Images can inform an ongoing assistant dialog: a vision-language model derives an environment state, which is added to the language-model prompt with the user’s request.
#Google patent application #US20260320311A1 #AI