The hundred-year conversation between people and machines

October 2, 2026 · 5 min read

The word robot is a hundred and six years old and the first industrial one went to work sixty-five years ago. Every era since has gotten something right that the next one kept, and something wrong that the next one fixed. Here is the whole conversation in six scenes, ending in the one you can have tonight for free.

The short answer

People have been arguing with machines about work since 1920, when a Czech playwright gave us the word, and building the machines since 1961, when the first one lifted hot castings in a New Jersey plant. The arc is not from weak to strong. The first robots were plenty strong. The arc is from machines that did one brutal thing behind a fence toward machines you can teach by hand, that find their own way across a floor, and, as of now, that give an honest answer about whether they belong in your plant at all. Each scene below is a real turn in that conversation.

1920: a word

Karel Čapek's play R.U.R. opened in Prague and gave the world a word borrowed from the Czech for forced labor. His robots were artificial workers who did everything, resented it, and ended the human race in the third act. The play is remembered for the word and forgotten for the argument, which was about dignity, not machinery. A hundred years on, the dignity argument is the one that still matters on a factory floor, and it is why this site keeps saying that the operator is the most important part of the cell.

1961: an arm

George Devol filed the patent in 1954 and Joseph Engelberger built the company, and in 1961 the first Unimate went to work at a General Motors plant in Ewing Township, New Jersey, pulling hot die castings out of a machine and stacking them. It was hydraulic, it weighed as much as a car, it did one job, and it did that job without getting burned. That is the whole thesis of industrial robotics in one sentence, and it has not changed. The arm takes the job that hurts.

The 1980s: the fence

By the eighties the arms were electric, fast and everywhere in automotive, welding bodies and painting them on lines a quarter mile long. They were also dangerous, because a machine that moves a hundred kilograms at speed does not care who is standing there, so the era's great contribution was the fence, the light curtain and the risk assessment. The standards that govern a cell today, R15.06 and ISO 10218, grew out of that decade's accidents. A buyer who thinks the paperwork is bureaucracy is arguing with people who learned otherwise.

The fence also kept robots out of small plants. A fenced cell with a programmer and a controls engineer paid back on a part that ran for years. A forty person shop with a hundred part numbers had nothing to put in it.

2008: the arm you could touch

In 2008 a Danish company sold its first collaborative arm, a small, light robot with power and force limiting that could, after a risk assessment, work beside a person without a fence, and that a machinist could program by moving it with their hands. Several makers followed and the matrix on this site now lists hundreds of them. The cobot did not make robots smarter. It made the setup cheap enough that a twenty piece batch could justify one, and it put the arm on a cart so one robot could serve several machines. That is why machine tending is the first robot in most small shops today and was almost nobody's first robot in 2005.

The 2010s: the robot that found its own way

Mobile robots used to follow tape and stop for anything in the way. In the 2010s lidar got cheap, mapping software matured, and a vehicle could learn a building once and then find its own way through it, around people, carts and the pallet somebody left in the aisle. The same ideas went into floor scrubbers that clean a terminal at three in the morning and hospital tuggers that take case carts to the operating rooms. The casebook on this site has the published accounts: two robots on bus routes through a New Jersey plant, a scrubber on 200,000 square feet a week at a Cincinnati airport, a molder whose robot loops the floor every half hour. Navigation turned the robot from a fixture into a colleague.

2026: the first honest answer

Every era above made the machine better. None of them changed the first conversation, which for a small plant was always with someone paid to sell one brand, or with a feasibility study it could not afford. What has changed in the last few years is that the first honest answer got cheap. A plant manager can describe a station in plain language at eleven at night and get a serious preliminary read: whether the physics work, what class of machine applies, what the manufacturer-published figures say, what it plausibly costs, and whether a fixture would do the job instead. On this site that conversation is with Joe, who searches a matrix of hundreds of machines, ranks on fit, discloses when a machine is a line we sell, and says no about as often as yes.

What each era got right

The sixties were right that the arm should take the job that hurts. The eighties were right that a moving machine needs a fence or an assessment, and that somebody has to sign. The cobot era was right that setup cost, not capability, is what kept robots out of small plants. The navigation era was right that the robot should come to the work. And the present is right, we think, that the most valuable thing to automate first is the honest answer, because every other era's machine was sold to a plant that could not afford to find out whether it needed one.

Čapek's robots rebelled because nobody asked them anything. Ours answer questions all day. Ask Joe about the station you have been wondering about, and see which era's lesson applies.

Related on this site

Every era of robotics kept what worked and fixed what did not: strength, then precision, then safety, then touch, then navigation, then the honest first answer. The last one is the only one a small plant was ever offered.

Put it to the test on your own job

Tell Joe what you are running and he will apply this to your numbers, including telling you when a robot is the wrong answer.

More notes