海角大神

They can walk and talk. How soon will humanlike robots be working beside us?

Humanoid robots walk like people. Carry things like people. Even dance with moves that look human. But can they grow up to do real work on an industrial scale?

That鈥檚 a question the robotics industry is struggling with as it looks to build a blue-collar workforce for the 21st century. Industry optimists point to a surge of interest from industry, faced with a potential labor shortage. Pessimists tick off the numerous challenges facing the makers of general-purpose humanoids, while more specific robot arms, hands, and delivery platforms are already doing more specialized, limited tasks in factories and distribution centers around the world.

The optimism and the skepticism were on display last month at the Boston Robotics Summit, where a handful of humanlike machines, powered by artificial intelligence, vied for attention with the many specialized robots on display at vendors鈥� booths.

鈥淗umanoid robots can be deployed, play a role, and work at the level of reliability that they鈥檙e doing useful work,鈥� Pras Velagapudi, chief technology officer of Agility Robotics, told conferees here at the summit. Its Digit humanoid has been tested in commercial facilities at various companies, including Amazon, an investor in Agility. 鈥淣ow, we鈥檙e expanding that out,鈥� he added.

In December, Agility signed with Mercado Libre, a commerce and fintech company, to use Digit at its fulfillment center in San Antonio. In February, Toyota Canada signed on to pilot the humanoid at its assembly operations, with plans to deploy it if the pilot succeeds.

What鈥檚 the positive outlook for humanoids?

One of the advantages of humanoids is that they are general purpose, able to move around an existing facility and do various jobs, as opposed to more traditional moving robots, which have more limited capabilities and often require a factory to be reconfigured. This flexibility could allow companies to move automation beyond very big assembly lines.

鈥淓xcept on the very few applications that have very large and stable scale, the norm is that almost all jobs are one of a kind,鈥� said Alberto Rodriguez, director of robot behavior for Boston Dynamics鈥� new humanoid, Atlas. 鈥淲hat鈥檚 driving our road map is building the technology that is necessary to solve that general-purpose case.鈥�

The potential for humanoids is huge. Morgan Stanley forecast last year that the market could reach $5 trillion by 2050, with nearly 1 billion such robots in operation. Venture capital funding is flooding into startups around the world building lifelike machines and the components to make them work.

What鈥檚 the pessimistic view?

Even optimists concede that many technical challenges remain. Walking robots need loads of sensors because cameras alone might not see a tripping hazard ahead. They need powerful computers to allow them to move and work in an unstructured environment. The electric power needs are so great that it鈥檚 hard for humanoids to work long periods at a time.

Safety, the ability to work with humans without wounding them or falling on them, is paramount and is not yet proven, especially for machines with legs, robotics experts say. Then, there鈥檚 the cost factor. If traditional robots can accomplish a specific problem at a fraction of the cost, will companies employ humanoids instead?

Eventually, the industry is likely to overcome these obstacles. But a big disagreement is over when that might occur. Serial entrepreneur Elon Musk said in January that he expected his Tesla company would start selling humanoid robots by the . That same month, by contrast, Rodney Brooks, one of the world鈥檚 premier robot scientists, predicted that robot dexterity would still prove 鈥減athetic compared to human hands beyond 2036.鈥� Thus, he predicted, the successful humanoid robots of the future will be more limited in what they can do than what鈥檚 being hyped today.

What is the industry predicting?

So far, the industry is banking on traditional robots. The $54 billion global industrial robotics market will double over the next five years, forecasts Mordor Intelligence. Humanoids represent only about $2 billion to $3 billion of that at present, Barclays Investment Bank calculates. That could surge to $200 billion or more, it adds, but that鈥檚 over a 10-year time frame that is hard to predict.

鈥淭here are some markets for humanoids that make a lot of sense,鈥� , chief technology officer of Titan Robotics, told the conferees here. 鈥淏ut we are in a place where custom-built, purpose-built robots make more sense.鈥�