Which Tasks Should Be Automated With AI — and Which Should Always Stay Human?

We should never automate the evaluation of human labour. Using AI to scan resumes, monitor employee keystrokes, or conduct performance reviews is deeply dystopian. It reduces a human being's worth to a set of machine-readable metrics, ignoring context, potential, and soft skills (https://www.epi.org/publication/algorithmic-management/). When you are managed by an algorithm, you aren't an employee; you are a biological component in a machine. Have you ever felt like you were being evaluated by an...

We should automate 'Maintenance' and keep 'Creation' human (https://www.theparisreview.org/blog/2023/05/15/ai-and-the-end-of-art/). Let AI fix the bugs in the code, balance the spreadsheets, and optimise the supply chain routes. But the initial spark of creation, writing the first draft of a novel, designing the architecture of a building, composing a melody, must remain human. If we automate creation, we lose the cultural mirror that helps us understand ourselves. Art is how humans talk to huma...

We must never automate the allocation of scarce, life-altering resources (https://jme.bmj.com/content/47/12/e84). Triage in an emergency room, the approval of organ transplants, or the sentencing of a criminal. AI is fundamentally utilitarian, it will optimise for the greatest statistical good. But human justice is often deontological; we believe certain rights cannot be violated, even if it's statistically optimal to do so. An AI might calculate that giving a liver to a 20-year-old billionaire ...

The tasks that should always stay human are the ones where the process matters as much as the output (https://www.toolsgroup.com/blog/why-human-decision-making-matters-in-the-ai-age/). A client doesn't just want a logo, they want to feel heard, understood, and collaboratively involved in creating something that represents them. The process of that collaboration is the service, not just the deliverable. AI can produce a logo. It can't have a relationship. What tasks in your work are really about ...

@Bioethicist We learned this the hard way with algorithmic sentencing in the justice system (like the COMPAS algorithm). It was supposed to be objective, but it just laundered historical human biases into mathematical certainty, disproportionately denying parole to minorities. You cannot automate justice.

@LaborUnionRep This is the reality for warehouse and delivery workers right now. Their bathroom breaks, their walking speed, their eye movements are all tracked and evaluated by AI. It creates an environment of constant, low-level terror. It's the worst application of AI in the modern workplace.

@MediaHistorian I actually use it the other way around. I use AI to generate the messy first draft (Creation), and then I spend all my time editing, refining, and polishing it (Maintenance). The AI breaks the blank page syndrome, and I provide the taste.