Sunday, 19 January 2025
25.5 C
Singapore

Meta’s AI blames hallucinations for the incorrect Trump rally response

Metaโ€™s AI assistant mistakenly claimed a Trump rally shooting didn't happen, highlighting ongoing challenges with AI-generated inaccuracies.

Metaโ€™s AI assistant recently made headlines for an error involving the attempted assassination of former President Donald Trump. The AI incorrectly stated that the event didnโ€™t happen, which a company executive has now attributed to the technologyโ€™s inherent limitations.

AI assistant’s incorrect response

In a blog post published on July 30, Joel Kaplan, Metaโ€™s global head of policy, described the AI’s responses to questions about the shooting as โ€œunfortunate.โ€ Initially, Meta AI was programmed to avoid responding to questions about the attempted assassination. However, after users began to notice this restriction, the company decided to remove it. Despite this change, the AI provided incorrect answers in a few instances, sometimes asserting that the event didnโ€™t occur. Kaplan assured that the company is actively working to correct these errors.

โ€œThese types of responses are called hallucinations, an industry-wide issue seen across all generative AI systems. Itโ€™s an ongoing challenge to see how AI handles real-time events in the future,โ€ Kaplan explained. He added, โ€œLike all generative AI systems, models can return inaccurate or inappropriate outputs. Weโ€™ll continue to address these issues and improve these features as they evolve and more people share their feedback.โ€

<blockquote class="twitter-tweet" data-media-max-width="560"><p lang="en" dir="ltr">Meta AI wonโ€™t give any details on the attempted ass*ss*nation.<br><br>Weโ€™re witnessing the suppression and coverup of one of the biggest most consequential stories in real time.<br><br>Simply unreal. <a href="https://t.co/BoBLZILp5M">pic.twitter.com/BoBLZILp5M</a></p>&mdash; Libs of TikTok (@libsoftiktok) <a href="https://twitter.com/libsoftiktok/status/1817654239587701050?ref_src=twsrc%5Etfw">July 28, 2024</a></blockquote> <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>

Broader industry challenges

This incident is not isolated to Meta. On the same day, Google had to refute claims that its search autocomplete feature was censoring results about the assassination attempt. Former President Trump commented on the situation in a post on Truth Social, accusing Meta and Google of attempting to influence the election. โ€œHere we go again, another attempt at RIGGING THE ELECTION!!! GO AFTER META AND GOOGLE,โ€ he wrote.

Since the launch of ChatGPT, the tech industry has been grappling with how to manage generative AIโ€™s tendency to produce false information. Some companies, including Meta, have tried to anchor their chatbots with quality data and real-time search results to mitigate these issues. However, this incident demonstrates the difficulty of overcoming the inherent design of large language models, which can sometimes generate inaccurate information.

Ongoing improvements

Metaโ€™s approach to this problem involves continuous improvements and user feedback. Kaplan highlighted that the company is committed to refining its AI systems to minimise inaccuracies. He emphasised that while generative AI has advanced significantly, it still faces challenges, especially when dealing with real-time events.

The situation underscores a broader issue within the AI industry: the balance between providing helpful, accurate information and managing the AI’s propensity for generating incorrect or misleading content. Companies like Meta and Google must find more effective ways to ensure their systems deliver reliable information as AI technology evolves.

Metaโ€™s commitment to addressing these challenges and improving AI systems is crucial. By doing so, the company aims to enhance the reliability of its AI assistants, ultimately providing users with more accurate and trustworthy responses.

Hot this week

Appleโ€™s sleek iPhone 17 Air is expected to launch this autumn

Appleโ€™s iPhone 17 Air, a slim model launching this autumn, may feature future techs like foldable-ready design, in-house modems, and AI advancements.

Amazon to acquire Indian BNPL startup Axio for over US$150M

Amazon is acquiring Indian BNPL startup Axio for over US$150M, strengthening its push into financial services in one of its fastest-growing markets.

Senator Ed Markey pushes for TikTok ban deadline extension

Senator Ed Markey is pushing to delay the TikTok ban deadline by 270 days, giving the platform time to address concerns before a shutdown on January 19.

More applicants but harder to hire: LinkedIn highlights hiring challenges in 2025

LinkedIn's 2025 research highlights hiring struggles in APAC, driven by a skills mismatch, rising AI demands, and new tools to address these challenges.

Nvidia criticises Bidenโ€™s AI chip rules while seeking Trumpโ€™s support

Nvidia criticises Bidenโ€™s new AI chip restrictions, aligning with Trumpโ€™s policies while highlighting risks to US innovation and global competitiveness.

ASUS introduces ProArt Display 5K PA27JCV for creative professionals

ASUS unveils the ProArt Display 5K PA27JCV, a 27-inch monitor offering 5K resolution, Delta E<2 colour accuracy, and advanced features for creators.

Character AI tests games on its platform to boost user engagement

Character AI introduces games to its platform to boost user engagement and enhance its entertainment offerings.

How to download your TikTok videos and data before the ban

The Supreme Court has upheld a TikTok ban, and hereโ€™s how you can back up your videos and data before it happens.

ChatGPTโ€™s head of product to testify in US antitrust case against Google

ChatGPTโ€™s head of product, Nick Turley, will testify in the US governmentโ€™s antitrust case against Google, addressing AI and competition issues.

Related Articles