00:00It might sound like science fiction, but U.S. tech company Sionic has developed a bionic hand that's being used
00:07to train robots.
00:08Our reporter Michael Morillier spoke to Sionic's CEO to find out how it all works.
00:14So I'm using these two buttons to move it right now, but that's the equivalent of having muscle sensors.
00:20So if you're missing your hand, we can place muscle sensors on your arm, and then you can use that
00:24to open and close the hand in different ways.
00:26So I can give you like a thumbs up, right? I can switch that over and make a pinch.
00:29I can go back and forth like that.
00:31But what's especially interesting about this is that we have up to 30 touch sensors throughout the entire hand,
00:38so one on the fingertip, one on the finger pad, two on the outside, two on the inside, and our
00:43users can feel that through a vibration.
00:46So when they pick something up on the table, they can actually feel the amount of force because of the
00:51touch sensors that are in the hand
00:52and the vibration that they feel on their skin.
00:55Our hand currently has six different movements that it can do, so it can flex and extend all five fingers
01:00and then rotate the thumb as well.
01:02So compared to like, you know, parallel jaw grippers that robots have been using for the last like 30, 40,
01:0850 years, right?
01:09It's a lot more dexterous now.
01:11But the problem was is that you didn't know how to control all those movements in the hand because the
01:16amount of compute you would need for that was way too expensive, way too complex.
01:22But with the breakthroughs that have come through with physical AI, it's unlocked this level of dexterity that now we
01:29can actually do all these more complicated tasks.
01:32Now, you've collected data from the people who've actually used the bionic hand, and you'll share some of that data
01:38with the robotics firm ABB.
01:41How will that data help them to train their robots?
01:45Yeah, so when our humans are doing tasks, right, we can get information generated by the ability hand that I'm
01:53holding here.
01:54So, for example, if I'm doing that pinch, we know exactly how much the fingers have moved, how much force
02:00the motors are producing, as well as how much pressure is being applied on the fingertip right now.
02:07And what we can do is we can actually sync that to meta Ray-Ban glasses, for example.
02:13And so while they're doing the task, we can actually see what that task is, and then they know exactly
02:18when they grab the object, how hard they're grabbing the object, and have it synced to what the object is
02:24from the camera data as well.
02:25So with ABB, we're actually taking the teachings from what we've learned from our human users doing those same tasks
02:33and using that data to train AI models to have the robots do the exact same thing.
02:39I think this is really the crucial question.
02:41Will this partnership help reduce costs for people who really need these bionic hands?
02:47Yeah, absolutely.
02:48And the thing is, there's more than 10 million people with hand amputations around the entire world, and less than
02:543% of them have access to affordable rehabilitative care.
02:59And in order to reduce that disparity, by having a beachhead market, right, like the robotic side, the volumes help
03:08reduce that cost so it can become affordable for all these people across the entire world who need them.
03:16And so what we build for humans benefits robots in the sense that the data can help train the robots
03:22to do all sorts of different things.
03:24But what we build for robots benefits humans because the volumes on those sides subsidize the cost for the humans
03:29so that more human users can get access to affordable bionic limbs.
03:34Dr. Adil Akhtar, thanks so much for talking to us.
03:37Thank you for having me.
Comments