If we want to invite robots from isolated factory halls into our messy homes, we first need to build them so that they can function in human environments. Learning about biological neural networks can help us figure out how to give artificial systems similar capabilities. Here, on the border between industry and academic research, is where Yulia Sandamirskaya works.
Yulia built her first robot out of a soapbox and a couple of electric toothbrushes at the age of seven or eight, she remembers. For her, robots are the most complex puzzle that a human can build: an incredible amount of interdisciplinary elements need to work together in harmony. This is a challenge worth dedicating one’s life to.
Robotics can also dramatically improve our lives, Yulia believes. She is not particularly fond of tedious routine tasks, and she would much rather delegate such work to robots. We already have great experiences with robots taking over straining and monotonous tasks in industrial settings. However, we have yet to see robots that work well with humans, in an environment full of people. Humans are unpredictable, fragile, emotional, and notoriously diverse communicators; it takes a lot for a machine to adapt to our changing needs.
“You don’t see robots roaming hospital corridors – yet”
When we build synthetic systems to work side-by-side with humans, it’s not a bad idea to look at how biology has solved similar challenges. After all, a tiny insect can navigate its environment flawlessly with mere 100,000 neurons. Neuromorphic computing is a discipline that studies biological systems and neural networks and then develops software and hardware inspired by them.
Yulia leads the Research Center for Cognitive Computing in Life Sciences at the Zurich University of Applied Sciences (ZHAW). Right now, she is putting together a new neuromorphic computing research group to advance the basic technologies needed for robots in human-centric environments. The group works in close collaboration with companies like NEURA Robotics, SynSense, and Intel – the last one a former employer of Yulia’s.
The second part of Yulia’s mission is to create a use case and market for the next robot generation. She already knows the setting: hospitals and elderly care homes. These facilities are a good middle ground between highly structured, disciplined factory floors and chaotic, unpredictable private homes. Hospital environments bustle with people, but the staff is used to following protocol.
Developing an agile robot won’t help if it doesn’t meet an actual need. This is why it’s important for Yulia to include hospital staff, patients, and facility experts in the design process. She has joined forces with a ZHAW colleague, Nicole Gerber, from the school’s hospitality and service management group. Nicole brings the project experience in introducing new technologies in hospitals, Yulia says.
“We need to go beyond today’s AI”
While Yulia finds the recent artificial intelligence (AI) developments inspiring, rapidly improving language models and other generative AI tools have little impact on her work. We won’t be able to include today’s AI into a robot’s control loops, because it’s too slow and it consumes too much energy, she explains. We can’t do much with a robot that needs to be plugged into the wall because its brain needs as much juice as its muscles.
Yulia recognizes four challenges that we need to solve in order to have robots by our side. First, we need to create accurate, fast, and efficient vision systems, so that robots will be able to understand their environment. Today’s smart cameras are too slow to enable real-time reactions, Yulia describes, which is why we need better chips and algorithms. This is the focus of Yulia’s new research group.
Second, we need to redesign the embodiment – the muscles of a robot. Industrial robots have very heavy and robust embodiments to enable high precision. Such musculature makes for very heavy and hard machines, which is a no-go in human-centric environments. Our robot helpers mustn’t hurt us when we bump into them accidentally!
Third, our robots need to get better at manipulating different kinds of objects. Human hands are still unbeatable; current robots need a stockpile of specialized grippers to manage variety. Fourth, robots must learn to understand spoken human language better. ChatGPT is already pretty good at understanding text, but AI audio comprehension lags behind.
“Ten years from now, our technology is subtler and our planet greener”
The future Yulia is working towards includes more technology: we will have better tools for exhausting work like mining, building, repairing, recycling, and maintenance. Yet we will hardly even notice these tools grinding on in the background; that’s how well they work. Our planet will be even greener than today, both in terms of aesthetics and sustainability. Even farther in the future, Yulia wants to see robots exploring other planets and finding new energy sources to conserve Earth’s resources.
Meanwhile, we humans get to concentrate on creative work and other fulfilling pastimes. All this is achievable, Yulia concludes, if we focus our smart minds and resources on building this future instead of wasting them on waging wars.
This article is partially based on this podcast episode.
Comments