 So we're here at Mobile World Congress 2019. Hi, so what's the latest? Hi, well ML is everywhere, 5G is everywhere and 5G and ML enabled devices are just becoming more and more real. Of course you do see a lot of hype, but what's amazed me really is the amount of reality behind the announcements and we're seeing more and more ML enabled machine learning enabled devices, particularly in the small microcontroller world that is just taking off massively. So it's possible to do AI in microcontrollers? Yes, so we've already released the compute library for Cortex-M based microcontrollers that's called SimpsysNN and there's been massive take-up of that and we see that just about everybody is using that. However about a week ago one of the things we released was something called Helium which is the V8.1M instructions head architecture extensions. So we're making a whole bunch of instructions head architecture extensions available on Cortex-M based microcontrollers to improve machine learning processing and it's really fantastic. We're looking at 15x improvement in up to 15x performance improvement on machine learning workloads and about 5x improvement on more traditional DSP type workloads and we feel that this is really going to power the revolution of machine learning based smart endpoint devices. So is this IP or is open source accelerating algorithms? So two things the software available is open source, it's software which we have seen massive numbers of downloads on. The V8.1M instructions head architecture is of course IP that we will be delivering in future releases of Cortex-M based microcontrollers. And V8-M has a lot of stuff doing that accelerates the AI even more than regular microcontrollers? So we've added a bunch of vector extension instructions to enable the sort of complex matrix arithmetic that you see very prevalent in machine learning workloads particularly neural network codes and they go really an awful lot faster as a result of that. So how is it possible to do AI in those tiny microcontrollers and what is it for? Well really the three main use cases we see that are absolutely taking off are the three Vs. So we have vibration, we have voice and we have vision. So vibration this is smart monitoring of industrial devices. Voice obviously more and more enabled devices you know the sort of Alexa voice assistants and things like that and vision of course we perceive so much of the world through what we see and smart devices will be doing the same thing. Now on Cortex-M we're already seeing people implementing significant capabilities on the existing devices today. What we're doing is we're announcing some future extensions which will make those devices even more powerful in the future. Future extensions so it's like in a next generation of microcontrollers. Next generation of Cortex-M based microcontrollers yes. So the microcontrollers those V8-M is not just about the security there's also some other stuff like AI. Security is incredibly important as our endpoint devices get more and more powerful than it is clearly the case that there is more and more data there that we want to secure. If you're enabling these devices access to your extremely personal information then you will want to know that that data is being kept secure. So we believe the combination of the ARM architecture and these new extensions together with the platform system architecture we've announced some time ago the security manifesto all of this together makes ARM the platform of choice for these endpoint devices. So V8-M as far as I remember is like two and a half years ago with something announced. That's about right yeah. So how has it been the adoption so far? You're seeing Cortex-M microcontrollers coming out now you're seeing the silicon coming out based on those devices and of course there are more to come. So the M23 and M33 are very popular? Yes yes so just to put you know throw a few numbers around just to give you some idea of this we believe we've got about 45 billion chips shipped containing Cortex-M based microcontrollers that's an awful lot of silicon. But most of those are not V8-M that's the newest of the newest right? That is yeah so what are they I don't know the breakdown of those so I couldn't say but but a very large proportion. V8 in the microcontrollers has been around for some time. And I was just looking at the press conferences recently with Qualcomm and the people using the Qualcomm Snapdragon 855 they're talking about 7 trillion transistors on a smartphone. So what's the latest with the AI in the smartphones? So pretty much all well actually all smartphones I know of the premium smartphones have announced that they are including neural network accelerators in those chips. Machine learning is very much a use case that is finding a home in those new smartphones and we're only really beginning to see those use cases take off. I mean everyone's doing smart you know face unlock and smart identification fingerprint identification things like that but in addition to that we're seeing a whole bunch of use cases kick off once you have downloadable applications which are utilizing these features. So we're seeing lots of awesome stuff especially with the camera but lots of awesome stuff yeah is it is it maybe true that we aren't seeing enough yet there's potentially lots more that is just imaginations. Absolutely we're just limited by our imagination but yes if you look at the machine learning use cases applied to smart video cameras people being able to give instructions to their video camera using voice you know zoom in on that guy blur out the background you know that sort of thing huge use cases for machine learning. All right.