ChatGPT, the Large Language Model invented by Open AI recently passed the Advanced Sommelier Exam (along with several AP Exams, a bar exam, GRE’s, the LSAT, etc.) I suppose it is just a matter of time before it passes the Master of Wine theory exam.
That’s grim news for people who get paid to write generic wine copy.
Of course, ChatGPT is just a very powerful prediction engine with the massive resource of the Internet as a data source. It doesn’t have insight, at least not yet, and it’s prone to hoovering up the same kind of false information we routinely find on the Internet. It won’t replace good wine writing (until maybe next year with the progress it’s making.)
But it should get us to re-evaluate some of our practices and activities that might soon be better performed by a machine.
To that end, Alder Yarrow made a great point last week in his post about ChatGPT.
At what point does the accumulation of factual knowledge become a useless exercise in the face of what a given profession, situation, or task requires? ….But when an LLM can ace the hardest exams that the gatekeepers of the wine industry put up as indications of ultimate expertise, we have to ask how well the educational curriculae and learning paths that we have established for wine reflect the realities of life.
I have long thought that wine education even at the higher levels relies far to much on memorizing a pile of facts that you can easily look up in a reference book. Does every sommelier really need to know all the sub-regions of Chianti Classico and their acreage under vine, cite every permissible grape in White Châteauneuf-du-Pape, or the varieties of yeast strains and their effect on thiol expression?
Of course, someone needs to know this information. The problem is that until you know what sector of the wine business you work in, you don’t know what set of facts you need to know. So wine education programs throw it all at you. But one of the “realities of life” is that when you cram facts in order to pass an exam you rapidly forget them. Testing for facts reveals no insight, depth, or creativity on the part of the student. It’s just a way of proving you put the time in to learn the material.
Learning to taste, appreciate, and evaluate wine and especially learning to communicate that appreciation should be the focus of wine education. Learning the various ways of conceptualizing wine and how to contextualize it, and discovering how to probe the various dimensions of the wine community for new information or trends are important at the higher levels of wine education.
All education requires knowledge of some facts. But facts without relevance or context are useless and too often wine education seems more like a ritual than an exploration. Now that we have machines to regurgitate facts maybe we can focus more on what really matters.