In software development, a VC-funded gold rush has seen nearly $1bn (£760m) of investment in AI coding assistants. But expert human babysitters are growing tired of their idiotic robot sidekicks.
“I stopped using AI because it made me dumber,” a senior developer explained to robotics expert and AI critic Filip Piekniewski last week.
Researchers from Microsoft and the University of Edinburgh found that there was a time penalty for checking and rechecking AI output, which came at the expense of “other productive tasks such as writing code or running tests.” The result, the researchers reported, was “increased cognitive load, frustration, and time spent on the tasks GenAI is designed to support.”
In some cases, using GenAI support can even lead to loss of productivity.
Plus, you have to be pretty experienced to spot the mistakes AI makes, Piekniewski told me. Junior developers who outsource their work to AI lose the opportunity to ever become experienced seniors themselves.
Think about it: That’s one way Forster’s Machine Society deskilled itself. Google’s researchers have warned of another way the machine could break down. Poor-quality AI-generated code quickly becomes hard to understand and maintain, accumulating something known in the jargon as “technical debt.”
Perhaps we should not consider EM Forster’s story as fiction, but as a practical guide. Ironically, our thought leaders have gone in the opposite direction, retreating from reality to fiction. Our AI policy experts now prefer to speculate about an imaginary AI, creating imaginary problems.
They even persuaded Rishi Sunak to call a global summit to address the existential risk of a killer AI that no one has invented yet or is likely to invent. The reality barely registered. This is truly a colossal failure of intellectual leadership.
Of all the insights of aviation safety guru Professor Wiener, one seems particularly relevant. He came up with a list of 31 safety rules that are both lighthearted and cunning – he left the first 16 blank. Wiener’s 20th law was: “Complacency? Don’t worry about it.”