in

Qualcomm Snapdragon 888+ & AI: How Smartphones use AI with PocketNow

i? I think that the topic has been brewed a lot for the past. What like five years, uh, not much. To be honest, so its probably really good that we have our guest on for today, uh, you can see him right in the middle right here on our three shot. We have ziad ziad asgar. Thank you so much for joining us from all the way from qualcomm. Now some of the questions and some of the things that we talk about in this episode was made possible by the snapdragon insiders community. If you are not already a part of this community, you can head over to snapdragoninsiders.com and you can get exclusive access to the latest snapdragon products and technology, so make sure you go and check that out busy out welcome, feel free to introduce yourselves to all of Our listeners and viewers um the floor is yours, yeah uh, you know what i do for qualcomm is essentially looking at all the different technologies that go into our products, and one of the key ones that were really focusing on nowadays is artificial intelligence. Music. You know for us when, when we were when we started the whole discussion about having this this interview and everything i i was like the first person to be like okay. We definitely do need to do this, mainly because im the first person that wants the most simplistic and basic answer for what is ai, or at least the state of ai, like right now so essentially, artificial intelligence is this umbrella term that we have coined that basically Talks about machines that have intelligence as in being able to sense whats happening around them, to be able to reason based on that and then to even take some actions.

So anything that is able to do that and, as you can tell, we are not there yet. Honestly, a lot of this thing, these things get bundled under artificial intelligence, but really there are subsets of artificial intelligence, the first one being what we call machine, learning so think of machine learning. As you know, when you typically program something you go ahead and say and use certain, if else statements to tell the machine what to do now, you dont have to explicitly train the machine to do that, but you can actually give it data that its able to Learn based on that, and then the last level of that, where we are today, is what we call deep learning, where we use neural networks to be able to do some processing that allows machine to take certain actions. I feel like a lot of people with their daily lives, using their smartphones or whatever the technology may be kind of takes all of this for granted. So i guess my question is what real world examples would you be able to give us um on how ai is truly making a difference in our daily experiences with the smartphone or really anything else, thats a great question but ill take at least you know one Example, think of it this way in the past the sensors, on a smartphone or on a security camera, they really used to be dumb. What that means is they would capture whatever light is coming into them, but they never knew what that light stood for with ai or with machine learning were giving those cameras, those microphones, the ability to be able to understand the data that theyre capturing that, as you Enable your devices to be able to understand the the stimulus that they are getting.

They can markly improve the use cases that we have on them all right im. Actually, this is like the perfect segue for the next question that i had because, probably where ive heard it most is in photography. So i do know that the processing the software portion of it and even the hardware have a lot to do with what we get in the final result, and i i would really like, if you would, you know if you could elaborate more on how ai takes Control into all this process, how do you teach it to be better like how? How does that make a difference? Sure so, like one great example, is i mean if you just look at the capabilities the cameras have today in terms of being able to capture images in low light conditions. I mean thats a very good example on how artificial intelligence has allowed people to take out a lot of the noise that might be there in that image. So what you do is that you train that camera by training it with long exposure, which means its able to get a lot more light in versus what its seeing with short exposure. You actually train that neural model to be able to take out a lot of the noise from the image, for example, but thats just one example. So there is like so many levels of improvement that we are seeing, especially on imaging and video um. If i could switch gears just a little bit, though we are talking about the capturing of photo and video, but um, can i ask you to dive a little bit into how ai is helping the consumption experience, because, as ive heard, ai is actually helping when it Comes to me viewing various forms of media content, whether its on youtube, whether its streaming netflix anything like that from a consumption perspective.

Another very important aspect that we see, especially as were working on some of our products that go into cloud inference. Engines uh. Just imagine that somebodys uploading, a video on you, know a platform like facebook, so many different levels of it right. Another key capability that ai provides is it can actually improve the quality of resolution of a video. So, for example, you might be in a location where you dont have that much of bandwidth on your device. Well, i can send and transmit that video to you at a lower quality, but within that device im actually able to upscale that and still be able to show that at much higher resolution with much better quality. We call this super resolution, so there are so many different levels of consumption. Experience also were improving. I think thats a perfect way for us to get into one of the cruxes of our discussion here, and that would be the snapdragon 888 plus, with this being your latest flagship, processor, um and the sixth generation. Now of your ai engine. Being a part of this brand new flagship, can you tell us more about what this new addition or this new version or update, is bringing to the table yeah so snapdragon, 88 or like we like to call it? The beast: uh is really a great product. Honestly, okay, its a bit of an inside joke, but honestly the capabilities on this product from every perspective, whether its artificial intelligence or graphics, or you know, uh camera.

You know, for example, on the camera. Now we can actually capture 120 12 megapixel images in one second. So this is basically increase of almost 30 percent. You know in now you couple that, with with the ai piece and with the ai, we have, you know, enabled almost 26 trillion operations per second come out with the 888 plus. Actually, it actually pushes that up to 32 trillion operations per second, what weve also done along with that is a completely new architecture where you have taken, you know the way we did it is. We have multiple accelerators inside our ai engine to be able to map it to a way a neural uh. You know engine looks like so we have what we call scalar vector – and you know tensor processors, without that, within that accelerator and weve increased the performance of all of those and at the same time we can do three times more processing in certain scenarios. For the same amount of given power, and along with that weve, basically redone, all the software stack to be able to leverage this amazing hardware that we put into this product. I guess the perception or the misconception that we have is that the more powerful the chip, the more power consuming uh youre, telling me that one of the benefits of the 888 plus is the complete opposite. Then. So what we excel at at qualcomm is exactly what you mentioned: to be able to do more processing at the lowest amount at the lowest power consumption, because that is how you make that device last the whole day and that same advantage and those same techniques that We have learned over many years.

We can actually apply them to those new areas and bring those benefits into those new end products. You know weve been diving into power efficiencies and everything and again i i think that its just that i think that the the paradigm that weve had lately is that the more power, the more cooling you need, tell me more about how you know how this ai, Actually like how does ai know when to be more efficient because im even noticing, for example, that my phone senses when to charge the way we were able to do this? I talked about the engine there, so we have very big engine thats for very intensive use cases, but then we have a very small ai engine. Also that sits in the sensing hub to be able to do that always on like processing right so thats. How were able to cover both? So if, if the ai and all of the data that its already using in order to perform its functions, is on device, obviously you have to provide it with said data, but as far as its ability to grow to learn even more. We talked about machine learning and deep learning before um, how much of it is required that you add more data to what the ai can perform with each new iteration, and how much is it able to actually learn on its own to evolve on its own? Its actually an area that we are really focused on, which is on device learning, so you would not do the fill up training that you do typically because that is very, very intensive and usually happens in server forms.

But what you can do is what we call uh. One example is federated learning, for example, so ill give you an example. When you use your keyboard and especially if youre using google keyboard, it actually uses something called federated learning, so its actually optimizing for the way you might use your keyboard or the words that you might use so its basically tweaking the model based on exactly what you Do on your device and then it might actually share a very, very synthesized version of that to the core model that sits somewhere in the server without exposing any of your private information. It actually enhances many of those capabilities. It customizes those experiences to you, the user and, i think thats, where were going in the future. This is the perfect segue, because, actually i wanted to talk about that feature i mean, is qualcomm working on next when it comes to this, like you mentioned, for example, the fact that youre working on stuff thats going to happen five years from now i mean we Dont mind any insider information on this podcast feel free to say anything. You want right here sure lets. Take the example of you playing a game in the future and youre playing that game using artificial or rather augmented reality. Typically, you hold a device and youre able to press certain buttons to play the game. No more. You can actually point at a character in the game. You can talk to the character using natural language processing because well thats a capability that we are doing based on how youre playing the game, the game can actually change the plot of the game, make it different, make it more interesting.

All of those are examples of what we have coming up and i always like to talk about uh. You know there is a notion or a concept of ai for good, so there was this engagement that we did some time ago with the with a company in india, but you can actually take your smartphone camera put a small adapter on top of it with a Lens and it can actually look inside the the eye of a person and be able to detect, for example, things like diabetic retinopathy, which is a condition which deteriorates the eye. Now this might be an environment where people dont have those capabilities or they dont have access to uh. You know healthcare very readily well, but something like this youre able to help those people that level of excitement that youre that youre showing us here is something i wanted to actually ask more personally: where do you want to see ai come through? What are you excited personally to really see it evolve into what i personally really am looking forward to is another angle or another dimension of application of ai. What were focusing a lot on is applying ai to technologies that we have multiple technologies into inside our products. So the levels of application of ai are basically at use case level, but its also a technology level, and now its going even at the product level, we can apply ai to be able to make our 5g modem even better.

Our leading 5g modems can now detect signal from even more complex conditions, with the application of ai, for example. So now we have a special segment here, uh that it was an open prompt that qualcomm put out for the community that we mentioned earlier. Um snapdragon insider shouts out to all of you who might be watching this or listening to this right now. So one of the questions that was posed to the audience was the snapdragon 888 plus 5g mobile platform is the ultimate in ai performance and its coming soon. To a device near you, what burning questions, do you have about ai and what it means for how you game stream listen or watch? So we did get a few responses that were going to go through right now, all right! So our first one here, as we know, ai basically learns on its own and keeps updating itself to get better. Is it possible that ota updates can be provided that enhances the capability of ai? What i was trying to explain is that we can do that over the air sort of a change, but we can also do that on device as time goes by, and i think if we keep privacy and all into uh in our minds, then probably as we Start to do more of the on device learning. I think that probably would be a better path, but of course there could be scenarios where the data is not of uh private nature and then absolutely it can be done uh too, so you can envision that you can actually provide a new new neural model to Be able to do better noise, uh suppression, for example, in an imaging scenario, so absolutely something thats doable.

Can i can i add to that question. Just uh. Has this been done already its already being done to to some extent, but i think the maybe the question that wanted yeah this there will be a lot more of that happening. I feel uh. So our second question here um, is one thing that we users arent satisfied with, is the battery no matter how big it is uh, so it is uh. So is it somehow possible that whatever were doing the ai would literally kill all of the background processes and concentrate on what were currently doing to enhance those optimizations and the battery thats? Actually a prime example of what ai can do from that? You know contextual picture that it has of the consumer right, for example, youre driving. Well, maybe your wi fi should not be running anymore right. There are things of that sort that can automatically be done depending on what the consumer is doing, and i think we actually have uh done some of that work and some of the improvements that we are actually doing on the device uses ai to be able to Understand what parts of the engine should be running, what shouldnt be running, what should be turned off in the device? All of those are actually very good applications of ai that are being done already and then our final one for uh, for this is um. There are some drawbacks of ai that we did see with tay, which is the microsoft ai that uh im not gon na, go through what happened there, but if someone bypasses the security and reverse engineers, the process making the the ai do similar things.

Can that be prevented? Can security be provided through ai if our data is being collected unknowingly and we get notified yeah? I think i kind of alluded to this uh earlier, but the application of ai, as it applies to newer and other technologies, is absolutely an area that we are focused on. Ai can actually look at what is happening on the device. Is there a lot of data? Thats being transferred is something on those lines happening, so that can absolutely be something that ai monitors. But i want to say that our products, the way we define them securities front and center, and we make sure that our devices and the content on that are super. Secure so yeah we have a lot of technologies baked in to make things more secure, but ai can enhance that further, absolutely whats. The ad we want to. Thank you once again for being a part of this episode, this podcast this video with us. Thank you. So much we would love to have you again um. I really appreciate to have you again awesome.

What do you think?

Written by freotech

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

Loading…

0

Steve Wozniak, Apple, Space hat Elon Musk Really Thinks About Steve Jobs

Qualcomm Snapdragon iPhone 13's A15 chip DESTROYS the Exynos 2200 & SD898! 🤯