<h2>Dozens of news websites generated by AI chatbots</h2>
<p>Just over two weeks ago NewsGuard published an <a href="https://www.newsguardtech.com/special-reports/newsbots-ai-generated-news-websites-proliferating/">alarming article on the rise of newsbots</a>. They have <a href="https://www.newsguardtech.com/press/newsguard-now-identifies-125-news-and-information-websites-generated-by-ai-develops-framework-for-defining-unreliable-ai-generated-news-and-information-sources/">identified 125 news and information sites</a> that appear almost entirely written by artificial intelligence software. They report AI chatbots, like ChatGPT and Google Bard, have been used to populate so-called content farms, creating large amounts of clickbait articles and optimising advertising revenue.</p>
<p>Although most news websites did not advance misinformation on a large scale, some of these websites generated falsehoods and prompted AI chatbots to produce misleading headlines and stories, likely to increase traffic to the website. Articles with headlines such as “Biden dead. Harris acting President, address 9 a.m.” published by a website called CelebritiesDeaths.com showcased the capability of AI chatbots for spreading misinformation. As technology improves, concerns about <a href="https://www.poynter.org/ethics-trust/2023/chatgpt-build-fake-news-organization-website/">quick-scaling fake news websites</a> have been raised, lowering the barrier and allowing anybody to start such an operation.</p>
<p>However, as NewsGuard highlighted, these content farms come with problems. All sites published at least one article containing error messages commonly produced by AI chatbots, like “as an AI large language model” and “I cannot complete this prompt”. They also often used bland, repetitive language, using “in conclusion” and “it is important to remember”. These error messages are set up for AI chatbots by their creators to serve as a content filter to prevent the user from generating malicious content, although these <a href="https://openaimaster.com/how-to-bypass-chatgpt-filter/">can be bypassed</a> too.</p>
<h2>How do AI chatbots like ChatGPT create misinformation?</h2>
<p>When OpenAI released their AI chatbot ChatGPT on 30 November 2022, they enabled all internet users to conduct dialogues with the AI for free. The user could prompt anything they liked, whether they were queries or tasks like “Write me a poem”, and they would receive a very convincing text in return.</p>
<p>So convincing that soon people realised it could be a powerful tool for spreading misinformation. Not everything an AI chatbot writes is factual; instead, the AI often makes up a confident response with made-up conclusions. Common hallucinations occur when tasked with maths equations or generating biographies for famous people. But AI chatbots can also be misled to create falsehoods. Professor Ethan Mollick <a href="https://www.npr.org/2022/12/19/1143912956/chatgpt-ai-chatbot-homework-academia">described ChatGPT as</a> “an omniscient, eager-to-please intern who sometimes lies to you”. When prompted, AI chatbots will please the user and can <a href="https://www.nytimes.com/2023/02/08/technology/ai-chatbots-disinformation.html">share conspiracy theories in increasingly credible and persuasive ways</a>. “The danger is that you can’t tell when it’s wrong unless you already know the answer,” <a href="https://twitter.com/random_walker/status/1598385725363261441?s=20&#x26;t=_dnuOM_AA5EgxWpGAS-8cw">said Princeton professor Arvind Narayanan</a>. With one false narrative in mind, an AI can be tasked to generate many convincing fake news stories.</p>
<h2>Where do AI news content farms come from?</h2>
<p>A <a href="https://en.wikipedia.org/wiki/Content_farm">content farm</a> is an organisation that generates a large amount of web content specifically designed to satisfy algorithms, and maximise visitors to their websites, to generate advertising revenue. Content farms are not inherently bad <a href="https://web.archive.org/web/20100413173743/http://econsultancy.com/blog/5725-usa-today-turns-to-the-content-farm-as-the-ship-sinks">but are generally considered so</a> for numerous reasons, as their content is seen as spam with low-quality content and bad conditions for writers. News content farms powered by AI chatbots offer a new paradigm to the criticisms of content farms. AI can be leveraged to churn out even more <a href="https://www.hypergridbusiness.com/2023/02/ai-farms-are-the-new-content-farms-and-are-just-as-bad/">useless fluff</a>, allowing articles to be generated faster and cheaper.</p>
<p>Any person with access to the internet can start a content farm. All you need is a website with a strong domain name. And with free access to tools to make a website, optimise <a href="https://en.wikipedia.org/wiki/Search_engine_optimization">SEO</a>, and even AI chatbots to generate content, anybody can get started in no time without having to be a good copywriter or needing a big team. Since anybody can own a content farm, the dangers of spreading misinformation are more prevalent, as any bad actor can decide to start one today. And because cheap, fake traffic to a website <a href="https://www.vice.com/en/article/9akeye/how-to-build-a-content-farm-in-20-minutes">can be bought</a>, growing a news content farm is made easy with enough capital.</p>
<p>Whether the organisations behind news content farms are bad actors depends on their motivation. However, it is easy to see how leveraging AI to generate misinformation on news content farms can be advantageous in influencing others. For example, Russia has been accused of using <a href="https://www.nytimes.com/2022/11/06/technology/russia-misinformation-midterms.html">social media bot accounts to influence presidential elections</a> in the USA. They did not have access to advanced AI chatbots in the past, and <a href="https://theintercept.com/2023/01/10/russia-twitter-bots-trump-election/">according to The Intercept</a>, their effect on the 2016 presidential election was negligible. The convincing text generated by AI chatbots opens up the possibility for more effective influence by bad actors.</p>
<h2>Other methods of AI-generated fake news</h2>
<p>So far, we have looked at AI chatbots and how they can be used to generate and spread misinformation. Besides AI chatbots, many other generative AI tools can help create content, and as we discovered, that can also help create misinformation.</p>
<p>A term that has been around for some years and that is commonly associated with fake news is deep fakes. AI-generated fake videos made by deep learning technology to swap faces on videos, fabricating what you see and what you hear. With deep fakes, videos of politicians, celebrities and pornographic material <a href="https://www.theguardian.com/technology/2020/jan/13/what-are-deepfakes-and-how-can-you-spot-them">can be manufactured</a>. A solution to deep fakes could be AI, able to detect AI-generated imagery, or a strategy focusing on the provenance of media, like digital watermarks, so that users can find the origin of such information. AI-generated fake videos made by deep learning technology to swap faces on videos, fabricating what you see and what you hear.</p>
<p>A photo of Pope Francis wearing a white puffer jacket <a href="https://www.buzzfeednews.com/article/chrisstokelwalker/pope-puffy-jacket-ai-midjourney-image-creator-interview">went viral</a> seven weeks ago. Many people were fooled by the image, which happened to be generated entirely using <a href="https://www.midjourney.com/home/">Midjourney</a> by internet user Pablo Xavier. Using simple text prompts, any image imaginable can be generated. <a href="https://www.lemonde.fr/en/les-decodeurs/article/2023/03/31/thought-the-pope-francis-puffer-photo-was-real-here-s-how-not-to-get-fooled-by-ai-next-time_6021336_8.html">These models are not perfect</a>, though. Pay attention to some of their weaknesses, such as weird hands, inconsistencies in headwear, and meaningless text.</p>
<p>While these tools are excellent for enabling new means of expressing creativity, AI-generated media also threatens the trustworthiness of digital information. The photo of Pope Francis was not the only image fooling people online. AI images showcasing <a href="https://www.bbc.com/news/world-us-canada-65069316">former US president Donald Trump being detained by police</a> went viral around the time he appeared in court for criminal charges.</p>
<h2>What can you do against AI-generated fake news?</h2>
<p>Misinformation has existed for a long time, but AI allows this to happen quicker. With the availability of more advanced AI tools that empower all internet users to prompt engineer content to spread misinformation, more than ever, we can do very little about it. As these AI models improve, they will become more convincing, and it will become easier to be fooled by them.</p>
<p><a href="https://theconversation.com/ai-tools-are-generating-convincing-misinformation-engaging-with-them-means-being-on-high-alert-202062">Fact-checking will be our primary defence in the future</a>, and critical thinking and verification skills are vital. Ask questions and verify the information you gather online. Always be sceptical of the content you come across, and verify it with other sources. Detection systems and watermarks on digital content offer another way to battle fake news. Platforms aim to label false content or AI-generated. However, these models have mistakes and do not always get it right. This can undermine user trust when they mislabel information and give people false confidence. Therefore, we must make access to transparent and objective information possible for everyone. Education in news literacy is instrumental as a foundation for the knowledge required to navigate the internet, and tools should be provided to make navigating the internet easier.</p>
<p>At Sourcer, we want to help you verify the credibility of news and empower you to navigate the complex media landscape confidently. We help you by making news transparent so you can make more informed decisions. Our browser extension analyses news articles you visit, allows you to compare multiple news articles from different sources quickly and helps you determine the credibility of news articles. Try it out if you want to improve your media literacy skills and develop a broader perspective on current news.</p>


Product

Bedrijf

Verbinden