For all the huge and hard-won improvements in robotics, it’s a bit hard to get away from the impression that robots are still “clunky” in many ways. They still look and in many ways act steampunk.
That’s not where they need to be, and it’s safe to say the sector is well aware of that fact.
Enter the word “neuromorphic”. You can expect to see a lot more of that expression as the technology overtakes old-style systems. It’s not really a new expression; it’s been sitting around in backroom jargon for a while, but now it’s center stage. Neural networks have been the foundation of AI development, and it has taken this long to reach their current state.
What’s critical about the expression neuromorphic is that this is a whole class of tech in one word. It means neural systems mimicking the human brain, but it’s far more complex. It’s all about functionality at the most basal level.
Things like navigation, situational awareness, and coordination of basic movement operations are essential for robots. The human brain does these things instantaneously through the autonomic nervous system. The “mechanical” brain can’t compete in terms of efficiency. Neuromorphic systems have now achieved that.
Queensland University of Technology (QUT) has come up with a system called LENS Locational Encoding with Neuromorphic Systems, which is 90% more energy efficient. The LENS system uses neuron-like electrical spikes to manage navigation and operational needs. Whether or not it can get to a truly autonomic operational level is the issue here.
QUT may also have hit the magic formula for realism in system design. The critical point here is based entirely on the economic realities of robotic systems.
They use far too much energy to do basic things. They’re not “world ready” in a lot of ways. Robots look great in controlled environments, but they’re still at toddler stage for other environments. That’s a severe operational limitation. Add cost of energy to severe operational limitations, and you see the problems.
The vast array of types of operational robots aren’t so much autonomous as having their hands held. They’re under some degree of external control, and they’re also networked. Neural systems are essential to free them from these constraints.
This is also still the Cambrian era of robotics. We’re at roughly trilobite stage. There’s a lot more to be done. Check out QUT’s robotics search information and you can see how far just about all aspects of robotic systems and designs have to go before they get to the science fiction level.
The neuromorphic approach is a fundamental issue:
Robots don’t have to be humanoid, but they do have to interact with humans. That is a problem, and the solutions must be trustworthy.
As AI keeps proving, human interactions with technologies are hardly perfect. Even simple terminology can get in the way.
A human brain works on any number of levels of proficiency and is highly reactive to incoming information and/or tangential information. Robots therefore need to have at least the processing speed to keep up with these erratic interactions. Neuromorphic systems can do that.
The AI overview of a search for “AI robotics” ironically proves the point. AI robots are inevitable. This overview defines an obstacle course of interactions and systems issues. Human and AI systems are very different. Working at different speeds and with often very different frames of reference are real stumbling blocks.
Neuromorphic tech will eventually remove these obstacles. LENS and LENS-like capabilities are a sort of beacon in the darkness. Energy efficiency equates to system efficiency equates to operational efficiency.
This is where robotics needs to go.
_________________________________________________
Disclaimer
The opinions expressed in this Op-Ed are those of the author. They do not purport to reflect the opinions or views of the Digital Journal or its members.
