Software developers have been the superstars of the information age. They have created the B2B software applications that have delivered dramatic productivity improvements to business. They have enabled the B2C consumer-app ecosystem that has opened up new possibilities like ride-sharing and upended many industries like retail. Their success has also led to politicians openly fretting about the growing gap between digital “haves” and “have nots”, while pushing for coding to be taught at schools.
But what if anyone could write an app with plain English?
As reported by Wired, in July 2020 entrepreneur Sharif Shameem created a simple app entirely by inputting natural language instructions into an artificial intelligence system called GPT-3. Within seconds of describing in plain English what he wanted to be built, GPT-3 spat out functioning code. I highly recommend watching the video on his twitter feed.
GPT-3, or Generative Pre-Trained Transformer 3, is a language model developed by OpenAI, an artificial intelligence laboratory in San Francisco. It was built by directing machine-learning algorithms to study the statistical patterns in almost a trillion words collected from the web and digitised books, including coding tutorials. It then uses this repository to respond to a text prompt by generating new text with similar statistical patterns.
Whilst there are still significant limitations in what GPT-3 can do, it is already being commercialised. For example, according to the Economist, Israeli software provider Tabnine leverages GPT-2 – an earlier iteration of GPT-3 – to provide software to speed up the coding process. Specifically, Tabnine’s software can detect that a user has begun to type code, and provide a pop-up list of suggested code to enable the user to autocomplete what they were trying to code. This should enable coders to save time by significantly cutting their keystrokes.
Whether knowledge of coding languages eventually becomes irrelevant altogether is still an open question. However, it is highly likely that, at a minimum, the coding process will become orders of magnitude more efficient. This, in turn, will have dramatic implications for technology vendors, non-technology businesses and society as a whole.
I see at least four clear trends which are likely to play out as a result of this.
At Potentia we constantly monitor how new technologies can be adopted in our portfolio companies. For example, we introduced Robotic Process Automation into our then-portfolio company Ascender in 2019, which resulted in Ascender delivering to clients more efficient payroll processing with fewer errors. We are eagerly watching the development of no-code and low-code development platforms and look forward to the opportunities they are likely to bring.