RGI on the Path to AGI

Artificial general intelligence, or AGI, is the presumed end game of the current artificial intelligence (AI) race. Many people are certain it is going to happen, with a large number of scientists and pundits expecting AGI “any day now.”

I think, however, that AGI is still pretty far away. There is significant work to be done to get to the point at which software can truly “think” like a human, and there are lots of gaps to fill. Perhaps the biggest gap is perception and tactile sensing, because, without a better understanding of the physical world, no software is going to reach the stage of general intelligence.

But what will be the revolution that comes before AGI? I think it might be RGI, or robotic general intelligence. If you have not been paying attention, the robot space has exploded lately. And much like the term “AI” has now been co-opted to equate to large language models (LLMs), “robot” is being co-opted to equate to humanoid robots that can perform real-world duties like folding laundry, assembling furniture or standing in line at Crumbl to bring home the cookie of the day for its owner. Much like the pre-LLM AI products of a few years ago have gotten pushed aside in terms of hype and focus, humanoid robots are pushing aside precursor robotic platforms. Figuratively for now, but probably literally in the near future.

Companies ranging from household brands like Google to startups like IntBot are racing to build the software brains to power humanoid form-factor robot bodies, all to deliver robots that can take over various jobs that, until recently, were assumed to be impractical to automate. This is driving RGI development, and we are seeing robots that have capabilities to refine tasks, and then immediately share the refinement steps they discovered with every other bot doing the same task. The net effect creates a robot workforce that can continuously increase efficiency, as would be expected, but this also creates learning and perception elements that I believe are ultimately required to drive anything we would label as “AGI.”

Today, these humanoid robots are just starting to be deployed at scale. In most cases, they are still in limited and controlled tests while developers of the hardware and software continue to refine the training and deployment process and optimize everything into a commercially viable platform. But we are seeing progressive companies recognizing this shift and getting in front of it.

The partnership between Figure and BMW is one public example. Figure recently announced that their Figure 03 models were replacing the Figure 02 models deployed at the BMW factory in Spartanburg, South Carolina. Closer to our industry, a leading computer and technology company has recently retained the security design and consulting firm ADRM to conduct a trial of various robot platforms at four different facilities to see how the company might be able to leverage robots, and possibly RGI, to extend the reach and effectiveness of its security operations.

If my thesis is correct, we will see general intelligence will show up first in bodies, not in browsers. Physical work has bounded success criteria, a paying customer and a data flywheel that cognitive AGI still lacks, and security is one beachhead where that transition potentially becomes visible.

As you are considering how AI can impact your organization, and how you can use it to your advantage in security, do not overlook the RGI opportunity. I expect we will see more and more organizations deploying robots in the coming years.

The views and opinions expressed in guest posts and/or profiles are those of the authors or sources and do not necessarily reflect the official policy or position of the Security Industry Association.

This article originally appeared in All Things AI, a newsletter presented by the SIA AI Advisory Board.