In the development of AI technology, we’ve been using the term “hallucination” to describe the phenomenon where models generate information not based on facts. However, this term tends to give a negative impression of technical errors.
In reality, this phenomenon can be seen as AI exploring new possibilities by filling in the gaps between training data. AI generates content based on statistical patterns, which sometimes results in outputs that may differ from facts but are nonetheless creative.
This characteristic is creating new value. A research team at Stanford Medical School has discovered new antibiotic candidates through AI “hallucinations” that couldn’t have been found using conventional methods. It’s also being utilized as a source of creative planning and unexpected ideas in businesses.
While this phenomenon carries risks of generating incorrect information and requires appropriate evaluation and management, wouldn’t it be more beneficial to view it as a creative characteristic of AI rather than simply a “mistake,” thereby maximizing its potential?
So here’s my proposal – what if we adopt the term “Creative Completion” which better captures its essence? This term reflects how AI creatively fills gaps between data points, highlighting the innovative aspect that can lead to novel discoveries.
This reframing might help us focus on the “possibilities” rather than the “limitations” of AI, allowing us to explore more creative forms of human-AI collaboration.
Have you experienced new ideas or solutions from unexpected AI outputs in your industry? Or do you have any good terms to describe this phenomenon? Please share your thoughts.
This idea emerged from a conversation with an English professor at my institution. I’m reminded of how wonderful it is to have various experts nearby.
