Another day another interesting development in the latest technology in the world of AI. BAYOU, a ‘deep learning’ tool being developed by a team of scientists at Rice University of Houston, Texas, is able to write code to order. The project has received funding from both Google parent company Alphabet and the US military and raises the question of whether even the occupation of software developer is under threat from technology advances. Could the profession cannibalise itself by building software that can build software?
A January 2017 report by McKinsey and Company detailed how 60% of tasks and 30% of current occupations could be automated by technology within the medium term future. A month earlier Bank of England Governor Mark Carney, then the Bank’s chief economist, warned 15 million UK jobs, almost half the country’s employment, were under threat from automation. At the time it was reported that the industries and professions considered to be at risk were admin, clerical and production jobs as well as rule-based occupations like accountancy.
It has always been assumed that the professionals that can sleep easy are those in the technology sector. The current generation of the kind of ambitious parents that used to send their children to extra-curricular math or foreign language classes are now packing them off to private schools that teach coding after school and at the weekend. “As long as little William and Lucy can write code, they’ll always have a good job”, is the implied assumption. Accountants and even lawyers might see their employment opportunities and earning potential ravaged by the tech revolution but at least those tasked with building their replacements are safe, right?
News of BAYOU may raise doubts around the solidity of the foundations that logic is built upon. The AI has read the Java source code and accompanying notes for around 1500 Android apps. The front end of the programme works like a search engine. That has created a database of around 100 million lines of Java code, accompanied by a parallel index of what each line of code does and the wider context of the application for which it was written. Users type in a few keywords describing what functionality needs to be programmed and BAYOU offers up the pre-written code that achieves the desired result.
Because BAYOU has the contextual information on what each app in its database was intended to do, as well as the details of the code architecture that enables the different functionality, the AI can actually put different sections of code together in a way that creates a whole new working piece of software.
The good news for software developers the world over, is that BAYOU is a long way away from potentially replacing the need for human developers. Rather, the team’s eventual goal is to make learning to code simpler and more intuitive. BAYOU is more an aid than a replacement and creates ‘sketches’ of programmes. Other than for very simple functionalities these will need to be pieced together into a larger whole and in most cases adjusted to the particular project. The AI is also envisioned as a resource developers can check their own code against. The project is also still in proof-of-concept testing.
However, the BAYOU project does raise the prospect that being able to code websites and other software-based products with reasonably complex functionalities will in the future have much lower barriers to entry. Several projects that hope to mean building software becomes something more like creating a complex Lego structure are in progress. The reality is that all but the most complex and innovative software architecture will, within the next decade, most probably be a case of putting together blocks of pre-written code rather than actually writing it. This will cut out much of the time required to learn different coding languages.
While the pace of the advances in the latest technology in the world is admittedly faster than ever, history has demonstrated that while professions can become obsolete as a result of new technology, new jobs are created to replace them. That doesn’t mean the transition is not sometimes painful but forecasts of a world in which humans don’t have to work, the positive scenario, or one of mass unemployment, the negative alternative, have so far always fallen far short. In the 1930s, the economist John Maynard Keynes predicted that a 15 hour working week would be the norm for his grandchildren.