AI is progressing drastically, and imagine they tell you you're fired because a robot will take your place. What are some jobs that can never be automated?

  • 3
    $\begingroup$ Which future are you talking about? 10+, 50+, 100+, 1000+ years? $\endgroup$
    – kenorb
    Commented Oct 1, 2016 at 11:16
  • 1
    $\begingroup$ Being a mother? $\endgroup$
    – SmallChess
    Commented Apr 12, 2017 at 1:59

6 Answers 6


The Oxford study from 2013 in The future of employment paper assess this and estimated the probability of computerisation for 702 detailed occupations using a Gaussian process classifier (using job data from the UK partially merged with data from US), and based on these estimates they identified three areas of computerisation bottleneck areas and nine skills that people are still needed for each profession, this includes:

  • Perception and Manipulation.

    • Finger dexterity.

      The ability to make precisely coordinated movements of the fingers of one or both hands to grasp, manipulate, or assemble very small objects.

    • Manual dexterity.

      The ability to quickly move your hand, your hand together with your arm, or your two hands to grasp, manipulate, or assemble objects.

    • The need for a cramped work space.

      How often does this job require working in cramped work spaces that requires getting into awkward positions?

  • Creative Intelligence.

    • Originality.

      The ability to come up with unusual or clever ideas about a given topic or situation, or to develop creative ways to solve a problem.

    • Fine arts.

      Knowledge of theory and techniques required to compose, produce, and perform works of music, dance, visual arts, drama, and sculpture.

  • Social Intelligence.

    • Social perceptiveness.

      Being aware of others’ reactions and understanding why they react as they do.

    • Negotiation.

      Bringing others together and trying to reconcile differences.

    • Persuasion.

      Persuading others to change their minds or behavior.

    • Assisting and caring for others.

      Providing personal assistance, medical attention, emotional support, or other personal care to others such as coworkers, customers, or patients.

Source: The future of employment: how susceptible are jobs to computerisation: Table 1.

What this study is basically saying, around 50% of all jobs will be replaced by robots in the next 20 years.

Based on the above study, the BBC assembled a handy guide that calculates which jobs are likely to be automated within the next two decades:

See also: replacedbyrobot.info website.

With this tool, you can check the prediction of over 700 jobs.



What are some jobs that can never be automated?


The key word here is "never". Technology is rapidly advancing, and while I can think of situations where jobs can't be killed in the short-term or even in the long-term, I can't think of a job that is 100%, totally immune to extinction. Surely they exist, but you can't be sure...anything can happen after all. As long as it's possible, that's what matters here. You can't prove a negative.

This whole question seems as foolhardy as predicting in the 1850s that airplanes would never be invented. You'd be right in assuming that airplanes would not be invented in the 1860s...or the 1870s...or even the 1880s...but eventually, airplanes would be invented.

What would be better is to provide a specific cut-off point ("will all jobs be automated by the year 2020?") that can allow us to try to extrapolate and predict based on current trends, but even that starts being difficult as you extend the cut-off point - My predictions about 2020 will be more accurate than my prediction about 2220. I think this type of question is truly unanswerable and can quickly decay to science-fiction speculation.

Some additional comments about Doxosphoi's answer:

Doxosophoi made some arguments for why current society might not accept the automation of all jobs (the need for the "personal touch" that only a human-like intelligence can provide), but that's no reason to assume that society will never accept automation. Technology can change and adapt, and humans can also change and adapt. Maybe a human might not care about a shrink who "rubs its eyes in the morning", dislike movies that are marred by that "vagaries of humanity" instead of personal customization, prefer politicians and soldiers that actually acts logically instead of acting like a falliable human, etc., etc. I mean, it's possible.

There's also the problem of the term "job". Technically, I am working by writing an answer on a StackExchange website, but I'm not getting paid for it, so it's not a real "job"...at best, it's just a hobby. I'm providing a valuable human touch, but since no one is giving me money, it's possible that this human touch may not be all that valuable in the first place: "never give out your labor for free, because then they'll take it for free".

Some of the techno-utopists (which I disagree with heavily) believes in a future where bots handle produce a lot of industrial goods and services, generating a lot of revenue that is then redistributed to the general population via some "Basic Income" scheme. This allows humans to do what they really want instead...such as hobbies? And what if the hobbies of the future are the "jobs" of today: shrinks, movie directors, politicians, soldiers, etc., etc. Instead of working for a paycheck, you're working in these jobs on a volunteer basis.

Obviously, no automation is being necessary to eliminate these hobbyists (no matter how good a bot is, free labor will always prevail), but they're not really jobs, are they? The bot is the one that is producing real value, and subsidizing the hobbies of all these other people. The idea of a "job" itself could be in jeopardy.

I don't think this scenario is likely either (in fact, I'd probably think it's just AI snake oil that will never actually happen). But it's possible and that's why I can't dismiss it outright. It could happen, just that I don't think it will.

And finally, the question is asking about whether a job can be automated, not whether it's a good idea to have it be automated, which is a completely different question. It's possible that we can build machines that can automate everything, and choose as a society not to use them for a variety of different reasons (such as the reasons that Doxosophoi mentioned).

  • 1
    $\begingroup$ I agree with your distinctions. I don't think we disagree here. And if we start stuffing entire earth simulations into computronium the size of a sugar cube, all bets are off. But if the speed of chips tapers off to biological levles of efficiency, we've probably got a good few centuries of employment left in us ;) $\endgroup$
    – Doxosophoi
    Commented Oct 1, 2016 at 5:25

If you were to completely automate a human, you'd just have another human, which defeats the purpose of the automation.

Any job that requires a "whole human," rather than just a human's hands, feet, or simple reasoning ability, will still require humans.

If I go to a shrink, one with Wikipedia-like knowledge would be great, but one that also actually knows what its like to rub its eyes in the morning would be even better. Why? Because solving some problems will require knowing what it is like to rub one's eyes in the morning.

If I go to a movie that was written, directed and produced by some form of automation, I may be able to suspend my disbelief and get carried away by the story, but something in me will fundamentally appreciate the movie less, if I know that the AI can produce an infinite number of these stories, completely arbitrarily. There is something about knowing that the story came from a mind that has been conditioned against the vagaries of humanity (ie, a "whole human"), that makes the story more appreciable.

If I call up a suicide hotline because I wan't someone to sympathize with me about my existential crisis, I'll want to talk to a "whole human" that can sympathize with my existential condition, not one that just regurgitates prior wisdom on life, heuristically matched against my problem state.

If I want to vote for a politician that can sympathize with the needs of the people, I'll want a "whole person" politician that can reflect on all the specifics that make life for a "whole person" hard or easy.

If I want soldiers to take the lives of humans, I want some sort of intelligence in that kill-chain that executes "whole person" analysis prior to pulling the trigger (a human).

If I want a conflict resolution specialist, capable of resolving complex cultural problems between humans, then I don't want just an AI that spits out the most likely solution based on prior solutions. I want an AI that can reason about prior solutions and all explicit and implicit problems between humans, in all human contexts, which requires a human or a perfect human simulacrum.

For any problem that requires consideration potentially across the whole spectrum of human context, we will want that solution to be generated by a "whole human" device. But if we automate the "whole human" then we haven't really outsourced the problem to automation but rather to a "whole automated human," which will, by necessity, have its own problems.

Sure, we'll probably create an artificial human intelligence (AHI) one day, but being optimized to automatically solve any given human problem without also having human problems... that's just AI snake-oil that will never exist - outside of some perverse matrix scenario, under an infinite oracle of some sort.

So, yes, there will be many jobs that still require humans - mostly human-to-human problems that require full knowledge of the human context.

  • 1
    $\begingroup$ The best one was the suicide prevention hotline +1 for that. If a suicidal person figured out that they were talking to a machine instead of a live person they could sympathize, they'd probably be more likely to give in, no matter what logical arguments the A.I. spit out. $\endgroup$ Commented Sep 30, 2016 at 22:55

Probably the only secure jobs are those where the audience enjoys watching live human craftsmanship take place in real time right before their eyes, like acting or standup comedy or musical virtuosity or playing a sport. Watching a robot do the same thing would be far less personally engaging since there's no human skill or artistry to appreciate or identify with, escpecially when the pressure is on or when human interpersonal dynamics are involved.

For example, why would anyone watch robots play poker? Or dance? Or do standup comedy about how hard it is to be [ethnic group or gender goes here]?


Anything that can be broken into set of instructions will be automated, and contained in a narrow trajectory. But we will have the ability to deep between those different skills.

human advantage

Wrote more about this thinking here

  • $\begingroup$ Please don't just provide a link. Instead, post the relevant information here. $\endgroup$
    – Mithical
    Commented Apr 13, 2017 at 11:49
  • $\begingroup$ It's a pretty long essay – not sure if I should be posting it here $\endgroup$
    – Nitzan
    Commented Apr 16, 2017 at 1:23

Truthfully, we don't know exactly how good AI can become, so we don't really know the answer to this question. But I see no reason - in principle - that AI can't become just as "intelligent" as a human, and correspondingly, I see few - if any - jobs that can't be automated.

That said, I suspect that a lot of human thought / behavior / intelligence is wrapped up in how we are embodied and how we experience the world as two legged, upright walking, biological machines with eyes, ears, noses, etc. So I suspect that AI might achieve parity with overall human intelligence, but may not also become capable of behaving like a human, or understanding certain things where the understanding is developed experientially. That may leave an opening for some jobs that require a very specific kind of "humanity", but that's all just speculation.


Not the answer you're looking for? Browse other questions tagged .