Hello and welcome to the August special edition of Eye on A.I.
The marriage of humanoid robots and artificial intelligence is a source of wonder and worries: Wonder at the prospect of the sci-fi-like machines and worries about the potential job losses (or worse).
It’s also an idea that has captivated some of the tech industry’s most powerful minds, from Amazon founder Jeff Bezos, who hosts an annual robotics conference, to Tesla CEO Elon Musk, whose company is developing a robot called Optimus that Musk claims could one day outnumber humans.
We’re still a long way from Musk’s prediction. But this week, a new droid made its public debut, taking us one step closer to the robotic age.
Apollo, as the new robot is named, is roughly human-sized, standing 5 foot 8 inches tall and weighing 160 pounds—and it’s designed to feel approachable and human-like. Its creator, Austin-based Apptronik, is positioning Apollo as a versatile robot suited for both working alongside humans (the company hopes interacting with Apollo will be akin to interacting with a friendly coworker) and taking over for them in dangerous work environments where humans would rather not be.
While initially focused on case and tote handling solutions—picking up stuff and moving it—in the logistics and manufacturing industries, Apollo is more broadly an all-purpose robotics platform available for developers to configure for different industries, such as construction, oil and gas, electronics production, elder care, retail, and home delivery. Research and development for Apollo came out of the DARPA Robotics Challenge, where two of the company’s cofounders worked with NASA Johnson Space Center on the Calgary robot. And then in 2016, the company spun out of the University of Texas at Austin’s Human Centered Robotics Lab to become Apptronik.
“Over the past seven years as a company, we’ve built over 13 unique iterations of different types of mobile robots, eight iterations of humanoid robots—each very different than the others and many times building many versions of each one of those iterations—two different full body exoskeletons or Iron Man suits, and then three mobile robotic arms. And basically, we’ve taken everything we’ve learned from each of those systems to build one of the first truly commercial humanoid robots for general purpose robots,” Apptronik CEO Jeff Cardenas told Eye on A.I.
Cardenas breaks down the role of A.I. in Apollo into two categories: cognitive intelligence (decision-making) and physical Intelligence (such as balancing, hand-eye coordination, and all the things that we do very naturally as humans). Long term, he says the vision is for Apollo to take advantage of new and different types of A.I. Emerging multimodal generative A.I. models that support not just text but various modes will be important to realizing a true general-purpose robot, he said.
Apptronik says the platform is expected to cost less than the price of many car models. That’s just the base, however, and customers would then need to build out their own versions of Apollo for their own purposes.
Perhaps more interesting is the second business model Apptronik is pursuing: Robots as a Service. The idea is that customers will be able to lease any number of Apollo humanoid robots and pay monthly according to the time they worked.
It sounds a lot like the arrangements between employers and human workers—and the kind of thing that will spark debates about job losses. Cardenas asserts Apollo is a solution to labor shortages affecting critical industries, pointing to fields like manufacturing that are suffering from retention issues and skill gaps that are leaving jobs unfulfilled.
“The question is, ‘Why don’t people want to do this work anymore?’ And a lot of times, the reason is because they [could] get injured, it’s a really difficult environment to be in a hot warehouse, especially here in Texas,” Cardenas said. “And so the idea is that you can now have the robots do the worst part of a particular task, and then you elevate the human so that they can do the things that they are best suited for and like doing the most.”
It’s the same argument made by generative A.I. developers and the industries embracing the technology, who say that LLMs will allow employees—like the striking Hollywood screenwriters—to focus on the most valuable creative work while freeing them from the grunt work. Critics however see a slippery slope that leads to the eventual extinction of their livelihoods.
Unlike a desk-bound scribe or software coder, however, a manual worker in a warehouse might not be so opposed to handing over some backbreaking labor to a machine. Indeed, it is possible that introducing these types of technologies could be part of improving work environments.
It will be interesting to see if and how companies integrate robots like Apollo, and how it affects physical labor, from wages and health care to the length of the workday and job availability.
“I think these are important debates for us to have in this sort of day and age,” Cardenas said, repeatedly comparing this moment to the introduction of the personal computer and saying it’s easy to be afraid.
These types of discussions often get flattened to be just a matter of robot-replaces-human, but it’s a more nuanced discussion. New jobs can be created and the way we work could change for the better. Yet as we debate these types of massive shifts that will change work, life, and society as we know it, it’s important to remember problems can—and usually do—require a variety of solutions.
Sage Lazzaro
sage.lazzaro@fortune.com
sagelazzaro.com
A.I. IN THE NEWS
Meta announces a new foundational model for translation across nearly 100 languages. Facebook’s parent company unveiled an impressive polyglot A.I. this week, with a new model that can translate dozens of languages in a mix of input and output combinations: Spanish speech translated into Vietnamese text, for instance. In its blog announcement on Tuesday, Meta positions the new model as part of its “approach to open science”—though while made available to researchers and developers, the model’s CC BY-NC 4.0 license forbids use for commercial purposes.
Nvidia keeps the A.I. party going. Soaring past analysts’ expectations, the firm reported $13.5 billion in earnings in its fiscal second quarter and year-over-year sales growth of 101%. As the company that makes the chips powering the generative A.I. boom, it’s clear Nvidia is emerging as a big winner of the current hype moment. “The race is on to adopt generative AI,” Nvidia CEO Jensen Huang said in the earnings statement.
Stephen King reacts to his books being used to train A.I. “These programmers can dump thousands of books into state-of-the-art digital blenders. Including, it seems, mine,” the author wrote in The Atlantic. “The real question is whether you get a sum that’s greater than the parts, when you pour back out. So far, the answer is no.” He goes on to say he’s not nervous about A.I. encroaching on his territory (yet) and ponders if it will ever be capable of true creativity.
