Age of AI, Meet the Robot Revolution | American Enterprise Institute

Item: “In the outbound dock of an Amazon warehouse near Nashville, a robotic arm named Cardinal on a recent day stacked packages, Tetris-style, into six-and-a-half-foot-high carts. Then Proteus, an autonomous platform, moved the carts to the loading bay, flashing electronic eyes designed to make the robot more appealing to human colleagues. As robots become more capable, they are performing an increasing number of tasks in warehouses and delivery centers with varying degrees of aptitude and speed. Machines can load and unload trucks. They can place goods on pallets and take them off. Robots can shift items around in inventory, pick up packages and move goods on warehouse floors. And they can do all this without a human minder guiding their every move.” – The New York Times, Nov. 19, 2024

Most of my essays about artificial intelligence feature an AI-generated image of a robot rather than attempting to illustrate sophisticated software. My justification for that switcheroo: Those advanced robots are probably run by AI. Of course, robots are more visually interesting than software, and I’m not above trying to grab your attention through a cool illustration.

That said, AI and robots will be a powerful technological combo that will shape the 21st century. This from the new analysis “The Rise of AI Robots” by a team of Citigroup analysts:

AI is not new (1950s). Robots are not new (1960s). Moving AI-robots are. In the last decade big leaps forward in AI are allowing robots to see, move, talk, learn, and act. … We are entering a new era in which AI-robots and humanoids will be moving all around us. Our analysis suggests there will likely be 1.3bn AI-robots by 2035 and 4bn by 2050.

There are three big reasons underlying Citi’s forecast:

🤖 AI Advances. A convenient confluence of AI breakthroughs (like large language models that can turn verbal commands into robot actions), improved skill (such as MIT’s robotic hand mastering 2,000 objects), and flexible business models (with robot labor costing as little as $2 per hour) is finally making capable, affordable robots a practical reality, according to Citi. A prime example is autonomous vehicles: Waymo already conducts over 150,000 weekly trips across Phoenix, Los Angeles, San Francisco, and Austin, and announced it’s coming to Miami in early 2025. AI improvement will also enhance less whizzy applications like robot vacuum cleaners, which are already in 20 percent of US households.

Most significantly, these advances are making humanoid robots feasible for the first time — this isn’t just hype from Elon Musk about his Teslabots — with capabilities that allow them to operate in human-designed environments and perform may different jobs in homes and industrial settings. Although, according to Citi, “the first use case area will be in the industrial sector, including manufacturing and warehouses” and then augmenting “care in hospitals and care-homes, we see the majority of demand coming from the elderly in their homes. (By the way, a new study of Japanese nursing homes finds robot adoption increases employment and worker retention, enables staff to focus more on high-touch care tasks, and improves both care quality and productivity.) Another primary use case for humanoids in households is help with cleaning: hoovering; using a dishwasher; tidying; with laundry.”

🤖 Economic Benefits. The economic impact appears substantial, particularly for humanoid robots. Citi calculates short payback periods when compared to human labor costs across various wage levels, from minimum-wage workers to skilled professionals like nurses. In cleaning applications alone, the market potential seems massive, with a forecast of 1.2 billion household and 25 million commercial cleaning robots by 2050. For AVs, there’s potential to reduce the 1.4 million annual traffic fatalities while increasing mobility and productivity. The humanoid robot market alone is projected to reach $7 trillion by 2050, according to Citi.

🤖 Adoption Trajectory. Citi sees adoption rates varying by region and application. Less developed regions are expected to account for two-thirds of the humanoid market, despite lower penetration rates, due to their larger populations. In industrial settings, penetration rates are forecast at 30 percent for developed regions, 20 percent for less developed, and 10 percent for least-developed regions. For households, the rates are projected at 22 percent, 11 percent, and five percent, respectively.

(“Given caring and cleaning robots will likely benefit many others, it is possible our penetration rate assumptions prove too low. We also note that we have not analyzed several large potential markets – such as military, safety, security, education, agriculture.” )

And while the US is the overall global AI leader, Asia shows growing dominance in robotics, according to Citi. The bank’s proprietary analysis reveals China’s overwhelming 78 percent share of robotics patents over the past two decades, alongside rising VC investment in Asian robotics companies.

Again, Citi:

Google’s chief economist Hal Varian once said, “a simple way to forecast the future is to look at what rich people have today.” In previous generations this included running water, flushing toilets, cars, washing machines, post-primary education, or music on demand. As we discussed in our AI Assistants report last year, AI can bring personal assistants, coaches, infinitely patient education tutors and health coaches to many. As noted in this report, AI robots can also offer people cleaners, butlers, chauffeurs, and carers. With appropriate cost, capabilities and guardrails, demand for these services could be very significant.

Recent Generative AI advances are pointing at cognitive tasks or creating art on demand. What many people really want is help with mundane tasks, such as cleaning, so they can do more thinking, art, and leisure. AI robots can help with this utopian angle.

Every revolution has its roadblocks

The report also examines a number of potential challenges or downsides to this AI-driven Robot Revolution, including manufacturing costs (high upfront costs with long payback periods), battery power (limited energy density vs growing AI power demands), infrastructure (need for comprehensive charging and maintenance networks), and legal accountability (unclear liability frameworks for autonomous decisions).

Another challenge is trust. A bad start: Citi cites global surveys that find trust in tech companies has fallen from 61 percent to 53 percent between 2019 and 2024, with an even steeper decline in the US, 50 percent to 35 percent. Then there’s AI’s inherent “black box” nature, where decision-making processes lack transparency and full explainability. Interestingly, humans tend to trust robots more when they appear less human-like. So The Jetsons’ Rosey good, Star Trek’s Commander Data bad.

Historically, the bank finds, acceptance and diffusion of new tech often faces setbacks but continues advancing nonetheless. (Nuclear energy in the US would be an exception, I guess.) Early transport innovations like railways and automobiles saw fatal accidents yet achieved widespread adoption. Safety measures evolved gradually — from driving licenses to seat belts — though lots of accidents still occur. “Implementing safety measures for robots that give users more trust and control over the technology is likely to prove critical to their successful integration into society. Like cars, this may be an iterative process that takes time.”

And will robots take all the jobs? Citi contends that although many robotic applications like self-driving cars and home cleaning robots will create more leisure time rather than replace jobs, other sectors will face significant disruption. Manufacturing seems to provide a sobering preview. The bank points to an MIT study that finds each industrial robot replaced 3.3 workers between 1990–2007. With robot costs falling 50 percent since 1990 and wages rising, the economics here are compelling: A $20,000 humanoid working 16-hour days could pay for itself in just 29 weeks at minimum-wage rates. That suggests both job displacement and wage suppression.

But the economic math isn’t quite that simple:

While some roles may be displaced by robots, the adoption of these technologies can also create new job opportunities such as robot maintenance technicians, AI system trainers, and data analysts who can interpret the data generated by robots. To mitigate the impact of job displacement, there is a critical need for targeted reskilling programs that help workers transition from old to new roles. … In addition to taking on tasks that were previously uneconomic, robots have the potential to enhance job quality by taking on more mundane or physically demanding tasks, allowing humans to focus on more complex, human facing, or creative aspects of their work.

As a reality check on the Citi analysis, it’s definitely worth checking the extensive new National Academies study, “Artificial Intelligence and the Future of Work” and what it has to say about robots. There’s quite a bit of agreement with Citi, including looking at how future advances machine learning and multimodal language models could accelerate robotic capabilities. Then there are potential job market challenges, with the study outlining a dramatic scenario where robots become so capable and cheap that they reduce the value of human labor to near zero. Lots of economic challenges around income distribution. In the near term, however, AI and robots are more likely to augment human workers than replace them entirely.

Finally, here are the concluding takeaways from the report: