A research charity recently found that up to three million low-skilled jobs in the UK could be lost to AI by 2035. There are a lot of studies like this out there right now, claiming that huge parts of the global job market will be replaced by AI, but they all seem to be missing something very important: the impact of this type of AI deployment. As we know, AI has a huge skills problem. It’s not that the AI isn’t skilled (of course it’s not, but that’s a story for another day). No, if anything, it significantly exacerbates the already serious problems with skills in the modern economy. This introduction of AI will not separate economic growth from labor, as declared. On the contrary, the overall labor force will become increasingly unskilled, causing devastating economic damage. Let me explain.
AI only really improves productivity in “low-skill” jobs, such as taking notes in meetings or providing customer service.
After all, AI gets things wrong all the time. The errors it makes are hilariously called “hallucinations”, even though it’s nothing more than a blatant PR attempt to anthropomorphize the probability machine. However, these errors make it extremely difficult to use AI to enhance skilled tasks. After all, monitoring AI used in this way and identifying and correcting its errors requires significant time and effort from skilled workers. In fact, the time and cost wasted on AI monitoring is often greater than the time and cost saved by AI. This is one of the main reasons MIT found that 95% of AI pilots do not perform well and why METR found that AI coding tools actually slow down experienced programmers..
This problem isn’t going away either. Data scientists have known for years that AI is just a probability machine with diminishing returns and can always get things wrong. In fact, OpenAI recently Latest research papers More data and computing power will not reduce the level of AI’s “illusion,” and there is currently no viable way to achieve it.
However, these hallucinations are less of a problem in “low-skill” jobs and tasks. The people doing these jobs are inexperienced and often make mistakes, but that’s okay because the jobs are later passed on to more skilled workers. In these applications, AI can smooth out the output of these lower-skilled workers and make them more accessible to higher-skilled workers. For example, an AI taking notes in a meeting might make some mistakes, but it might make fewer mistakes than someone with less experience with the corporate language.
So does this mean that AI can augment or automate these “low-skill” tasks and jobs?
