“Just one word . . . plastics” is the career advice offered by the family friend Mr McGuire to a bemused Benjamin Braddock, played by Dustin Hoffman, in the 1967 movie The Graduate. A modern-day Mr McGuire might well advocate coding as the route to decently paid job security. Or he might have done before the tech industry axed 200,000 jobs in the past 12 months — and the clamour around the artificial intelligence platform ChatGPT raised awareness that machines can write code, too. Computers, it seems, may soon be programming themselves. So is what we’ve been telling our kids about safe jobs all wrong?
Parental panic would be premature. Widespread lay-offs may indeed signal that the days of perpetual growth in Big Tech companies are ending and they are starting to behave more like banks — hiring in good times and firing in bad. Some are, for the first time, being pressed by activist investors to raise profitability. But after their extraordinary hiring sprees during lockdown, the job cuts represent only a modest retrenchment. Many of the axed roles are in sales and marketing, not programming. And if the arms race among tech companies to secure computer science graduates is abating, other previously outgunned sectors — from engineering to media and finance — will be happy to pick up the spare talent.
The need for software will only grow. The pandemic drove demand from consumers needing to work, shop, educate and entertain themselves at home — and from businesses struggling to control supply chains. Now organisations are looking to IT to blunt the impact of inflation. Robotics, automation, and the digitisation of even everyday products will require a lot more code. And a lot of legacy systems will need replacing.
Rather than being a big threat to coders, generative AI can help. Computers are unlikely to be writing their own programs any time soon, but Copilot, a kind of super-smart autocomplete function for coding from Microsoft’s GitHub division, is already raising productivity. Coders write a prompt on what they’re aiming to do and Copilot suggests lines of code. Andrej Karpathy, a former director of AI at Tesla, tweeted last month that Copilot was writing 80 per cent of his code, with 80 per cent accuracy. Programmers say the tool both speeds up their work and frees them to be more creative.
Advances are also opening coding to non-specialists. Low-code or no-code development platforms allow users to create application software by using a graphical interface. Microsoft’s CEO Satya Nadella suggests this could help close skills gaps: someone with expertise in, say, media or logistics, but no coding knowledge, can get involved in developing business apps. This shifts the balance a little towards business units and away from IT.
Those who understand coding from scratch will still be needed — not least to oversee the AI and correct any machine-generated mistakes. But as the grunt work is automated away, their skills will have to move up the value curve. The US Bureau of Labor Statistics projects US employment among “programmers”, narrowly defined, will fall 10 per cent over the next decade, but sees the number of software developers, including those responsible for planning, quality assurance, testing and integration, rising by a quarter.
If automation makes the best IT brains more productive, this is a good thing. Scaling the capabilities of talented people is likely to lead to more breakthroughs. Computers taking over more technical work as humans move up to a higher level of abstraction has been happening since the dawn of computing. For the graduates of today, that means coding is still a good place to start.