AI-Generated Images and Misinformation: A Double-Edged Sword
Artificial Intelligence (AI) provides a varied range of services and continues to catch intrigue and experimentation. It has altered how we create and consume content. Specific prompts can now be used to create desired images enhancing experiences of storytelling and even education. However, as this content can influence public perception, its potential to cause misinformation must be noted as well. The realistic nature of the images can make it hard to discern as artificially generated by the untrained eye. As AI operates by analysing the data it was trained on previously to deliver, the lack of contextual knowledge and human biases (while framing prompts) also come into play. The stakes are higher whilst dabbling with subjects such as history, as there is a fine line between the creation of content with the intent of mere entertainment and the spread of misinformation owing to biases and lack of veracity left unchecked. AI-generated images enhance storytelling but can also spread misinformation, especially in historical contexts. For instance, an AI-generated image of London during the Black Death might include inaccurate details, misleading viewers about the past.
The Rise of AI-Generated Historical Images as Entertainment
Recently, generated images and videos of various historical instances along with the point of view of the people present have been floating all over the internet. Some of them include the streets of London during the Black Death in the 1300s in England, the eruption of Mount Vesuvius at Pompeii etc. Hogne and Dan, two creators who operate accounts named POV Lab and Time Traveller POV on TikTok state that they create such videos as they feel that seeing the past through a first-person perspective is an interesting way to bring history back to life while highlighting the cool parts, helping the audience learn something new. Mostly sensationalised for visual impact and storytelling, such content has been called out by historians for inconsistencies with respect to details particular of the time. Presently, artists admit to their creations being inaccurate, reasoning them to be more of an artistic interpretation than fact-checked documentaries.
It is important to note that AI models may inaccurately depict objects (issues with lateral inversion), people(anatomical implausibilities), or scenes due to "present-ist" bias. As noted by Lauren Tilton, an associate professor of digital humanities at the University of Richmond, many AI models primarily rely on data from the last 15 years, making them prone to modern-day distortions especially when analysing and creating historical content. The idea is to spark interest rather than replace genuine historical facts while it is assumed that engagement with these images and videos is partly a product of the fascination with upcoming AI tools. Apart from this, there are also chatbots like Hello History and Charater.ai which enable simulations of interacting with historical figures that have piqued curiosity.
Although it makes for an interesting perspective, one cannot ignore that our inherent biases play a role in how we perceive the information presented. Dangerous consequences include feeding into conspiracy theories and the erasure of facts as information is geared particularly toward garnering attention and providing entertainment. Furthermore, exposure of such content to an impressionable audience with a lesser attention span increases the gravity of the matter. In such cases, information regarding the sources used for creation becomes an important factor.
Acknowledging the risks posed by AI-generated images and their susceptibility to create misinformation, the Government of Spain has taken a step in regulating the AI content created. It has passed a bill (for regulating AI-Generated content) that mandates the labelling of AI-generated images and failure to do so would warrant massive fines (up to $38 million or 7% of turnover on companies). The idea is to ensure that content creators label their content which would help to spot images that are artificially created from those that are not.
The Way Forward: Navigating AI and Misinformation
While AI-generated images make for exciting possibilities for storytelling and enabling intrigue, their potential to spread misinformation should not be overlooked. To address these challenges, certain measures should be encouraged.
- Media Literacy and Awareness – In this day and age critical thinking and media literacy among consumers of content is imperative. Awareness, understanding, and access to tools that aid in detecting AI-generated content can prove to be helpful.
- AI Transparency and Labeling – Implementing regulations similar to Spain’s bill on labelling content could be a guiding crutch for people who have yet to learn to tell apart AI-generated content from others.
- Ethical AI Development – AI developers must prioritize ethical considerations in training using diverse and historically accurate datasets and sources which would minimise biases.
As AI continues to evolve, balancing innovation with responsibility is essential. By taking proactive measures in the early stages, we can harness AI's potential while safeguarding the integrity and trust of the sources while generating images.
References:
- https://www.npr.org/2023/06/07/1180768459/how-to-identify-ai-generated-deepfake-images
- https://www.nbcnews.com/tech/tech-news/ai-image-misinformation-surged-google-research-finds-rcna154333
- https://www.bbc.com/news/articles/cy87076pdw3o
- https://newskarnataka.com/technology/government-releases-guide-to-help-citizens-identify-ai-generated-images/21052024/
- https://www.technologyreview.com/2023/04/11/1071104/ai-helping-historians-analyze-past/
- https://www.psypost.org/ai-models-struggle-with-expert-level-global-history-knowledge/
- https://www.youtube.com/watch?v=M65IYIWlqes&t=2597s
- https://www.vice.com/en/article/people-are-creating-records-of-fake-historical-events-using-ai/?utm_source=chatgpt.com
- https://www.reuters.com/technology/artificial-intelligence/spain-impose-massive-fines-not-labelling-ai-generated-content-2025-03-11/?utm_source=chatgpt.com
- https://www.theguardian.com/film/2024/sep/13/documentary-ai-guidelines?utm_source=chatgpt.com