:54It's going to encourage us to be lazy, to simply take, you know, one source as it, right?

04:03When you ask AI systems to describe their own weaknesses in news gathering,

04:08they acknowledge a number of them.

04:10Take OpenAI's new search engine, Atlas.

04:13When we asked what challenges it faces in gathering and delivering news,

04:17it replied, short answer, everything.

04:20Long answer, buckle up.

04:22It then listed issues such as accuracy, information verification, and information overload,

04:29adding that there is more content than any human brain can handle.

04:33And when news outlets issue corrections or retractions, can AI keep up with those changes?

04:39Atlas told us, short answer, no.

04:42Saying there was no indication it tracks or flags corrections to stories that it crawls.

04:48Even though we have this plethora now, it is really eroded trust.

04:52We had more trust in journalists, the institution of journalism in the past,

04:58with fewer sources than we do now.

05:02And Blevins says that even though users can customize these AI algorithms to fit their interests,

05:09including news, remember that AI has its own interests, too.

05:13I would say, you know, a bigger concern would be, you know, what is, you know,

05:20the algorithm really programmed to do?

05:22You know, what is, you know, there's a commercial interest here, right?

05:26And that is to keep you engaged.

05:28So, you know, the longer that, you know, it can keep you in front of the screen,

05:34you know, clicking on items, making comments, sharing those, you know, those items,

05:41those are all, you know, signs of engagement.

05:44The latest Pew Research survey shows 62% of Americans interact with artificial intelligence

05:51at least once a week, making it a massive market that is rapidly growing.

05:57A big thank you to Jeffrey Blevins for taking the time to talk with us about AI

06:02and how to use it responsibly when consuming news.

06:07And related to this story, Straight Arrow News recently reported on an international study

06:12that found 45% of AI responses to news prompts had at least one significant issue

06:18and 81% had some form of issue.

06:22You can find that story on our website or mobile app by searching AI News Report.

06:28For Straight Arrow News, I'm Cara Rucker.

Category

Transcript

Be the first to comment

Recommended