Robot Futures #3 – the AI 'broligarchs' may be digging their own graves. Here's why

Earlier this week I spoke to 3Laws’ PhD, co-founder and CEO, Andrew Singletary, whose company is working on critical areas in robotics: safety, interoperability, and functional communication between different types of robots and other autonomous and cyber-physical devices.

The aim is not to get robots to understand Isaac Asimov’s fictional Three Laws of Robotics, on an intellectual level, as you might imagine from the name. Creating an ethical robot is an order of magnitude larger a task than simply making a machine that is safe, and which has simple functional limits – for example, ‘don’t move beyond this geofenced point’, ‘avoid colliding with objects’, and so on.

For years, robotics has been characterised by the kind of principled, iterative, safety-focused, standards-based work that Singletary and his colleagues exemplify; work that seeks to identify real problems and create workable solutions to them, often collaboratively and globally.

But recently, the focus on domestic humanoids by some high-profile AI CEOs, plus the sudden influx of billions of dollars from impatient investors, has seen extravagant claims made for general-purpose robots’ abilities. Some robots may be released into our homes this year, such as 1X’s Neo (if you have a spare $20,000 or $490 a month for a subscription).

But the claims made for super-intelligent, general-purpose humanoids are unlikely to be realised as quickly as hype suggests, which risks leaving an excited public disappointed. Some robots may even have to be tele-operated in our homes by offshore workers. After all, Neo was tele-operated in its publicity videos to drums up advance sales: it was unable to perform any of those tasks autonomously.

This is because the challenges of creating a genuinely intelligent, truly general-purpose dextrous robot are simply too complex at present: such robots need to have a three-dimensional world model, visual intelligence, and Vision Language Action Models, Large Behavior Models, and other systems that help them turn verbal instructions into actions and then predict the likely outcomes.

Physics is fact

Put simply, robots need to understand physics, but not as a Large Language Model  (LLM) does: it needs to know, experientially, what happens when it moves through physical space and interacts with objects and people.

All that requires a lot of training and a mass of data, not all of which can be simulated accurately enough to create a safe, trustworthy machine. And yet the AI hype machine – which is an increasingly toxic and political device – has convinced many of the public, and a lot of credulous investors, that all this is within robots’ clumsy grasp.

Is Singletary worried by this, especially given that humanoids intended for industrial applications may also be working in our homes? Is the tech bros’ social media gravity warping the market around them? And might that compromise good progress towards interoperability and standards? He says:

Yeah, I can share my personal sentiment on the matter. It's really their own graves they're digging, but not the industry's grave. I'm not worried about it from the industry’s perspective. I think they're setting unrealistic expectations for their investment. And I do think it'll cause a bubble. And a lot of people foresee that bubble bursting if the humanoid companies are not able to meet the timelines they're putting out. And a lot of people familiar with the industry are highly sceptical – and not just for safety reasons, but also for a broad variety of others.

But the robotics industry is much more than just the humanoid hype cycle – it’s the AMRs [Autonomous Mobile Robots] that are already populating logistics facilities today, for example. But I certainly align with broader sentiment in the industry that the timelines they’re setting are wildly unrealistic. It reminds me of Elon Musk saying that full self-driving is coming this year for… how many years in a row?

At the time of writing, the tech bro of tech bros Elon Musk has announced that he is re-focusing Tesla’s business on humanoid robots, and few would deny that he has the cash to keep failing until he succeeds – just as he did with SpaceX’s rockets.

But there is a related question amid all the hype: how to get the robotics world onboard with the drive for safety, standards, and interoperability, when commercial pressures are as likely to impede it as accelerate it? And how global are these conversations? After all, China, South Korea, and Japan are in many ways more significant players than the US in physical robots.

Speaking about his moves to co-create the new communications standard, Singletary says:

At our first meeting, we had representatives across Asia. With its original work on creating an AMR interoperability standard, [Boston-based independent hub] MassRobotics took it upon themselves to create that framework. But this [new, broader standard] is something where, from the start, we’re facilitating a group coming together and starting the discussion. We very much want it to be an international group of experts lending their opinions and formulating a framework, as opposed to ‘We're going to do it all ourselves and see what the industry thinks’.

Challenges

But as ever, there are practical challenges in our complex, physical, human world, he explains:

ISO has this challenge: how to hold meetings when you've got people across every timezone on Earth? I'm dealing with that challenge right now, because the humanoid safety standard committee is meeting currently in Korea. I'm a part of that effort, but I was not able to travel to Korea for it, so I'm having to stay up late nights instead.

Other challenges lie ahead too, he says:

Certification is a big milestone that we're about to achieve. We've been working with a third party who's been certifying our software to IEC 61508, which is the backbone of software safety in the industry. That will enable us to take the next step of productisation. Once we have that stamp, once we don't have to convince every customer that we've fulfilled these requirements and there's a third party backing us, it'll be easier to reach a broader set of customers.

My take

For all our sakes, let’s hope that is true.