Why T-Shaped Skills Matter More Than Ever in the Era of Generative AI
Shapes are often used as a way to describe the diverse skillsets that modern graduates need to succeed in today’s rapidly changing job market.
One of the most common concepts is that of “T-shaped” graduates, who have a deep expertise in a specific field (represented by the vertical stroke of the T) as well as a broad range of general skills and knowledge that can be applied across multiple fields (represented by the horizontal stroke of the T).
The first “official” reference to T-shaped skills, or a T-shaped person, was made by David Guest back in 1991 in an article in The Independent entitled “The hunt is on for the Renaissance Man of computing." The concept gained real popularity after the CEO of IDEO Design Consultancy firm - Tim Brown - endorsed the idea when looking over applicants’ resumes. According to Tim Brown, using the search for T-shaped skills/a T-shaped person helps build the very best interdisciplinary teams within a company.
I was reminded of T-shaped graduates by a fascinating Stanford HAI panel discussion on Generative AI and Education, featuring Associate Professor Percy Liang, Director, Center for Research of Foundational Models.
I think there's a growing fear that Foundation models will replace the need to learn certain skills. Instead of multiplying numbers or writing code, one can simply ask a model to do it for them. However, if we look at the history of technology, calculators, software for solving calculus problems, spell checkers, and even Stack Overflow have helped people work more efficiently and lessen the need for developing proficiency in certain skills.
My personal take on this is that students should still learn everything about how the world works from first principles, but they don't need to be proficient in all skills (italics mine). There is a difference between understanding how to multiply large numbers and being able to do it quickly. Foundation models can do the former for you, and you can move on to the next thing.
Instead of focusing on learning how to write code line by line, one should focus on writing good specifications and tests, which is what people should be doing anyway. Similarly, instead of just focusing on writing essays word by word, one can develop ideas in the form of an outline. There are many possibilities if you can get the basic level of skills automated in some sense.
As the type of work you're doing shifts more towards oversight and quality control, there is still a tremendous amount of skill required, but it is different than what is required at the lowest level. At each point in time, there is an element of responsibility. This is already happening in management, where there are programmers, team leads, managers, and executives, each with their own responsibilities.
In other words, as AI reduces the need for proficiency across a range of skill sets, the horizontal part of the T becomes more and more relevant.
For example, a healthcare professional who is knowledgeable about AI-powered diagnostic tools may need to have a deep understanding of medical terminology and disease pathology (represented by the vertical stroke of the T), as well as a broad range of skills and knowledge related to data analysis, machine learning algorithms, and ethical considerations in AI (represented by the horizontal stroke of the T).
In the era of generative AI, it is imperative to reconsider the importance of the horizontal part of the T-shaped graduate, who may face a "sandwich" workflow in their future careers. This process involves three steps: first, a human initiates a creative idea and provides a prompt to the AI. Second, the AI generates a menu of options based on the prompt. Finally, the human selects an option, modifies it, and adds personal touches. This "sandwich" workflow is a departure from traditional work processes and demands a fresh approach to developing the skills needed for success in an emergent AI-impacted job market.