Illustration: Tag Hartman-Simkins / Futurism. Source: Getty Images
A veteran programmer shares his brutally honest thoughts on the role of AI in the workplace. This is as much an indictment of the technology as it is an indictment of organizations that have been lazy in implementing AI.
In an X rant that has been praised in online programming circles, programmer Dax Raad said that what holds software companies back is not the speed at which they can churn out code, but the quality of their ideas. This is a problem that AI cannot solve, even though the industry is emphasizing its ability to significantly improve productivity.
“Organizations rarely have good ideas, and it actually helps that ideas are expensive to implement,” wrote Raad, whose company OpenAuth sells AI tools.
And workers aren’t using AI to become 10x more efficient, he continued. Rather, “they use it to mass-produce tasks with less energy consumption.”
Even worse, “the two people on the team who actually try it out get fed up with the crappy code they’ve all written and quit right away.”
“Even if we could produce work faster, bureaucracy and many other realities of shipping the real thing remain bottlenecks,” Lard concluded.
There are several studies that back up Mr. Lard’s scathing assessment. Ongoing research is reported harvard business review A study that monitored 200 employees at a U.S. technology company found that rather than reducing the workload of employees, AI actually enhanced their work. Using AI to speed up tasks turns out to be a double-edged sword. This is because it causes “workload creep,” where AI increases expectations of how quickly workers have to churn out things, which in turn creates a vicious cycle in which workers become even more dependent on AI to meet ever-greater demands. The result is worker fatigue, burnout, and poor work quality. It is not a hallmark of a thriving organization.
Another study documented how AI led employees to misidentify low-quality “work failures” that appeared to be good work but actually required someone else downstream to fix it. Not only did it slow everything down, but it also created resentment among co-workers, with some admitting that receiving work product from co-workers made them feel less valued.
As Raad makes clear, AI is not a panacea. Even if AI improves productivity, that productivity can be illusory. How often do AI models produce bad code? And what happens if that bad code goes unnoticed? Perhaps, as Raad suggests, it was a good thing that the idea was “costly to implement.” Because engineers had to think creatively about problems. You don’t have to entertain every impulse. What are 1,000 ideas killed by AI rather than a few promising ones that you have honed with time and care? The former may seem more productive when they are actually a collection of dead ends.
Furthermore, if employees become dependent on AI, it is unlikely that creativity will be rewarded or fostered. As many experts have warned, this is another form of cognitive offloading, where important functions of the brain, including critical thinking, are outsourced to technology.
However, this is not what tech companies are promoting. Nvidia CEO Jensen Huang reportedly told employees it would be “insane” not to use AI to complete every conceivable task. Microsoft’s AI CEO Mustafa Suleiman claims that AI is already so effective that within 18 months, virtually all white-collar jobs will be automated. And Microsoft and Google both boast that more than a quarter of their code is generated by AI.
But no matter how useful or not these AI tools are, they cannot work miracles. At the end of the day, it’s humans who run the tough ship.
Even though AI can be used to “produce jobs faster,” “bureaucracy and many other realities of shipping the real thing are still bottlenecks,” Lard said.
Learn more about AI: Large-scale survey of CEOs and other executives reveals frightening findings about AI’s impact on productivity
