Home / Technology / AI Risks Lives: Rhubarb Errors Echo in Digital Age
AI Risks Lives: Rhubarb Errors Echo in Digital Age
24 Mar
Summary
- Generative AI can produce misinformation faster than it can be corrected.
- AI models prioritize plausible-sounding sentences over factual accuracy.
- AI errors can have real-world consequences, from medical to dietary.
Generative AI, including tools like ChatGPT and Claude, does not function as a traditional search engine. Instead, these platforms create text based on patterns in vast datasets, sometimes generating plausible but incorrect information. This is analogous to historical instances where British government pamphlets erroneously advised the public to eat poisonous rhubarb leaves, leading to illness and death.
Despite corrections, the original misinformation resurfaced and was re-disseminated, highlighting the persistent danger of flawed data. Large language models, the technology behind these AI chatbots, are designed to predict the next word in a sequence, not to ascertain truth. Consequently, they can "hallucinate," presenting false statements with the same confidence as accurate ones.
Research indicates that generative AI misrepresents news up to 45% of the time. This unreliability carries significant real-world risks, including AI failing in medical emergency recognition and potentially generating dangerous advice, such as recipes that could produce toxic gases or non-existent hiking routes. Safeguards include seeking information predating AI-contaminated content, with tools available to filter for pre-November 2022 data.
