5. Gimme A[I] break
Even large language models need a “brain break.” KitKat’s most recent work in Canada, from Courage, shows users to tell AI to “Have a break, and then…” before asking it a question, which can yield a more accurate answer. For example, when instructed to “Take a deep breath and work on this [math] problem step by step,” large language models (LLMs) yielded their highest accuracy score of 80.2%, per the study, which was conducted by Google’s AI research lab, DeepMind. The “Have AI Break” is a clever take on KitKat’s classic “Have a break” tagline at a time when AI is getting a lot of buzz.
More news: Read Super Bowl ad scripts generated by AI