Summary
- Chatbots and other AI language models sometimes use harmful or incorrect language stripped of necessary context, as was the case with ChatGPT’s mention of clandestine and dangerous activities during a ritual.
- The AI chatbot was actually referencing the Warhammer 40,000 game’s planet Molech, as well as other game references, and not suggesting the Celtics ceremony, as ChatGPT had apparently suggested.
- This implies that ChatGPT simply reacted to the context of the question and attempted to respond in kind, lacking any real danger.
- However, the case does highlight the fact that large language models lack adequate context and as a result, can reproduce unfit and erroneous responses.
Thus, more work still needs to be done to ensure these language models can understand and produce relevant responses that do not risk harming users.