The Return of Language

For the past decade, we have been fed a diet of STEM, encouraged to raise our maths, science, and engineering standards if we wanted to compete with the world’s best. Western schools and universities placed a renewed emphasis on the teaching of STEM subjects, and as societies we worried about our performance relative to other countries, and about the diversity issues in all these subjects and related professions.

One particular component of STEM, computer science, assumed outsized significance. With the rise of the GAFA giants (Google, Apple, Facebook, and Amazon) as software ate the world, coders and coding developed cult status as the source of much if not all innovation. Silicon Valley culture prized the ‘technical founder’, from whose keyboard came legends written in Java, Python, and Ruby; those who lacked such skills were ‘non-technical’, the most modern of slights.

None of this is to downplay the huge achievements of science, maths, engineering and technology in the 21st century, which include organising the world’s information, connecting more than 5 billion people, and vaccinating many of them against pandemic illness and other diseases. But if you weren’t a scientist, a mathematician, or an engineer of some kind, you might have been forgiven for feeling a little disenfranchised. And as so many of us hit our maths ceiling well before university, that’s a lot of disenfranchised people.

Ironically, it is thanks to the very brilliance of mathematicians, computer scientists, and the code they created that old fashioned ‘language’ is back. The deep learning algorithms and transformer models created by brilliant minds at Google, Meta, and OpenAI among others in the past few years haven’t created Large Number Models, but Large Language Models (LLMs), and in doing so, they have opened up a whole new world of opportunities for people skilled in the use of language as opposed to numbers and code.

Code. The word can’t help but imply secrecy, inaccessibility, and hidden knowledge. But we all have access to language. We can all communicate in our mother tongue. Some better than others, maybe, but language isn’t confined to a clique or a club, it is available to us all.

Andrej Karpathy, one of the founding members of OpenAI, may have put it best in a Tweet on 24 January 2023, when he said “The hottest new programming language is English.” He went on to explain that whereas the programmer base was hitherto limited to those with advanced coding skills, “…this new programming paradigm has the potential to expand the number of programmers to ~1.5B people”, i.e. anyone who speaks English (Author’s note: this was in early 2023, soon after the release of ChatGPT, and so the number should be increased to include speakers of any language which can be used to prompt a capable LLM).

A friend who runs a successful software development agency told me recently that his #1 coder had not personally written a line of code for six months. He was still producing great code, but uses the AI tools now available to him to increase the pace and quality of his work. There are clearly still advantages in having expertise within a specialist domain. But as our ability to use natural language to prompt AI tools extends into new domains, we are democratising access and lowering the barriers to entry into them.

In writing this note, I stumbled across Jensen Huang, founder of Nvidia, speaking at a recent summit about the importance of learning things other than computer science, arguing that ‘human’ is the new programming language, and that we are all now programmers. He is right. Anyone with domain expertise and the ability to use language creatively now has the power of advanced computing at their fingertips, and the talent pool of innovators is now deeper and wider than it has ever been. The iPhone put a supercomputer in our pockets; LLMs give all of us the ability to programme it. Language is back.