We associate various colors, shapes, and sounds with different tastes, and these connections can subtly influence our perceptions. It seems that AI is also capable of forming similar associations.
Research by an international team of scientists suggests that generative AI systems can create associations between colors, shapes, and sounds with different objects. The rapid development of AI prompted scientists Carlos Velasco, Charles Spence, and their colleague Kosuke Motoki to investigate whether generative AIs trained on human data would identify the same associations. They asked ChatGPT and Google’s Gemini to respond to the same prompts previously given to human participants.
By analyzing the results of hundreds of conversations in English, Spanish, and Japanese, the researchers concluded that AI indeed mirrors patterns characteristic of human perception. However, there were some differences observed among the various versions of AI they used.
When asked which color is the sweetest, the AI responded that pink is associated with sweetness, likely due to its connection with treats like cotton candy and bubble gum. It also mentioned that green is associated with sourness, white with saltiness, and black with bitterness.
In the future, the researchers plan to explore whether generative AIs can propose new hypotheses for other cross-modal associations not yet documented in scientific literature, which can then be tested on humans. This approach could prove valuable for marketers looking to design products or packaging that align with existing perceptual associations.
