Higher education after ChatGPT

Higher education after ChatGPT

These days, it is virtually impossible to have a conversation with students without someone mentioning ChatGPT. In a way, it makes sense: much of what we ask from students is for them be able to summarize, compare, and assess common methods, ideas, concepts, frameworks, etc., in the discipline. These are the things that ChatGPT does well; hence, their concern.

Reactions have run the gamut from panic to interest. Some, such the New York City Department of Education, the International Conference on Machine Learning (ICML), or the Chinese education authorities, have chosen to ban its use. A measure that everyone can agree will be hard to enforce.

We have also seen technological proposals designed to “detect” whether a text has been generated with ChatGPT, such as GPTZero, models that use indicators such as perplexity and burstiness, that use a text’s entropy, to determine whether it is uniform and matches the student’s type of writing. Readers can be forgiven for thinking that this, too, is unlikely to work; I agree.

These attitudes are nothing new, nor are they useful. Over the decades, we have seen how technology has allowed us to augment human capabilities; every time, it has been met with similar reactions. The advent of calculators would cause people to forget how to add or multiply or do square roots, yet we should keep explaining manual algorithms, some said. Excel and, especially, Wikipedia had a similar impact.

But reality is notoriously stubborn. No matter how hard we try to stop the inevitable, we will not succeed. Generative language models, such as ChatGPT, are not only here to stay, but will progress and improve quickly, driven by stiff competition in the tech industry.

Just as with Wikipedia or Excel, we must integrate them into our learning models, which will be less rote, less oriented toward summarizing, toward recreating existing knowledge over and over again, and more oriented toward applying it to specific problems and discussing its applicability.

Questions such as which innovation methodology is most appropriate in this case or what are the pros and cons of applying missions, competitive projects or generating disruptive innovation will replace all those questions that used to begin with “Describe….”

Enhancing our capabilities with intellectual tools results in higher quality work, with a greater contribution, and that is part of the challenge.

Esteve Almirall, Operations, innovation and Data Science, Esade