 We're here at the MediaTek and you have a face recognition, deep learning. Is this in the X20? Yes, it's a part of an X series of videos. So what is a deep learning? What can it do? It is a kind of artificial intelligence and it can do object, scene, image recognition. For example, bookstore and boxing ring. So this is not on the cloud? Yes, this is not on the cloud. It can do it on device. So it is safer with privacy. So it's on the chip? How is it possible to do this on the chip? You can use. We develop an open SDK to all the application developers and our customers. So that deep learning model like convey and tensor flow can easily run on our platform. So it's a neural network, convolutional, recurrent. What does it mean? What is that? It is the deep learning model. So it can learn the features from big data and know how to recognize and classify features in a more efficient and accuracy way. And can we test the camera also? There's another function you can do with the people? Always people, sure. Let me show you. Let's try in a random person. It can tell the different age and facial expression. For example, this guy is around 30 years old. And this lady is around 20 years old. Alright. So it can recognize people. We can point it at me for example. Yes, I think you are 20. Yeah. Okay. How much? Probably 20 years old. Really? Yes, 20 to 30 years old. Alright. So in the future, your partners can work together with MediaTek to make a better deep learning CPU? I think it's GPU because the computation of if learning is heavy. GPU. And in GPU. So we do some optimization. Is it Mali? Yes, Mali or imagination. We optimize for both GPUs. Which one is better? Both are good. 10 times better than before optimized. Alright. So how much storage is needed on the device? And you put a lot of storage to do learning? We will do memory footprint optimize. But for how many storage we can save, we are still evaluating. Like for this application there you need a 200 megabyte or 2 gigabyte or 20 megabyte or how much? Yes, probably near 10, not 10, several megabytes but maybe 60 or 70 megabytes. And this you put on the EMMC? EMMC. You put this on the phone and then you can add more and more and it's offline. No need for the cloud? Yeah, it is offline. But what I just said is about the DRAM. DRAM? Yes DRAM, not storage. Alright. So isn't it a RAM? Yes. So as for how much storage we need, we are still evaluating. We are still optimizing. So it's ready for this year? People can have deep learning? Of course. Very soon. And there will be many different applications for deep learning on a smart phone? Yes.