Loading summary
A
In March 1982, General Motors announced a rapid and aggressive conversion to Robotics. By 1990, GM wanted 14,000 robots in their factories doing everything from painting to welding to assembly. Nowadays today we dream of robots in the factories doing everything end to end in the dark. Lights out. Guess what? GM dreamed the same 40 years ago and they spent an estimated $60 billion to try to make it reality. In today's video we look at General Motors and their dreams of the automated all robot factory. This video is brought to you by the agenometry Patreon. George C. Deval Jr. Was a self taught American inventor and entrepreneur. In the 1930s he invented a photoelectric control for a garage door which became the basis of an industrial controls company. He then started thinking about magnetic recording and teachable machines. The idea came to record onto magnetic tape a person's actions as they worked a machine tool called a lathe. The recording would let a machine play those movements back like a player piano, enabling mass manufacture of a widget. In the early 1950s, Duval added a hand to this machine to make it a generalized manipulator. To Duvall, this made the device more than just another numerically controlled machine tool. It can be programmed to perform many different tasks given the tools making it quite versatile. In 1954 he files for a patent for this machine. Granted in 1961, it is the first US patent for what we now call a robot arm. But he didn't call that in the filing. He referred to it as universal automation or unimation. Duvall tried to sell the idea. IBM was briefly interested but had went all in on computers. Then Duvall was connected to Joseph Engelberger, chief engineer of the aircraft products division of an old school railroad supply company called Manning Maxwell and more. Engelberger recalled Duvall being unable to restrain himself from spilling the whole dream. That dream being that working robots can get humans out of boring, dirty and dangerous jobs. Most businessmen in those days would turn heel and run upon hearing such a weird dream. But Engelberger was game. An agreement was struck for Manning, Maxwell and Moore. I should just call him Manny Maximo to license Duvall's patents. But then Manny Maximo got acquired and that acquirer didn't care for the whole robot thing. So in 1958, Engelberger got the Consolidated Diesel Electric Company to buy the division and it was renamed to consolidated controls. Three years later, in 1961, the team completed the unimated a hydraulic powered robot arm. It looked like a tank with an arm instead of a gun turret. You first had to teach it, but it can store up to 200 sequential movements in its drum memory. Articles published at the time noted that this made the Unimate more versatile and far less likely than custom designed single purpose machines to become obsolete. In 1962, Condec partnered with the railroad carmaker Pullman to reorganize the division's Unimate related assets and patents into a jointly owned venture called Unimation. Unusually, I saw several different Unimation origin stories. In one, Duvall and Engelberger first met at a cocktail party and started the company. After that, another story has Manny Maximo laying the two off. The story I told you comes from Duvall himself. In a 1981 interview with the magazine Robotics Age, I consider it the closest to what actually happened. In 1960, General Motors bought one of the first unimate Robots for about $18,000, way below its cost of $65,000, and installed it on a factory line in New Jersey. What did it do? There was this machine called a die casting machine, which forces molten metal into a mold under pressure. Upon completion, the shaped metal part, called a casting, needs to be taken out, inspected, and then cooled down in water before moving on down the line. The Unimate did that job. It first positions its fingers on the die casting machine, waiting for the casting to be done. When the machine opens up, the Unimate's fingers reach in, clamp down on the hot, hot casting, and pull it out of the machine. The robot arm then holds the die casting in front of an infrared monitor for inspection. If that checks out, then the robot dips the casting into a water bath. If it does not check out, then the robot notifies a human who has to deal with it. This used to be done by humans with tongs and was not a very well liked gig, which is probably why they wanted to automate it, though I note that the average American worker made about $2.30 an hour in the early 1960s at an $18,000 sticker price. It takes about four years to get your money back. Engelberger went all out to sell the Unimate, writing articles, going to trade shows, and regularly appearing on talk shows on tv. The robot washed windows, poured coffee, and opened curtains. Despite his efforts, by 1965 consolidated controls only got about 30 unimate orders in total. Then in the late 1960s, unimate robots and their competitors were equipped to spot weld. Spot welding is a basic form of welding where two sheets of metal are welded together at one spot without any filler. An average car may need thousands of spot welds it turns out that the unimate robots are well suited for spot welding. The task itself is self contained, boring, repetitive and dangerous. It is also positionally forgiving. The spot is pretty wide and frames and fixtures hold the metal sheet steady for the weld. This is an important requirement since the hydraulically powered, unimated and other robots of the era did not have the finest motor control. Tests in 1968 found that the average human did about 108 spot welds per hour with a 20% rejection rate. The robot did 135 with a rejection rate of 2% or lower. Workers often complained that management was trying to make them into robots. Now here comes a real robot that that was just better at it and could do it all day without complaining and tiring. The robot also doesn't come in late or hungover to work. In 1967, Unimation's parent company, Condec, said that they had about 60 robots in operation in the United States and were selling 10 to 15 each month. Condec president Norm Schlafler said to security analysts at a meeting, five years from now there should be 5,000 unimates in operation and the day will come when entire factories will be robot operated. In the late 1960s, General Motor started to lose ground to foreign imports. West German and Japanese companies like Volkswagen, Datsun and Toyota benefited from labor costs half or fourth of what American workers got. The domestic carmaker struggled to find ways to compete. In 1968, General Motors unveiled their response, their import fighter, later named the Chevrolet Vega. The Vega was a subcompact car designed for good mileage and a low cost of about $1,900. To develop it, GM computerized the design, reuse parts from other car models and drastically reduce the number of parts from 3,500 for a similarly sized car to just 1,200. On the production side, GM invested tens of millions of dollars into automating and computerizing various processes at the Vega plant in Lordstown, Ohio. Most prominently, there were 26 unimate robots to do 95% of the car's 3,900 spot welds. The Vega sold well at the start, but ultimately fell far short of GM's expectations due to faulty design, reliability issues, bad engineering, plus labor strife, particularly a nasty March 1972 strike. It might be tempting to blame the poor robots for this strife, but I think it's a bit more complicated than that. For the most part, GM's labor union, the United Auto Workers, did not seem to complain all that much about the robots. Specifically not that they had much choice. In 1950, the UAW set a precedent of giving up the right to contest production process changes on the shop floor in exchange for wage rise guarantees tied to productivity gains. The real problems were the higher expectations from having those robots to be cost competitive with the imports. GM set a goal for their Lordstown plant to produce 100 Vega cars an hour. Ordinarily, a line did about 55 cars each hour. Previously, the UAW had argued against 70 or even 60 cars an hour, saying that it raised the risk of accumulated defects, worker injury and errors. 100 cars per hour meant workers had to do their tasks in 36 seconds and rather than the minute or so that they had used to have. It is important to note here that management neglected critical points where human effort remained necessary. Yes, unimate robots automated the actual spot welding, but a human still had to clamp the metal sheets onto a frame before the machine can start. So in this case, automation didn't quite make things faster. There were other issues. The workforce was unusually young, average age about 24 years old. They didn't want to become robots and sought ways to assert their humanity, slowing or stopping the line to read the newspaper or smoke. Management tried hard to control the line, worsening relations. The tipping point was when the company folded the factory under one assembly group and eliminated what they saw as duplicate jobs. In March 1972, the Lordstown workers walked off the line for 18 days, costing GM $141 million in production and the workers $11 million in wages. Management accused the workers of sabotage and sloppy production. The union retorted that the Vega was a defective product. They were probably both right, but it didn't help the greater cause of selling more American cars. The Vega suffered several design and component reliability issues, leading to multiple recalls in 1972 to repair engine and rear axle defects. In 1977, GM finally put the Vega line out of its misery. Now, up until then, robots were either hydraulic or pneumatic powered. And doing it this way gave these robots raw strength, which suited them for physical tasks like lifting hot metal. But hydraulics lacked the accuracy for more fine motor tasks, limiting their general usefulness in the factory. Hydraulics also made these robots large, loud and unreliable. Messy, too. Oftentimes, the system leaked, spilling hydraulic fluid onto the floors and making them sticky. People don't like sticky floors. Out in Sweden, ASEA the the iconic company now part of the Swiss firm abb, had seen several factory customers across Europe buy unimate robots. Asea's CEO, Kurt Nikkellen, wanted his company to make a robot too, but felt that the Unimate was too big, loud, noisy and leaky to be practical. Was there a better way? In 1971, Asea assigned two of their top engineers to develop a robot arm from the ground. Up to that team, there was one key question they had to should they go hydraulic, pneumatic or electric? Electric servo motor technologies were then maturing, enabled in part by the emergence of power semiconductors like the thyristor. Asea's team did a bake off between hydraulics and electric, and the latter won easily. To control the thing, ASCA turned to a new category of chips. In 1971, intel released the Intel 8008, the first microprocessor. The 8008 fit well because those electric motors can respond instantly and directly to the electric signals sent by the processor. By comparison with hydraulics, you always have to consider the complexity of the system's fluids intermediating between signal and action. It is where much of the lag and inaccuracy comes from. Moreover, the 8008 can be programmed and reprogrammed without needing to rejigger any wires inside or on the circuit board. This made the device flexible and teachable, like the Unimate was, but even more so. In October 1973, ASEA presented the IRB6. The name stands for Industrial Robot 6 kilogram payload. Priced at 350,000 francs, the system offered high accuracy while also being small, quiet and power efficient. It didn't quite fly off the shelves like ASEA originally wanted, but its two big developments, electric drive motors and microprocessors, forever changed the roles that robots can have in the factory. It opened the door for them to do finer tasks like painting, assembly and arc welding. Arc welding uses an arc, a little lightning bolt, basically to weld one metal to another, often with the help of a filler material. It requires more sophistication and skill than spot welding end it enabled what the Japanese called work cells, where a robot pairs with a numerically controlled machine tool like a lathe. Self contained work cells were seen as the atom of the factory, a building block to automate more complex work. Udomation, for their part, saw the all electric trend coming and made moves to adapt to it. In 1977, Utimation acquired a small company called Vicarm, founded by the robotics pioneer Victor Scheinmann. Scheinman was a Stanford professor who invented the Stanford arm, a small, all electric robot arm with superior accuracy and freedom of motion. GM saw the arm and wanted it for delicate operations like screwing in Light bulbs and assembling brake cylinders. However, Scheinmann was unsure whether his small operation can supply such a giant, so he sold out to Unimation. In 1979, GM and Unimation together unveiled what they dubbed the Programmable universal machine for assembly or the Puma. Puma was small, about 92kg and it was not all that strong. It starts losing precision with loads of about 2.2 kg, which does not seem like a lot. But GM's internal surveys found that 90 to 95% of the parts handled on the assembly line weighed less than that. The arm quickly gained hype as an autonomous solution for labor cost issues. Easily reprogrammable, GM envisioned the robots as being interchangeable with any human on the line. And at $25,000 with an eight year working life, it would cost almost $10 per hour, less than a human. $4.80 compared to $14. GM's manager of manufacturing development, Richard Beaker said Puma is not truly a universal assembly machine, but but it comes close to being a universal system when applied to relatively small products. Looking to challenge the Japanese, GM in late 1979 planned out a massive $40 billion six year capital expansion plan. The car brands would be redesigned and the factories basically rebuilt from the ground up. With automation in mind, GM unveiled this robot led plan as it thrashed through one of its most serious crises. Things really got going in early 1979 with the Iranian revolution. Oil exports from Iran stopped overnight, sparking panic in the markets and amongst ordinary people. Gas prices spiked to record levels, leading to rationing and long lines. The American carmakers were caught flat footed. They had gone all in on large gas guzzlers. When gasoline prices abruptly soared. Consumers flocked to smaller, more fuel efficient Japanese cars. The net result was the American domestic car industry suffering two straight years of drastically falling sales. In 1980, they sold just 6.6 million cars, their worst number since 1961. While the imports surged to capture 26.5% of the market. GM, Ford and Chrysler were devastated. In late 1979, they idled nearly 100,000 workers. Idling soon turned into factory closures, with tens of thousands of layoffs across the American Midwest. Layoffs at the carmakers quickly spread to both their dealers and suppliers. In total, the US government estimated that nearly 800,000 jobs were affected. Ford lost $3.3 billion between 1979 and 1982, including 1.5 billion in just 1980. At the time, it was the largest single year loss in American corporate history. They didn't hold that record for long. A week later Chrysler announced a gobsmacking $1.7 billion loss for that same year. They had nearly collapsed into bankruptcy necessitating a $400 million loan guarantee bailout from the US government in January 1980. For their part, General Motors announced an annual loss of $763 million in 1980, their first unprofitable year since 1921. GM had reigned as America's most profitable corporation for decades. This was a huge loss of face. The next year GM announced a new chairman, the 10th in its history. The 55 year old Roger Bonham Smith. A quiet man described as a good cash and capital manager. Smith joined GM as an accounting clerk in 1949, back when Alfred Sloan still ran things. He then worked his way up through the public relations and financial divisions. Ironically enough, Smith was definitely unready for the public facing side of leading America's largest company. He was horrifically gaffe prone. Like voting for executive bonuses right after layoffs. Cutting his half a million dollar salary by just $135 a week as solidarity and saying things like Every time labor goes up $1 an hour, 1,000 more robots become economical. But Smith did recognize that GM needed shaking up and that is what he went and did. Some of the changes were organizational. For instance the unpopular 1984 reorganization which removed autonomy from GMs various brands. This led to a corporate led strategy known as platform sharing where different car brands share common design and production. Car enthusiasts really hate it. They say it strips away each brand's soul and uniqueness. Other changes were financial major cuts in wages and benefits as well as closing down several factories. Such moves are probably why Smith even today is particularly disliked by rank and file GM workers. Though most relevant to our story, Smith publicly acknowledged Japanese manufacturing superiority and sought to prepare the company for that reality. For instance, he did the NUMMI joint venture with Toyota to try and learn their production systems. And he pushed ahead into factory automation and technology solutions to reclaim GM's manufacturing mojo. Allocating billions of dollars to do was a weird time for the domestic carmakers. The early 1980s downturn kicked off much reflection on on what the future held for gm, Ford and Chrysler. Some in the industry believed that consumers gas price induced shift towards smaller, lighter and more fuel efficient cars would be permanent. If so, then the Americans were hosed. How can the Americans compete with the Japanese? Considering what appeared to be a massive disparity in wages, GM and the Americans looked to the robots. The Japanese led the world in industrial robot adoption far higher than that in the United States. I want to first caution you about these numbers because the definition of a robot differs between counters. But in 1974, Japan had a thousand robots and the United States 1,200. By 1980, Japan had 14,250 industrial robots. America, by comparison, trailed far behind with just 3,400. By 1982, there would be over 30,000 Japanese robots. Another report from June 1981 from the National Labor Organization said that Japan had 58,000 robots, accounting for 80% of the world robot population. If you read the news at the time, this robot gap granted the Japanese superpowers. In early 1980, the Washington Post reported that Nissan had a plant in zama turning out 1,300 Datsun cars a day. With just 67 employees, Robot supposedly did 97% of all the work. That article was later debunked. Turns out the Nissan Zama factory actually had 2,000 workers, not 67. US managers actually visiting Japanese factories did not see the fully robotic cyberpunk utopia they had expected. But anyway, if you can't beat them, join em. GM concluded that it cannot close its tremendous cost gap with Japan unless they produce this technology enriched plant capable of banging out cars as cheaply as even the Japanese can. Such a vision cannot be fulfilled without smaller all electric robots. Fortunately, Unimation had one ready to go. The aforementioned Puma. Perfect timing, right? But Unimation struggled with scaling up production. GM publicly promised monthly production of 60 Puma robots by the end of 1979. They failed to hit that. Yes, scaling up is hard. Building a small software guided robot requires a whole different skill set than making a big heavy hydraulic Gundam. GM complained a lot about the Puma's build quality and design, but it can be done. The core problem seemed more to be that Engelberger and his company, for whatever reason, could not move on from the big hydraulic robots. Management believed the hydraulics was where the real money was. And that wasn't going to change anytime soon. In 1980, Victor Scheinman resigned from Unimation, frustrated by the lack of progress and chronic underinvestment in the Puma robot line. Without Victor there, relations between Unimation and GM rapidly went downhill. When GM insisted on electric robots, Engelberger went to Smith himself and insisted on on the importance of hydraulic driven robots. In the end, GM dropped Unimation and the Puma from the robotics plans. A major loss considering GM wanted to buy 9,000 Pumas for its factories. The heavily indebted Unimation was forced to go IPO in 1981 to pay back its debt, and was then quickly bought up for $102 million by the declining electric giant Westinghouse Electric. What is that quote Steve Jobs once said? A players hire A players, but B players hire C players. Joining the Westinghouse Clown show and its flurry of strategic moves caused all of unimation's best talents to stampede out the door. Engelberger himself resigned in 1984, supposedly in tears. Thus marks a robot pioneer's decline into irrelevance. So who did General Motors decide to partner with? Instead, they chose an obscure but rising robotics player, the Japanese numerical control giant Fanuc. Fanuc is one of Japan's most technologically important companies. We discussed Fanuc's founding and story in a prior video. Founded in the 1950s as a subsidiary of Fujitsu, they became global leaders in a niche but vital product category called computer numerical control or CNC controller modules. Fanuc dominated this CNC controller market, but was nowhere near as big in industrial robots. Just 2.5% of Fanuc's 1979 revenue came from robots ranked by sales. In Japan, Fanuc was a distant sixth behind leaders like Hitachi and Kawasaki, and in terms of U.S. sales numbers, they badly trailed Cincinnati, Millikron and DeVilbiss. So why Fanuc? Well, to start, someone in sixth place is way more willing to accommodate you than someone in first place. They also had good technology. The their all electric models 2 and 1, powered by Fanuc's own CNC controllers, were generally well received, though did not sell in numbers. Fanuc also went all in on the vision of the all robot fully automated Factory. Founder and CEO Dr. Suimon Inaba just personally really liked robots almost as much as he liked the color yellow. In Japan they nicknamed him the Emperor of Robots. Some companies might sell more robots, but Fanuc made their robots do more. By the end of 1979, Fanuc had 15 robots in his DC Servo motor factory, loading and unloading components, doing the work of 13 humans for many hours on end. In 1981, Fanuc built and moved into a large factory in a forest at the base of Mount Fuji, later named the Fanuc Forest. They turned this factory into a showroom of what robots can do. The factory later gained substantial attention in the American press for having robots make robots or being lights off. So I am bemused by people being impressed again by articles today of dark factories in the People's Republic of China. History does indeed repeat itself. Again, the robot's capabilities were somewhat exaggerated almost to the level of Fabrication. Inaba later said that he was a bit embarrassed to hear claims that Fanuc robots were autonomously building robots. Critical aspects like robot and CNC module assembly were still done manually, but the robots did improve productivity. Flexible manufacturing work cells allowed Fanuc to raise their productivity by some 1.5 to 5 times, depending on the metric. Fanuc workers did less overtime, getting to turn off the lights and go home. In the end, I think General Motors wanted to partner with someone with good technology who shared the vision and was willing to tailor the roadmap to what GM wanted. Fanuc fit the bill. GM, Fanuc Robotics or just GMF, launched in mid 1982 with $5 million from each party. So a straight 5050 equity share. The joint venture was based in the United States and all its top management, including its CEO, a former top executive from GM's overseas group named Eric Mittelstadt, were in the United States. The manufacturing was initially done in Japan while a factory was built. GMF Robotics product lineup had contributions from both sides. General Motors contributed a numerically controlled painter robot. Fanuc brought several medium sized robots. Informed by GM's demands, GMF quickly expanded their lineup from there. Boosted by General Motors purchases. GMF Robotics sales skyrocketed. In 1983, its first full year of operations, they sold 332 robots. The company rapidly scaled its employee count from 60 to 20, 275, including 100 engineers and 100 technicians. In 1984, sales surged to 1,200 robots and over $100 million in sales, making it America's largest robot company with 26% market share of the US industry. Second place, Cincinnati Millikron had half as much market share. A few competitors complained that GMF's success was entirely because of its cozy ties to GM. One they have the inside track at GM because most of the GMF people came from GM and can get audiences with their old buddies. Meanwhile, in 1983, GM's fortunes turned in a good way. Political pressure in wake of the 1979 oil crisis and the domestic automaker crisis led to the Japanese voluntarily restraining the number of cars they exported to the United States. With the Japanese still in the process of setting up their American factories, gm, Ford and Chrysler got badly needed breathing room. Additionally, gas prices plunged in 1980 as an oil glut developed in the market. This revitalized the American consumer's thirst for big cars, causing the Detroit carmaker sales to surge in 1983 and 1984. GM in particular benefited on both the top and bottom lines. Revenues recovered and surged to new heights. $84 billion in 1984, which would adjust to $264 billion today. And since GM had substantially cut costs, I.e. layoffs and factory closures during the hard times of the prior two years, profits surged to record levels. 3.7 billion in 1983 and then 4.5 billion in 1984. One might imagine that low gasoline prices might lead GM to scrap their plans for making a cheap, small car to counteract the Japanese. But either the company felt that the low gas price situation was temporary or they lacked proper perspective. Thus, in late 1983, Chairman Smith publicly announced Project Saturn. Stop me if you've heard this one already, but General Motors conceives Saturn as its response to Japanese imports in the subcompact market. A smaller, cheaper car with good mileage design, with fewer simpler parts. The larger story of Saturn is beyond the scope of this video, which is already spiraling out of control. But high technology and robotics featured heavily in the manufacturing strategy. GM envisioned a computerized factory. Robots from GMF will work in concert with tools and other equipment. A central computer can detect and respond to assembly line problems in real time. Such an automated factory would produce Saturns at costs that even Japan cannot match. Saturn was scheduled to come out in 1990. Roger Smith regularly acknowledged that its success depended on technology the that did not yet exist. In an April 1985 interview he said, we have confidence in our people, we have confidence in our suppliers, we have confidence in our research organizations and I am positive we don't have vehicles on the drawing board today that we feel uncertain about the technology being available. I do wonder though, if Smith was so confident in gm, why did he feel so compelled to go and buy computer and systems engineering competence? In June 1984, GM spent $2.5 billion to acquire Ross Perot's computer services company Electronic Data Systems. It was GM's largest non automotive acquisition thus far and the biggest electronics related purchase in corporate history. The two companies immediately started working together closely to overhaul GM's systems. GM augmented the EDS purchase with minority stakes in various computer vision and artificial intelligence companies. Notably, they paid $3 million in mid-1984 for a small stake in a Silicon Valley startup called Technolog, which specialized in the AI technique of the day, expert systems. Today we call these rule based AI. They tried to codify human knowledge into a series of rules in order to emulate human decision making for complex questions. GM said in its annual report that it wanted to use expert systems to help dealers diagnose problems to cut maintenance costs. In 1984 alone, GM invested in five different computer vision companies, one of which was View Engineering, a commercial machine vision pioneer. They had a measurement machine that scanned an object in three dimensions using vision in order to understand its features. It is pretty obvious why GM invested so much into so many machine vision startups. They identified 44,000 potential robot opportunities in the factory, including 12,000 just in assembly. None of those work if the robots cannot see and inspect what they are picking up. GMF CEO Mittelstadt himself was on the record saying back in 1983 that if a robot doesn't have vision by 1990, it will not be called the robot anyway. General Motors capped off this wild spending spree in 1985 when they beat out Ford and Boeing to acquire the defense and electronics technology company hughes aircraft for $5.2 billion in cash and stock. The strategy behind this was that Hughes had a lot of special electronics and defense technologies, and GM wanted to apply that to making cars, I guess. In late 1985, General Motors opened the Detroit Hamtramck assembly plant, their most advanced plant and the forerunner to Saturn. Hamtramck opened with 260 GMF robots, the most of any GM plant, to augment 5,000 human workers. A GM spokesman said that the technology moves so fast that that since construction began in 1980, significant parts of the plant had to be revamped six times. The plant had all sorts of technology goodies. 60 automated rovers to carry auto parts around the factory, following little wires installed into the floors. Barcodes and computer chips were set up to track inventory throughout the whole factory. All the information is stored in a central computer set up by EDs notifying workers when new shipments have arrived. Seeing robots, I'm guessing like the ones I mentioned from View Engineering, were set up to guide other robots as those robots install doors, windshields, rear windows, sealant around the window frames, rear axles and the tires. And in the paint shop, helper robots are programmed to open and close doors so that other spray painting robots can go inside the car to paint interiors, Chairman Smith said in a speech during the opening ceremony. Automaking is on the threshold of a technological revolution that will dwarf all our past accomplishments. Inspired and almost miraculous inventions and innovations are ready to usher in a new and golden age of opportunity. The Hamtramck plant was expected to produce about 60 new cars per hour. A year later, it could only do about 35 per hour, far short of expectations. Much of this was blamed on technology problems. The automated rovers meant to carry parts around the factory failed right off the bat and it took several months to fix the computer controlled spray painting robots went haywire. On occasion, they started painting each other rather than the cars, leading to several hundred vehicles being so badly painted that GM had to ship them to a nearby Cadillac plant for a redo. Another GM plant in the city of Flint was installed with computer vision enabled robots to align windshields onto Buick cars. Unfortunately, the robots failed to properly perceive physical depth and shattered the windshields. Yet another painful learning lesson was that robots can work fine on their own, but f up real bad when networked within larger systems. This explains why the Japanese had work cells. If one fails, you can isolate it. An example, each car in the factory is tagged with a transponder to identify it and its attributes, so the central computer network can know what program to run at each station, thereby making it very flexible. This automatic vehicle identification system, however, was too slow to keep up with the actual assembly line in the real world, thus creating delays where cars, Rover, AGVs and robots all waited around for the right command, doing nothing. Robots might not come in late to work, but they can be just as good as humans at sitting around. Slacking speed was a consistent issue. Even when things did work, they often worked too slowly. A very good example of this was Consight, a vision based automatic inspection technology that GM started working on back in 1979. It focused two light sources on items passing by on a conveyor belt. When it saw the right item, then a robot can pick it up and transfer it to a work area. Yeah, sure it worked, but it was expensive and slow. Here's one more. How many robots does it take to screw in a light bulb? They did a bake off between a robot and a person. The problem was that the robot, a puma in this case, can't feel the bulb as it screws it in. So to keep from smashing the thing, it moves very slow. The human meat bot might cost more, but is far more productive. When asked about these, General Motors played it off as early teething issues, saying that such issues were to be expected for any new factory. A spokesman said Hamtramck was a very aggressive application of new technology. Maybe a little too aggressive. Some people have looked at it as a disaster startup and it wasn't. Basically, it's been a little slower but satisfactory, given all the complex elements involved. As I said earlier, Saturn and GM's robotic revolution all depended on the technology working out. The 1986 experience showed that that it clearly wasn't. Despite a million man hours of training, and refocusing, Hamtramck had automated only 5% of total assembly operations. Smith told the press that things were on the right track. And to their credit, Hamtramck did eventually become an excellent plant. But in that moment, General Motors had bitten off more than they can chew. On a larger macro scale, GM was feeling pressure. While GM was blowing what Businessweek estimated to be $60 billion on technology, and Capex, Ford and Chrysler, for sheer lack of capital, just focused on cutting costs. So when the market turned, they were wonderfully positioned. In 1985, Ford launched the Taurus, a car that revitalized the entire company and saved it from bankruptcy. And Chrysler, well, they cut deeply enough to finally turn a profit. So after all that spent on buying companies, robots and Exotic Vision Technologies, GM's ROI was a decline in domestic market share from 48% in 1978 to 37% in 1986, plus a bunch of very expensive new plants that were hardly more efficient than the old stuff. There were other serious concerns. Part of GM's business model, set by the legend Alfred Sloan himself, was that their nameplate brands must be distinct by vision. By the mid-1980s, the platform sharing had gotten to a point where every car looked the same. A rollback was necessary. In 1986, General Motors postponed Project Saturn. The majority of the project's $3.5 billion spend was postponed. High volume production, meaning 500,000 cars or more, was pushed back from 1990 to 1992. As a result, GM canceled $88 million of robot orders from GMF. Considering that the automaker represented 70% of the joint venture's orders, the company had no choice but lay off 200 of its 700 employees. Smith retired in 1990. He can point to big profits throughout his tenure, but his legacy is that of decline. The company's fall began and continued throughout his watch. In the early 1990s, a deep, sharp recession due to the Gulf War caused GM to once again turn record breaking losses $4.45 billion in 1991. They would also close over 20 factories and cut 74,000 jobs in 1992. Smith's successor as chairman finally rolled back the 1984 reorganization. They took a financial bath that year, turning out a $23.5 billion loss due to charges related to retiree health benefits. As part of this reorganization, GM decided that it no longer wanted to be in the robotics business. They sold off their share of the partnership to Fanuc. Several years later, Fanuc merged that into their larger North American structure, Fanuc America. The partnership with GM lifted the Japanese CNC giant up from the cellar, leaving it as one of the top robot makers in the world. Pre China, that is General Motors. Lesson is a classic one. Trying to fix problems with automation often just gives you automated problems. Lots of people point to nummi, the GM Toyota joint venture, as a counterexample. When it opened, it had a few robots, mostly for welding the car and painting the outsides of the car, but nothing hyper techy like Hamtramck. No computerized inventory system, no fancy rovers, no robots open doors for other robots. Toyota, however, successfully translated their principles surrounding continuous process improvement to turn NUMMI into one of GM's most productive plants. It took some time, but GM eventually adopted several of those principles, including the one where where automation is deployed not to replace operators, but augment them. At least when it comes to cars, the dreams of totally automated production in the lights out factory remain as such, just dreams. Alright everyone, that's it for tonight. Thanks for watching. Subscribe to the channel, sign up for the Patreon and I'll see you guys next time.
Host: Jon Y
Date: May 10, 2026
In this episode, Jon Y explores the ambitious—yet ultimately troubled—history of General Motors’ (GM) drive toward fully automated “lights out” factories powered by robotics. Tracing the journey from early robot arms in the 1950s to GM’s massive investments in the 1980s, Jon examines technological milestones, corporate decisions, unexpected pitfalls, and the human consequences of GM’s $60 billion robot dream. He draws lessons for today’s automation efforts and reflects on what GM’s quest says about the intersection of technology, management, and manufacturing reality.
George Devol’s Vision:
Unimate’s Early Challenges:
"That dream being that working robots can get humans out of boring, dirty and dangerous jobs." — Jon Y, paraphrasing Engelberger and Devol [03:06]
Spot Welding Automation:
Lordstown’s Automated Revolution:
"It is important to note here that management neglected critical points where human effort remained necessary. Yes, Unimate robots automated the actual spot welding, but a human still had to clamp the metal sheets onto a frame before the machine can start." — Jon Y [14:15]
Technological Innovation in Robotics:
The Puma & Early Robot Arms:
"Puma is not truly a universal assembly machine, but...it comes close to being a universal system when applied to relatively small products." — Richard Beaker, GM’s manager of manufacturing development [23:12]
Oil Crisis & Market Collapse:
America vs. Japan—The Robot Race:
"Every time labor goes up $1 an hour, 1,000 more robots become economical." — Chairman Roger B. Smith [24:22]
Failure with Unimation:
Turning to Fanuc:
"Trying to fix problems with automation often just gives you automated problems." — Jon Y [46:48]
Hamtramck & Project Saturn:
Technical Failures:
"Robots might not come in late to work, but they can be just as good as humans at sitting around. Slacking." — Jon Y [42:48]
"How many robots does it take to screw in a lightbulb? ...the robot...moves very slow. The human meat bot might cost more, but is far more productive." — Jon Y [43:03]
Shortfalls and Retrenchment:
End of GM’s Robotics Dream:
"At least when it comes to cars, the dreams of totally automated production in the lights out factory remain as such, just dreams." — Jon Y [48:20]
On Selling Robot Dreams:
"Most businessmen in those days would turn heel and run upon hearing such a weird dream. But Engelberger was game." [03:21]
On Management Blind Spots:
"Management neglected critical points where human effort remained necessary...In this case, automation didn't quite make things faster." [14:15]
On Tech and Reality:
"If a robot doesn't have vision by 1990, it will not be called the robot anyway." — Eric Mittelstadt [38:00]
On the True Cost of Automation:
"Trying to fix problems with automation often just gives you automated problems." [46:48]
Jon Y’s exploration demonstrates that GM’s pursuit of total robotic automation, while visionary, ran aground on technical, managerial, and organizational shoals. The real lesson is that “trying to fix problems with automation often just gives you automated problems.” Success came not from lights-out factories, but from thoughtful integration of automation to support, rather than replace, skilled human workers—embodied most in the lessons GM learned through their later partnership with Toyota at NUMMI.