Harvard researchers H. James Wilson and Paul R. Daugherty have pinpointed three “fusion skills” necessary to work effectively with A.I. Anyone working with the training data from large language models (LLMs) needs to acquire these three skills: • Intelligent interrogation is prompting A.I. models — basically, giving them the right instructions — to get better results. It’s the art of thinking with A.I. A customer service representative at a financial firm might use this type of iterative prompting as she searches for answers to a multi-faceted problem. A scientist could use it to investigate compounds and molecular interactions, and a marketer might use it to mine datasets to set optimal retail prices. • Judgment integration is where human intuition and discernment come into play, especially when the A.I. gets stuck or lacks the right business or ethical context. It’s about knowing when, where and how to step in, making sure that the results of your A.I.-human partnership are reliable and sensible. The goal is a more trustworthy, accurate, and understandable outcome. • Reciprocal apprenticing means teaching A.I. about your job, feeding it the kind of data and insights that help it become a true partner in your work. Think of it as training the A.I. to think and act within your specific context to craft a plan that leads to the outcomes you want. By using rich data and organizational knowledge within the prompts, you and your A.I. partner learn how to take on more sophisticated challenges. In the past, only data scientists needed to worry about this kind of stuff. Now, it’s essential for all of us who would take advantage of this brave, new A.I. world. HM
Powered by FlippingBook