ColdFusion - Proof That The Future is Here!

Starting point is 00:00:00 You are watching Cold Fusion TV. There's something going like, selfie. Hi, welcome to the first Cold Fusion video of 2017. I was originally going to release a video on Sony, but it's taking longer than expected. So in the meantime, we're going to start with something a little bit fun here. Today, let's look at five innovative technologies from recent months.

Starting point is 00:00:25 I think some of these will really surprise you. Let's get into it. Let's start off small, very small. Number 5. The $1 paper microscope. Education in developing nations is a rising phenomena, but it's about to get a lot cheaper for some keen young scientists. With a little invention called the Foldscope, it's an optical microscope developed by Manu Prakash, a biophysicist at Stanford University. It costs about $1 to produce. The Foldscope's aim is to provide cheap and easy to use scientific tools for the developing world.

Starting point is 00:00:59 And yes, what you're looking at are some actual images from the $1 device. Students can view the images live, but if they want to take a photo of their discovery, there's an attachment that allows for phones to be used as a camera or a display device. The Foldscope is mainly made out of paper but includes small spherical lenses, an LED light, and some prisms to provide magnifications from 140 times to 2,000 times. With student microscopes ranging anywhere from 100 to 1,000 US dollars, a $1.00 option isn't the worst thing in the world. Number 4. Giving devices a sense of touch.

Starting point is 00:01:40 Now this one is a strange one. When I first saw this, I thought it was fake so I had to do a bit of further research to verify. And it really does seem to be real. So here's the deal. Back in 2015, Google released a radar chip called Sully. Its purpose was to allow for easy detection of touchless gestures. Researchers at the University of St Andrews has successfully used this chip along with their in-house recognition software to actually classify and detect objects and materials. In other words,

Starting point is 00:02:10 the system knows exactly what objects is being placed on its surface. The project is called Radar Cat. This stands for radar categorization for input and interaction. The software mainly uses machine learning to understand and get better trained at recognizing objects. Although it's in its early stages, there is already some talk of some possibilities. For example, robots that often come into contact with humans could use this system to understand. that they're touching human skin and that they should be extra gentle at that time. My guess is it would take something like a huge database of recognized objects stored up in the cloud to make this anywhere near as good as identifying objects as a human, but it's still cool to see.

Starting point is 00:02:50 Number 3. An AI can determine what's happening in a scene purely by listening. Researchers at MIT created a neural network and fed it 26 terabytes of video data from Flickr. The research has found that the neural network can interpret natural sounds in terms of images and categories. For example, if it heard a scene of birds tweeting and leaves, it would note a picture that with forests and nature scenery. Okay, so next we're going to do a little bit of an experiment. The following is going to be a bunch of clips with the audio playing but with blurred visuals. See if you can guess what's going on in the scene just by the audio.

Starting point is 00:03:26 Let me know in the comment section how many correct guesses you get. As it turns out, this artificial intelligence called SoundNet, called Soundnet, interesting name, is about, on average, 93% as accurate as guessing as a human. So in this case, the AI is actually doing a fairly good job when compared to humans. Some uses suggested by the researchers include a cell phone switching itself to silent automatically when it recognises that it's in a cinema hole, or just generally giving situational awareness to autonomous robots. So at this point, it might be worth suggesting that if you're confused about AI or just

Starting point is 00:04:55 keep thinking of SkyNet every time it's mentioned, it might be time for you to find out exactly what it is. And for that reason, I've done a video exactly on this topic that defines what AI is from first principles. There'll be a link below. From watching that video, you'll find out that AI is actually all around us and has been for years. In fact, if you're not subscribed to this channel and are watching this video, chances are that YouTube's AI actually got you here. All right, moving on. Number two, simulating the future of a scene. is pretty damn cool. And I mean straight out of science fiction. Imagine giving a still image to an AI and it gives you back a video prediction of what happens next in the scene. This is exactly what

Starting point is 00:05:38 researchers at MIT have done using deep learning neural networks of course. They train the AI on 2 million unlabeled videos and it can now predict the future motions of scenes it's given or even create scenes that it imagines from scratch and I use the term imagine very lightly. The way it works is probably actually the coolest bit. The deep learning method they use is something called adversarial learning. It involves training two competing neural networks. One network generates the video and the other decides if that video generated is fake or real. Sometimes real videos are slipped in with the generated ones. Over time, the AI that generates the fake videos learns to fool the other AI into thinking that they're real. According to the researchers, as the video generating AI gets smarter, it can

Starting point is 00:06:25 create videos that represent scenes from beaches, train stations, hospitals and golf courses. There are some quirks though. For some reason, the AI tends to make humans look much larger than real life. And right now the videos are only 1.5 seconds long, but as the algorithm is improved, the researchers hope to make this time longer. This invention comes hot off the heels of yet another MIT experiment. This one's interesting as well. They made an AI watch TV shows like The Office and Desperate Housewives. After a while, it could determine what actions the humans are most likely to do next in a scene, whether it was a handshake, a hug, or a kiss, and it got very accurate at doing it.

Starting point is 00:07:04 But an AI predicting human behavior, this one I'm a little less keen on. Alright, to number one. A huge leap forward in bionic legs. This invention is number one, not just because it's a great advancement in the field of bionics, but also the story behind it. It's pretty amazing. This is Hugh Heur, a champion, mountain climber and a figure well known in the sport. In 1982 at age 17 he lost both of his legs while climbing due to frostbite but instead of giving up he went

Starting point is 00:07:36 ahead and built his own legs and began climbing again. Hugh had reached his old climbing ability within a year and with determination he even surpassed it. He was so good that he became a threat to other climbers and they said that he had an unfair advantage. This made him see that technology could really improve and regain quality of life. As he kept working and tweaking the technology himself, he became inspired by this experience and took his studies seriously, pushed by the thought of what he could create with a higher education. Eventually, he would become a pioneer in the field of Bionics at MIT. Even to this day, most Bionic limbs are passive, becoming more of a hindrance requiring up to 20% more

Starting point is 00:08:19 energy to use. Hugh Hur would change this in 2016 when he received the Princess of Astorius Award for technical and scientific research for his latest pair of limbs. Instead of weighing down a person, these limbs actively inject energy back into the movements of that person. As a mechanical engineering graduate myself, I find the work by Hugh and his team

Starting point is 00:08:41 absolutely incredible. Here are some highlights. Our key strategy in the design of bionic appendages is to look into nature. So we studied how the calf muscle works, for example, and how the calf muscle

Starting point is 00:08:56 is controlled by the spinal cord using neural reflexes. We programmed that capability on the small computers that are underneath the shell and the bionic limb. Each bionic limb has three computers, the brains of the system, various sensors to sense its environment, and a muscle tendon-like actuator, a motor system. And when I walk, it adapts automatically with variations in speed and terrain.

Starting point is 00:09:25 in a way that's completely natural. We actually embedded on chip the fundamentals of how muscles are controlled that span the ankle drain. The main breakthrough here is that most the processing on the legs are done instantly, meaning that there's no delay in reaction time

Starting point is 00:09:42 to fast changes made by the person wearing them. It works so well that there's almost no training period required. It feels natural. We normally just walk on a level ground and try to walk right up the hill. Oh my. Pushing you up?

Starting point is 00:10:00 Yes, it's not even described up. It's pushing you right up. You guys just don't know what it's like not to have that. In order to do this, Hugh and his team had to study the intricacies of walking in detail. For example, how does the foot and ankle subtly roll with each step? How are the tendons working at each point of each stride? Hugh and his team also modeled the stiffness properties of muscles, bone and skin, using a method called finite element analysis.

Starting point is 00:10:34 As a side point, coincidentally, this is exactly what I did for my third year project at university, but it was with the movement of human brain tissue during trauma. My experience gives me further insight into what these guys are doing, and I know for a fact that this is no easy pie. This modeling is done, so each limb has a perfect 3D printed fit, custom made for each individual. So I just love designing Bionix. I love every few months upgrading my own legs so I have better balance, I can move faster. You know, it's interesting because my biological body as I age gets worse and worse

Starting point is 00:11:14 due to age-related degeneration, but the artificial part of my body gets better in time. So I predict when I'm 80 years old, my balance will be better than an 18-year-old. That's a lot of fun. Each year, there are 3 to 4 million debilitated amputees across the world. Up to this point, there were no real prosthesis made by actual amputees. When you think about it, who better to understand the physical and engineering challenges than an amputee biologist mechanical engineer? It's a very rare find, but it was just what the industry needed.

Starting point is 00:11:47 Hughes team is already pushing forward, looking at prosthetics attached straight to the bone and limbs controlled directly with brain impulses. Overall, this is great progress and with some scaling in production and adequate funding, this could restore the quality of life to millions of people around the globe. Wishing Hugh and his team all the best. And with that, that wraps up five interesting breakthrough technologies in recent months. If you liked what you saw, give this video a thumbs up and feel free to share it with someone who would be interested. Sharing is caring.

Starting point is 00:12:29 And don't forget to subscribe if you're new to this channel. So thanks for watching guys, this has been to Go-Go-Go. You've been watching Cold Fusion. And until next time, I'll see you again soon for the next video. Cheers. Have a good one. It's me thinking.

ColdFusion - Proof That The Future is Here!

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.