About two minutes · no cost · no commitment
CorpusMarket Robotics
Robots taking over jobs: what is known and what is guesswork
Talk of robots taking over jobs tends to swing between panic and dismissal. This page keeps to what can be said with some confidence about hands-on work, marks clearly where it turns into reasoning, and takes up the fair question of whether recording work for robot training makes things move faster.
By Linus Näslund, Co-Founder
- No cost to apply
- Nothing shared before we talk
- Reply within 2 business days
Contents06
What has actually been automated so far
Worry about robots taking over jobs is older than the robots themselves, so it helps to start with what has happened.
Machines have taken over a great deal of physical work, and the kind of work is consistent. Welding and painting on car assembly lines. Stacking pallets. Moving totes across warehouse floors. Filling, capping and labeling in bottling plants. Milking cows. Vacuuming floors and mowing lawns.
Two patterns show up in that list.
Tasks, not whole jobs. A machine takes one task. The job it belonged to usually continues with a different mix of duties. A warehouse with robots still has people, and they do less walking and more picking, checking and fixing. Sometimes the number of people falls, sometimes it does not, and that depends on the business as much as the machine.
Structured places first. Every item on the list happens where the surroundings can be arranged to suit the machine: a fenced cell, a flat floor, parts that arrive in the same position, good lighting. The robot does not adapt to the world. The world is adapted to the robot.
Where the surroundings cannot be arranged, automation has moved slowly. That covers most hands-on work outside a factory.
Why hands-on work in unstructured places is the hard case
A service plumber walks into a different basement on every call. The pipe is where a previous owner put it. The shutoff is seized. The space is too tight for the wrench that would normally be used. None of that is unusual, and all of it is hard for a machine.
There are three reasons, and they are technical.
- Robots learn from examples, and the examples do not exist. The models that control new robots are trained on recordings of tasks being done. For most trades there are almost no recordings. How robots learn explains the method.
- Hands are far ahead of grippers. Feeling a thread catch, judging how tight is tight enough and working by touch where you cannot see are routine for a person and mostly unsolved for a robot.
- Every site is different. A robot that works in one kitchen often fails in the next. Coping with variety is the central unsolved problem.
The machines being sold today reflect this. 1X, which sells a home robot called NEO, says that for chores the robot does not know yet, one of its experts can supervise it remotely. The company is describing honestly where the technology stands.
Which tasks look most and least exposed
What follows is reasoning from how the technology works. It is not a forecast, and nobody should plan a career or a business on it alone.
| More exposed | Less exposed |
|---|---|
| Repeats the same way each time | Different every time |
| Done in one fixed, controlled place | Done on other people's sites, indoors and out |
| Rigid objects of known shape | Soft, wet, sticky or irregular materials |
| Light contact, low precision of touch | Force and feel: threading, seating, tensioning |
| Short, with a clear end | Long, with diagnosis and decisions along the way |
| No one else involved | Customers, coworkers and other trades in the mix |
Read as tasks, this suggests that moving, sorting, stacking and simple machine tending are closest. Finding a fault, fitting to what is there, finishing work that has to look right, and anything done in a cramped or changing space are furthest.
Most hands-on jobs contain tasks from both columns. A line cook does repetitive prep and also runs a station during a rush. A warehouse worker moves boxes and also sorts out the damaged, the mislabeled and the missing. The likelier near-term change is that the mix inside a job shifts.
None of this makes any job safe. Technology has surprised people before, in both directions.
Why forecasts differ so much
Search for "what jobs will AI replace" or "jobs AI cannot replace" and you will find confident lists that contradict each other. The reasons are worth knowing.
- They count different things. A study of which tasks a machine could do in theory gives a much larger number than a count of jobs that have gone.
- Capable is not the same as cheaper. A robot that can do a task still has to cost less than the alternative, be maintained, and fit into a working day.
- Software and physical work get mixed together. Progress in AI that writes and analyzes has been fast. Progress in machines that handle objects has been much slower. Many forecasts treat them as one trend.
- Timing is a guess. Small differences in assumed speed produce very different pictures a decade out.
- The forecaster may have an interest. Robot companies, investors and critics all have reasons to describe the future a certain way.
A fair reading is that the direction is clear, the pace is unknown, and precise dates and percentages deserve suspicion.
Does recording your work to train robots speed this up?
People ask this, and they should. CorpusMarket introduces businesses to a programme that pays for recordings of hands-on work, so here are both sides.
The case that it does. Robots improve by learning from recordings of people working. More and better recordings help the models improve. A business that records is adding, in a small way, to that progress. Saying otherwise would not be honest.
The case for proportion. The work labs want recorded is the work robots are furthest from doing, which is why they are asking for it. One business's recordings are a very small share of what these models need, and the footage is being collected with or without any single business. A recording also shows what the hands did. It does not carry the judgment behind the choice.
What is unknown. How much data is enough, how long it will take, and what happens to particular jobs along the way. Nobody has those answers, including the labs.
Who decides. The business decides whether to take part. Each worker decides whether to record, and in the programme we introduce businesses to, recording is voluntary for every worker. Some will see it as being paid for knowledge that would otherwise go unrecorded. Others would not record at any price. Both are reasonable positions.
If you manage people, the question will come up, and it goes better raised early and answered plainly. Our page on recording at work covers how that conversation usually goes, along with filming and privacy.
If you are weighing it up
For a view of one field, see robotics in construction or robots in warehouses. For how the programme runs and what pay depends on, see how it works and get paid to record your work.
If you want to talk it through, doubts included, tell us about your business. It takes two minutes and commits you to nothing. We reply within 2 business days, and nothing is shared outside CorpusMarket before we have spoken with you.
Questions and answers
Frequently asked questions
Will robots take my job?
Nobody can answer that for one person. The pattern so far is that machines take over particular tasks, starting with repetitive ones in controlled settings, and the job changes around them. How far and how fast that goes for hands-on work is not known.
Will AI replace construction workers?
The technology as it stands does not point that way. Robots on job sites do narrow, repetitive tasks on prepared surfaces, and the fitting, adjusting and problem-solving that fill a crew's day are what robots handle worst. That describes today and is not a promise about the future.
What jobs are safe from AI?
No job can honestly be called safe. Work that is varied, done by hand, in places that differ every time and that involves dealing with people looks least exposed with current technology. That is a judgment about how hard the work is for a machine, not a guarantee.
What jobs will AI replace first?
The clearest pattern is tasks, not jobs. The tasks that go first are repetitive, predictable and done in one fixed place, whether that is moving totes in a warehouse or producing routine documents on a computer.
Does recording my work to train robots put my job at risk?
It adds to the data robots learn from, so it would be wrong to say it has no effect. It is also true that the work labs want recorded is the work robots are furthest from doing. Each business and each worker weighs that for themselves, and recording is voluntary.
Why do predictions about robots and jobs disagree so much?
They measure different things. Some count tasks a machine could do in principle, others count jobs that have actually disappeared. They also assume different speeds of progress and different costs, and small changes in those assumptions give very different answers.
More pages
Read next
Employee body cameras and recording at work
Putting a camera on an employee raises fair questions, and they deserve straight answers. This page describes how recording for robot training usually works, what tends to be asked about it, and how those questions are gone through with each business before recording starts. It is a description, not legal advice.
Read moreHow robots learn: imitation learning and diffusion policy explained
Most robots that do something new today learned it by imitation learning, which means copying recorded examples of the task. This guide explains that method and the others around it without the math, and ends with what each one needs in data.
Read moreGet paid to record your work for AI
Robotics labs pay for video of people doing hands-on work, filmed from the worker's own point of view. This page is for businesses whose crews do that work every day. It explains what gets recorded, what the pay depends on and what to settle before anyone puts on a camera.
Read moreNext step
Tell us about your business
If your crew does hands-on work, a robotics lab may pay for video of it. Send your details and we will tell you whether it fits and what happens next.
- No cost to apply
- Nothing shared before we talk
- Reply within 2 business days