The “planning” part of city planning traditionally focused on developing strategies, policies and land use patterns that would lead to a future condition that a given community desired. You tell us the future city you want, and we (city planners) will do our best to figure out how to get there.
That sounds straightforward enough, but there were always gremlins lurking beneath the surface. To begin with, the “community” was rarely in agreement about what the ideal future should be like. But more importantly, the strategies, policies and land use plans were based on assumptions about the economy, population change, the environment, and a dozen other factors that were often incorrect. We planned for an idealized future that was frequently sabotaged by inaccurate assumptions, unforeseen events, shifts in public sentiment, or just bad luck. Consequently, decision makers would often ignore city planning efforts – deriding them as just wishful thinking or out of touch with reality.
In recent years, planners have tried to address these shortcomings by incorporating elements of what is known as scenario planning. Instead of focusing on a single desired future based on a single set of assumptions about the global forces that shape communities, scenario planning explores a range of futures by acknowledging that every influential factor has a plus-or-minus deviation from what we think is most likely, and that there are occasionally unforeseen events that cities must be resilient enough to handle. My next several posts are going to be written in the spirit of this new planning approach – exploring a “what if” scenario that might reshape the future of our communities in ways that simple extrapolations of the past might miss.
Each topic is plausible enough to come true, but is certainly not guaranteed. More importantly, each topic – if it did come to fruition – is powerful enough to cause significant ripples through our society and urban forms. They would be – to borrow a common tech term – “disruptive” in that they would force uncomfortable changes. As always, I will focus primarily on midwestern cities, but the actual impact would be much more widespread and so I may touch on more global impacts as well. My goal isn’t to predict the future with any particular level of precision, but rather to encourage the kind of “outside the box” thinking that we often say we need but rarely put into actual practice.
The “What If” Future of Robotic Technology
Robots are already more numerous than most people know. Just in the U.S. alone there are hundreds of thousands of robots deployed in industrial settings such as warehousing and fulfillment centers, vehicle manufacturing, and computer chip fabrication. For safety reasons, however, many of these robots are separated from humans which accounts for the perception that robots are a rarity. Aside from the occasional Roomba robot vacuum, most people don’t encounter any type of autonomous robot outside of movies and television. Unfortunately, C-3PO from Star Wars, the cyborg from Terminator, or Sonny from I, Robot aren’t great sources of information on current robotic technology.
 |
| The Tesla Optimus Robot |
The other fact shaping public perception is that the vast majority of robots are single purpose (or at least limited purpose) machines that can perform tasks with great strength and precision, but which can’t generalize those abilities to other environments or to similar tasks without extensive reprogramming. That, however, is about to change.
What if robots were equipped with multiple sensors that could “see” their surroundings, “feel” surfaces and objects, and “smell” chemicals or spoiled food? What if robots were controlled by artificial intelligence models that could process verbal commands, respond with conversational answers, act with appropriate caution around people and pets, and problem-solve when presented with an unexpected obstacle? What if such robots could be mass produced at prices that ranged from perhaps $10,000 for relatively simple robots to perhaps $100,000 for sophisticated ones?
In many respects, my “what if” speculations are already taking place. The combination of robotic technology and artificial intelligence will undoubtedly happen because entrepreneurs around the globe are solving that problem as you read this post (supported by billions of dollars in venture capital). The real question is how self-sufficient and useful will such robots be in the next five to ten years if they are produced at a price that can create and sustain a mass consumer market? My “what if” scenario assumes that robots will exist with the following parameters:
Sufficient intelligence to understand verbal commands, hold conversations with people, verbally report results, and adapt actions to a reasonable range of environments;
Ability to move within the expected physical environment with human-like agility and speed, and recover from unexpected trips or falls;
Specialized sensors to perceive their surroundings at a level sufficient to move independently, measure environmental factors as needed for their purpose, and interact with a reasonable range of people and objects;
Adequate power supply for several hours of normal work load, plus the ability to autonomously recharge as needed, and then return to work;
Sufficient physical resilience to adapt to different environments including different terrain, different weather conditions, different light levels, and a different landscape of obstacles; and
A price that is low enough to generate mass market demand and which is perceived as being at least somewhat lower than the equivalent cost of human labor.
At present, such a robot does not exist, but current robotic technology is tantalizingly close in many respects. What is unknown is whether that technology will continue to evolve rapidly the way computers and cellphones evolved a few decades ago, or whether the development of the desired abilities will be agonizingly slow like the glacial progress of fully autonomous driving. Oddly enough, Elon Musk is already making astonishing predictions about his Optimus robots which eerily echo his 2016 prediction that an autonomous Tesla vehicle would be able to drive coast to coast without human intervention by the end of 2017. That obviously didn’t happen and is likely years away from happening – an important lesson about the unknowable pace of technological progress.
Where Technology Currently Stands
There are literally dozens of companies scrambling to build an economically viable robot with enough generalized intelligence to be useful. I’m going to describe two examples that are reasonably close to meeting the parameters I described for my what-if scenario, but there are a variety of others that are roughly equivalent. The two I have picked stand out because they have actually been deployed in real work environments for paying customers, while most of the others exist only as prototypes.
Spot by Boston Dynamics. As you might have guessed from the name, the Spot robot has a form factor similar to a dog. It walks on four legs, weighs roughly 75 pounds, and measures about 2 feet tall and four feet long. It has sufficient agility to handle different types of surfaces, and many types of terrain (including stairs), and walks at a normal human pace. It can be equipped with a wide variety of sensors and a robotic arm that can pick up objects, open doors, spray fire suppression chemicals and much more. Spot has been combined with Large Language Models such as ChatGPT so that it can understand verbal commands or read a list of tasks, perform the appropriate actions, and digitally return results.
 |
| Spot by Boston Dynamics |
Spot has been deployed at over 1,000 sites around the globe, often in industrial settings where it can cover a large facility looking for leaks, overheating equipment, security breaches, or other safety anomalies. Spot is essentially a watchman that doesn’t get bored or distracted. It is also being used to assist bomb squads or rescue squads to analyze a dangerous situation before sending in a human responder. Spot fits in spaces that would be difficult for a human and can tolerate toxic chemicals. Finally, Spot is designed to work safely around people within limits.
As impressive as all of this sounds, Spot falls short of my list of parameters in several areas. To begin with, Spot’s level of intelligence doesn’t allow it to have conversations or rise to the level of understanding a new environment or a new set of tasks without a fair amount of custom training or remote direction from a human. Second, Spot is too expensive to be considered a mass market product – the cost of a fully configured unit can reach $300,000. Finally, Spot normally lasts only 90 minutes before needing to be recharged. While it can recharge itself autonomously, that is a significant limitation on the types of tasks it can handle.
Digit by Agility Robotics. Digit is a “humanoid” robot designed primarily for handling materials or bins of products in an industrial setting. The term “humanoid” means that digit is roughly the size and shape of a human (approximately 6 feet tall and weighing 200 pounds), walks like a human (with two legs), and picks objects up like a human (with two arms). It does not look particularly human, however, in contrast to the robots from several other manufacturers. The primary goal of a humanoid robot is to be able to do productive work in an environment designed for humans (e.g. on a factory floor where shelves, conveyor belts, and product bins are designed for human workers).
 |
| Digit by Agility Robotics |
Digit is actively deployed doing real work for real customers such as Amazon, GXO Logistics, and Toyota. It is primarily used to move products or bins of products from one place to another, stacking or sorting as needed. Lifting capacity is approximately 50 pounds, which is the upper limit that OSHA allows for humans. The current model (version 4) is restricted to a “work cell” environment where direct human interaction is not allowed, but the next version (due at the end of 2026) will be able to recognize human workers, calculate their path of movement, and either pause until they have passed or redirect to avoid contact. Digit is designed to work for 4 hours before self-docking at a station to recharge.
Digit can use Artificial Intelligence Large Language Models to understand verbal commands but it does not respond verbally. Instead, it has LED “eyes” that can communicate some status information. In most factory settings, however, Digit’s actions are controlled by Agility Robotics’ fleet management software. The estimated cost to produce Digit is over $200,000 but Agility Robotics uses a robot-as-a-service leasing model instead of direct sales. The company expects to produce version 5 in 2027 at a scale of thousands of units per year which should significantly reduce unit prices.
The fact that Digit can balance on two feet while moving a heavy payload is impressive, but it also falls short of my “what if” parameters in several ways. To begin with, Digit moves much more slowly than a human worker – maybe a third of the speed. Yes, it doesn’t get tired or bored, but it isn’t equivalent to a person. Second, it is not really designed to do general purpose work or hold conversations with people; it was designed to do a limited range of tasks well but no more. Third, it is too expensive to have mass market potential. Costs may come down in the future but they aren’t very close yet.
These two examples don’t quite meet my list of parameters, but they don’t seem that far off. Plus, YouTube is full of videos showing robots dancing or doing karate moves, so my “what if” scenario can’t be that far away, right? Well, maybe, but don’t bet the farm just yet. Most robot designers face three significant roadblocks. First, the more multi-skilled a robot becomes, the more complex it is mechanically. That complexity brings with it problems with fragility and durability. Second, any product that includes artificial intelligence has to deal with the issue of “hallucinations” – a situation in which the artificial intelligence engine fills a gap in its knowledge with a false assumption or a completely made up “fact.” A robot that takes actions based on false information (rather than asking a question before proceeding) could be disastrous.
Finally, a robot may work well in one environment, but scaling to a variety of customers or adding new features requires additional engineering, re-calibration and exception handling. [1] The goal of adding artificial intelligence to a robot (sometime referred to as “embedded AI”) is to enable a robot to “learn” a task in a way that allows that knowledge to be generalized so that the robot can be moved to a different environment or be given a somewhat different task and be able to adapt to that change on its own. So far, no one has reached that goal with a mass-produced product.
The computing model that powers Artificial Intelligence engines such as ChatGPT is known as a Large Language Model (or LLM). LLMs learn to communicate, find information, process ideas and summarize results by digesting millions of articles, books, reports and other documents off the internet or in private databases. In essence, LLMs learn by emulating what humans have already created and there are lots of examples to use.
Roboticists wanting to create embedded AI in their robot have to add two additional elements. The first is vision – the ability to scan the physical environment and understand the objects, equipment, and beings that it contains. This is typically done with a combination of cameras, LIDAR, and other sensors, but that information must be processed into object types that the robot understands and it must be constantly updated as the environment changes. The second element is action – the ability to take instructions, apply those instructions to the environmental objects surrounding the robot, and use its legs, arms, hands, etc. to do something useful. This, of course, involves sending all sorts of commands to various motors, actuators, sensors and other physical elements of the robot to create movement or action of some type. So instead of LLMs, roboticists are building VLAs – Vision-Language-Action models. Unfortunately, VLAs are extremely complicated and there is relatively little existing data to learn from.
Say, for example, you wanted to create a robot that could change a flat tire on a car. You could have the robot watch YouTube videos to get a general sense of how cars work, why a flat tire is bad, and the steps involved in replacing a flat tire with a spare. You could have a robot scan and manipulate all of the elements involved (tire, lug nuts, jack, etc.). You could have the robot watch an expert change a tire in real life. You could have an engineer manually take a robot through the dozens of actions needed to change a tire step by step. You could run computer simulations using different kinds of cars and different types of roadside settings. The problem is that none of that information exists in digital form (aside from the YouTube videos) and even once it is created, the robot would have to perform the action repeatedly (and fail repeatedly) to really learn to change a tire successfully with different types of cars in different situations. Do all of that dozens of times with different tasks and you might eventually build a VLA for a robot to autonomously provide roadside assistance to a stranded motorist. And, by the way, you would have to build in safety protocols so that the robot doesn’t accidentally hit one of the passengers in the head with the lug wrench.
But . . . What If?
So embedding AI in a robot is going to be difficult, but clearly the thousands of people working on the issue think it can be done. If we assume that robots are eventually created that meet my list of parameters, what impact would that have on our society and our cities? I’ll give my take on what I think is most likely by exploring three different scenarios but keep in mind that this type of future could play out in a variety of ways, so my opinion is not necessarily more accurate than anyone else’s. I’m also going to skip warehouses and factories which already use robots extensively and are likely to become even more robot-centric in the future.
To begin with, I think it is likely that even robots with embedded AI are going to be designed for relatively specialized roles rather than be capable of doing everything that a human can do – at least for the foreseeable future. They will have a much broader range of skills than current robots, but there will still be limits. Spot the robot dog, for example, might be able to not only identify a piece of equipment that is leaking oil, but be able to shut down the equipment before more damage is done, reroute the workflow to other equipment, initiate a work order to have the equipment repaired, place warning cones around the leak so that no one slips on the oily floor, and summarize all of these actions in an email to its “supervisor”. But this type of Spot isn’t going to go home with you after work and babysit your kids or paint your house.
While the current obsession with humanoid robots seems to capture a lot of the press attention, I think that there will be a wide variety of form factors. Humanoid robots use lots of computational power and electrical energy just balancing on two legs and mimicking the movement of human arms and hands. That might make sense in some use cases, but in many other situations it might be more efficient for a robot to move on wheels, or to take the form of an eight-legged spider or a four-legged dog. So expect robots with not only different shapes, but also different sensors, arm configurations, and hand capabilities to match their intended function.
This specialization and the limits of production capacity mean that robots won’t suddenly appear everywhere doing everything. They will be phased in over time which means that there will be decisions made about where AI-enable robots will be initially deployed. I can see this playing out in a couple of different ways. The first is that decisions are based on a return-on-investment algorithm. In what roles could a robot replace or substantially increase the productivity of a human worker such that the investment in the robot is rapidly recovered and a profit is made. This approach would focus on identifying work roles where mass produced robots can perform the work satisfactorily and where the cost of the human labor being replaced is sufficiently high to cover the cost. Obviously, as the cost of robots drops and their range of skills increases, the list of possible roles would increase.
The second approach would be to deploy AI embedded robots where it does the most good for society. A frequently cited rule of thumb is that robots should focus on jobs that meet one or more of the three Ds – Dull, Dirty or Dangerous. These are jobs that humans often don’t want and where employers are often looking to reduce liability exposure. Robots are common in warehouses, for example, because the injury rate is double what it is across all private industries. In addition, the work is often very repetitive and demand is expanding faster than the workforce. [2] Mining operations might be an example of roles that are dirty enough to create health hazards for human workers but be fine for robot workers. Industrial jobs involving toxic chemicals, high levels of heat, or strong electrical fields would qualify as dangerous. While this approach has some obvious appeal, it would probably require the type of top-down decision making that isn’t very compatible with our free-enterprise form of economy.
Field Survey/Inventory. Both the public and private sectors tend to be better at building assets to serve their customers than they are at tracking those assets and keeping them in good condition. Doing periodic surveys of asset condition tends to be both boring and time consuming (i.e. expensive). In my experience, cities tend to either use summer interns for this type of work and hope that they are conscientious enough to produce good results, or they hire a consulting firm to do the work but then wait 10 years before they can afford to do it again, or they do nothing until something breaks in a catastrophic way. AI enabled robots would basically allow for a more-or-less continuous process of asset evaluation in ways that would improve both frequency and thoroughness.
The City of Hoboken, New Jersey, recently deployed Daxbot robots to survey the condition of sidewalks and pedestrian ramps. The sophisticated sensors and cameras will provide much more detailed information than the visual surveys that most cities do and it will be more thorough than the common approach of surveying with city staff “as time allows”. [3] These robots have a very specific purpose and will only be in use for a few weeks, but I envision more autonomous robots with more elaborate sensor combinations being a permanent fixture in many Public Works departments. Cities need to inventory not only sidewalk condition, but they need to know whether storm sewer inlets are clogged, curbs are broken, gas lines are leaking, or street signs have been damaged or lost their reflectivity. A truly AI enabled robot could do all of this in a single pass. This type of survey robot could automatically update asset management databases, highlight areas where repair work is most critical, and counteract claims that public improvement programs ignore the poorer parts of town. By the way, the Police Department wouldn’t mind if the robot picked up parking violations at the same time.
 |
| Daxbot Robots in Hoboken, New Jersey |
The private sector has just as much need and probably more economic justification for investing in such a program. Railroads, for example, have a tremendous amount of liability if faulty tracks lead to derailments, broken signals lead to vehicular accidents, or structural bridge defects lead to a catastrophic collapse. Once robots learn to move autonomously without disrupting traffic and do their work without endangering people, pets or property, I think survey robots could become commonplace in a variety of industries. Organizations will tout that the robots help them protect the public in a cost efficient manner, but their real economic justification will be reducing accident payouts and lawsuits related to equipment and infrastructure that was poorly maintained.
Security/Public Safety. Americans were about twice as likely to have been the victim of a violent crime in 1993 than in 2022, and yet in 23 of 27 Gallup surveys since 1993 at least 60 percent of Americans have said there was more crime in the U.S. than the previous year. The perception is that crime is getting worse (particularly in urban areas) but the reality (according to FBI statistics) is that the crime rate has dropped significantly. [4]
There are a variety of reasons for this mismatch between perception and reality, but one of the significant ones is the sensationalized coverage by social media or the local news of violent (but rare) crime incidents. In addition, people tend to be poor judges of environments that are safe versus dangerous, often assuming incorrectly that the presence of graffiti or homeless people in an urban area is a reliable indicator of danger. Still, perception is reality in many ways so we clamor for visible signs that more efforts are being made to protect our safety.
Unfortunately, police departments have a difficult time recruiting new personnel. Roughly ten percent of all police positions are unfilled and that trend has been persistent for some time. Consequently, neighborhoods, business improvement districts, private companies, and many other organizations have turned to private security services to supplement what they perceive as inadequate police protection. More than 1.2 million people are employed as private security guards compared with less than 700,000 sworn police officers. [5] Some have estimated that the private security industry will grow at a compounded annual rate of 7 percent for the next decade. And yet there is relatively little information on how well trained private security personnel are or how effectively they can handle a serious crime incident. If a violent event occurs, exactly how much faith do we have that a poorly paid private cop with a gun but little training will do the right thing?
Enter AI enabled robots with a variety of form factors and sensor packages designed to supplement human security services. Police Departments could also make use of AI robots but the American public has been particularly resistant to such efforts – perhaps we have watched too many sci-fi movies where robot cops go rogue (e.g. RoboCop, Chappie or I, Robot). So for the near term I think AI enabled robots will be primarily a private sector initiative.
Robots with artificial intelligence would have a number of advantages over a typical private security employee. To begin with, they wouldn’t get distracted by their phones or take shortcuts on their rounds when the weather is bad. Second, they could be equipped with sophisticated sensors that exceed the perceptive ability of humans – such as thermal sensors that could detect someone lurking behind the landscaping. 360-degree cameras and Lidar imaging would enable better decision making during an incident and provide more detailed evidence if prosecution resulted. Third, they could use facial recognition software to rapidly distinguish an employee from an intruder, or to pick known “bad actors” out of a crowd of people. Much of this can be done now with fixed cameras and a control room with experienced security personnel, but robots would allow that ability to be mobile which would improve coverage, and add a verbal interface backed by artificial intelligence that would allow questions to be asked, warnings to be issued, and directions given (in multiple languages if needed). Many routine interactions could be handled without human intervention, thus freeing security personnel to focus on real emergency situations.
The boom in data center construction has drawn a lot of interest for a variety of reasons, but one of the largely unexplored impacts is that sprawling buildings full of high-value electronics are being built in many rural areas with minimal police resources. Hence, there will be a rapidly expanding demand for security services and AI enabled robots would be a perfect supplement to human guards. One of the disadvantages of traditional cameras, motion-sensors and similar alarm systems is that they generate a huge number of false alarms – so many, in fact, that most police departments charge a fee for each false alarm response and generally treat alarms as a low priority event. A security robot powered by artificial intelligence could be the first line of response so that false alarms are drastically reduced or completely eliminated, and so that effective action is taken much more quickly.
Construction. The construction industry is facing a bit of a dichotomy. The design side of things is almost entirely digital, with larger building designs including advanced 3D modeling known as a Building Information Model (or BIM). Blueprints are a relic of the past and plans are put in paper form only when absolutely necessary. The physical construction of the building, on the other hand, is only begrudgingly going digital and much of the work is done by hand with techniques that are only marginally more automated than they were 30 years ago. AI enabled robots might finally close that gap, and move several “dirty” and “dangerous” jobs from the human realm to that of robotics.
Landscape architects and civil engineers, for example, go to great lengths to design the ideal topography for large building sites. That design gets translated into reality by surveyors placing stakes at locations across the site with elevation information, followed by large, earth moving equipment that scrapes dirt away in some areas and piles dirt up in other areas. In many cases, a highly precise plan is turned into a landform that only approximates the actual design. In site grading, “close enough” often rules the day. The future, however, is likely to involve autonomous earth moving equipment that is GPS and laser guided so that its path and blade height shapes the dirt into the desired form with much greater precision. Survey stakes will be unnecessary since the AI brain of each vehicle will contain a 3D map of both the existing terrain and the future design. Caterpillar, to reinforce my point, recently announced an entire line of AI enabled grading equipment. [6]
Many years ago, I worked summers in the construction industry to help pay my way through college. At the time, I was surprised at how few workers could actually read the construction plans. Most relied upon the foreman to tell them what to do, assisted by measuring tapes, construction stakes, chalk lines, and a variety of other locational markers. The resulting building was close to the intended design but rarely exactly right. Given the complexity of digital plans and modern building requirements, I doubt things have gotten any better. Enter computer controlled robots such as the FieldPrinter by Dusty Robotics which literally prints the floor plans for new construction, tenant finishes, or remodels on concrete or wooden floors (including architectural notes and symbols). Construction workers just build what they see on the floor. The FloorPrinter, by the way, is accurate to within one-sixteenth of an inch.
 |
| The FloorPrinter By Dusty Robotics |
Whenever I visit New York City, I’m always amazed by how many sidewalks and buildings are covered by scaffolding. The reason is New York’s Facade Inspection and Safety Program which requires tall buildings to have their facade safety-inspected every five years. If the inspection reveals problems that could affect the safety of pedestrians on the sidewalk below, a protective “sidewalk shed” must be installed until the problem is fixed. The city has current permits for roughly 8,500 shed/scaffolding installations and almost 2,000 of those have been in place for more than two years. What is needed – but what doesn’t, to the best of my knowledge, exist yet – is a robot that can “crawl” the surface of a building and use a variety of sensors and cameras to document facade conditions. Ideally, the robot would then weld, paint, re-anchor or tuck point the facade materials as needed to correct the problem.
The Bottom Line
Robots have had a variety of advantages over humans for decades. They can do actions with great strength and precision for hours at a time without getting tired or bored or injured. They can be equipped with sensors that allow them to “see” in the dark, or detect faint vibrations or heat anomalies that humans would miss. Until recently, however, robots needed humans to determine what actions needed to be done in a given situation, to program precisely how each action should be completed, and to find meaning in the data that the robot’s sensors collected.
Adding artificial intelligence expands the potential of robots exponentially. Soon, robots will be able to converse with people, understand generalized instructions, scan the internet or company databases for relevant data, scan the environment around them to become spatially aware of nearby objects, people and equipment, and determine the best actions to take given all of that information. If AI systems like ChatGPT are an existential threat to white collar jobs, then AI enabled robots will be an equivalent threat to blue collar jobs. Regardless of what we think might be best for our society, AI and AI enabled robots are powerful genies that won’t go back into their lamp.
Roughly a year ago, Figure AI demonstrated their humanoid robot folding clothes from a laundry basket under the direction of their Helix VLA model. That robot undoubtedly cost hundreds of thousands of dollars to produce, but I wouldn’t pay even $10,000 for a robot to fold my clothes. But what if it could learn to do all of my laundry and cleaning chores by simply observing me? What if it could share breaking news stories that it selected based on my interests and helped me talk through my reactions so that I had a deeper understanding of what the story might mean to my life? What if it could scan all of the travel and hotel options for my upcoming trip to Paris and give me a list of the best three given my budget and travel preferences? How much would I pay for that robot?
As with all technologies, AI enabled robots will become more useful, more reliable, and less costly as the technology advances and production scales up. The ultimate impact on our lives and cities will be shaped by what we value the most and what tasks we are willing to shift to some form of robot instead of doing ourselves. Under my what-if scenario, I think AI robots would become ubiquitous fairly quickly, to the point where we wouldn’t think twice about seeing a security robot at the bank or a robotic cleaning crew at the office. I think that robots will adapt to our physical environment fairly well, so I don’t expect major changes to buildings or public spaces – other than robots being likely to keep them in better condition than humans do now. But they will be another thing that takes up space, moves somewhat differently than we do, and thus requires some adaptations on our part just as we have adapted to e-scooters and e-bikes encroaching on our streets and sidewalks.
Despite all of the talk about robots helping humans raise everyone’s standard of living by being more productive, there will almost assuredly be winners and losers – especially people who lose their jobs and don’t have the skills to work at whatever new types of jobs are created. In a best case scenario, boomer retirements, labor force participation declines, dropping birth rates and restrictive immigration rules might offset the rise in workplace robots. In other words, people gradually phase out of the workforce and robots gradually take their place.
Reality might be much messier, however, if robots destroy jobs faster than the labor force can adjust. That scenario would increase unemployment rates dramatically and destroy people’s faith in capitalism and current government policies regarding the economy and labor. Those who want to work but have been replaced by AI robots might generate a significant amount of political unrest. The rise of Democratic Socialists and ideas such as universal basic income and wealth taxes might be early harbingers of a significant shift in our reliance on the free market and labor meritocracy. AI enabled robots are likely to be a major boost to overall productivity but those gains might benefit a select few and leave the vast majority worse off in which case labor unrest might reach new highs.
Notes:
Olesya Krindach; “Practical VLA: Why Vision-Language-Action Models Matter for Real Robotics Business”; April 2026; Medium; https://medium.com/@olesyakrindach/practical-vla-why-vision-language-action-models-matter-for-real-robotics-business-8770206fca21
Shaun Edwards; “Robotic Solutions For the Three D’s”; March 2026; Occupational Health & Safety; https://ohsonline.com/articles/2026/03/16/robotic-solutions-for-the-three-ds.aspx?Page=1
Alexa Herrara and Nick Caloway; “Robots are hitting the street of Hoboken, N.J., to improve accessibility for pedestrians”; July 2026; CBS; https://www.cbsnews.com/newyork/news/hoboken-nj-robots-pedestrian-accessibility-data-daxbots/
John Gramlich and Kirsten Eddy; “The link between local news coverage and Americans’ perception of crime”; August 2024; Pew Research Center; https://www.pewresearch.org/short-reads/2024/08/29/the-link-between-local-news-coverage-and-americans-perceptions-of-crime/
Ben Grunwald; “Why private security is on the rise”; November 2025; Duke Law, Duke University; https://law.duke.edu/news/why-private-security-rise
“Caterpillar Unveils the Next Era of Autonomy In Construction”; January 2026; Caterpillar; https://www.caterpillar.com/en/news/corporate-press-releases/h/next-era-autonomy.html
