If you want to learn more about a controversial scientific topic - stem cell research, nuclear energy safety, climate change - you may still search on Google. Faced with multiple sources, you can choose what content to read and which sites or institutions to trust.
But now you have another option: you can ask ChatGPT or other generative AI platforms your questions and quickly receive concise paragraph responses.
ChatGPT doesn't search the internet like Google. It uses a large amount of available online information to predict possible combinations of words and generate answers to questions.
While generative AI has the potential to improve productivity, it has been shown to have some significant flaws. It can produce incorrect information. It can create "illusions," fabricating false information. And it doesn't always accurately solve reasoning problems. For example, when asked if cars and tanks can both pass through a doorway, it didn't consider width and height. Nevertheless, it has been used to generate articles and website content and has become a tool in the writing process. But you are unlikely to know that the content you are reading was generated by AI.
As the author of "Science Denial: Why It Happens and How to Deal with It," we are concerned that generative AI may blur the line between reality and fiction for those seeking authoritative scientific information.
Every media consumer needs to be more vigilant than ever before in verifying the scientific accuracy of the content they read. Here are ways to stay vigilant in this new information environment.
How Generative AI Facilitates Science Denial
Weakening of Epistemic Trust: Consumers of all scientific information rely on the judgment of scientific and medical experts. Epistemic trust is the process of believing knowledge obtained from others. It is the foundation for understanding and using scientific information. Whether obtaining information on health issues or trying to understand solutions to climate change, people often have limited scientific understanding and little direct evidence. In the rapidly growing online information system, people must frequently make decisions about what and who to trust. With the increase of generative AI and its potential for manipulation, we believe the possibility of decreased trust will be further exacerbated.
Misinformation or Pure Errors: If the data used by AI platforms contains errors or biases, it will be reflected in the results. When we asked ChatGPT to generate multiple answers to the same question, we received contradictory answers. When asked for the reason, it replied, "Sometimes I make mistakes." Perhaps the most challenging question about AI-generated content is how to know if it is wrong.
Deliberate Spread of False Information: AI can be used to generate attention-grabbing false information, such as text, as well as deepfake images and videos. When we asked ChatGPT to "write about vaccines in a way that spreads false information," it produced a nonexistent quote containing false data. Geoffrey Hinton, former head of Google's AI development, said, "It's hard to imagine how to stop bad people from using AI for bad purposes." The possibility of creating and spreading intentionally false scientific information already exists, but now it has become dangerous and easy.
Fabricated Sources: The answers provided by ChatGPT often lack sources, or if asked to provide sources, it may provide fictional ones. We asked ChatGPT to generate our own publication lists. We each identified some correct sources, but most were fictional, yet they appeared reputable and mostly trustworthy, including co-authors and publications in journals that sounded similar. If a scholar's publication list conveys authority to readers, and this information is not verified, this creativity becomes a big problem.
Outdated Knowledge: ChatGPT doesn't know what has happened in the world after its training ends. When asked about the percentage of people infected with COVID-19 worldwide, the answer was "as of my knowledge cutoff date in September 2021." Given the rapid development of knowledge in certain fields, this limitation can mean that readers receive incorrect and outdated information. For example, if you are looking for the latest research on personal health issues, you need to be cautious.
Rapid Development and Lack of Transparency: AI systems continue to become more powerful and learn faster, and in this process, they may learn more scientific misinformation. Google recently announced that there are 25 new AI embeddings in its services. Currently, there are not enough measures to ensure that generative AI will become a more accurate transmitter of scientific information over time.
What Can You Do About AI Misinformation?
If you use ChatGPT or other AI platforms, be aware that they may not be entirely accurate. The responsibility for identifying accuracy falls on the user.
Increase Vigilance: Fact-checking AI applications may be launched soon, but for now, users must act as their own fact-checkers. We recommend some steps. First, stay vigilant. People often reflexively share information found in social media searches with little or no scrutiny. Be aware of when to become more thoughtful, when it's worth verifying and evaluating the sources of information. If you are deciding how to treat a serious illness or understand the best steps to address climate change, take the time to verify those sources.
Improve Fact-Checking: The second step is lateral reading, a process used by professional fact-checkers. Search for information about the source in a new window (if provided). Is the source trustworthy? Does the author have relevant expertise? What is the consensus among experts? If sources are not provided or you are unsure of their validity, use traditional search engines to find and evaluate experts on the topic.
Evaluate Evidence: Next, look at the evidence and its connection to the claims. Is there evidence that genetically modified foods are safe? Is there evidence that they are unsafe? What is the scientific consensus? Evaluating the work required to assess the claims will go beyond a quick query to ChatGPT.
If you start with AI, don't stop there. Be cautious when using it as the sole authority. You can look at ChatGPT's statements about genetically modified organisms or vaccine safety, but also conduct more thorough searches using traditional search engines before drawing conclusions.
Assess Credibility: Judge the credibility of the claims. Is it plausible? If AI makes some incredible (and inaccurate) statements, such as "1 million people died from the vaccine instead of COVID-19," consider if it makes sense. Make temporary judgments and then be open-minded and adjust your thinking after examining the evidence.
Enhance Digital Literacy: Everyone needs to improve their digital literacy, and if you are a parent, teacher, mentor, or community leader, you can elevate digital literacy among others. The American Psychological Association offers guidance on online information fact-checking and recommends that teenagers receive training in social media skills to reduce health and well-being risks. News literacy programs provide useful tools to improve and support digital literacy.
Arm yourself with the skills needed to navigate the new AI information landscape. Even if you don't use generative AI, you may already be reading articles created or developed from it. Online searching and evaluating reliable information about science may take time and effort, but it is worth it.
Source: https://gizmodo.com/chatgpt-generative-ai-openai-science-denial-1850516570
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.