 Welcome to Embedded World 2018, where we're focusing on ST innovation in the world of artificial intelligence. I feel this gentleman here, Daniel O'Pow, has been working in for 25 years. Yes, indeed. You've seen many developments over two decades plus. What's ST bringing to the party now? Well Chris, this is STM32CubeMx.ai. It's a real game changer because it helps our customer to be more productive and efficient by deploying any artificial neural networks they develop on STM32 microcontroller, any kind of. Listen, explain to me very simply, what's the big deal about putting a neural net on a microcontroller? Well, because this is a key technology enabling true intelligence close to the sensor for IoT to take benefit of. Because by 2020 there will be 50 billion of devices attached to the cloud, many hundreds of billion of sensors. And so we need really, and there is a strong market demand to have intelligence close to the sensor as possible. And where does ST come in then? Our customer experiences three types of gaps. First is productivity of software development. Second is interoperability between cloud and embedded system. And also because it takes a lot of skills to develop efficient neural network for resource-constrained embedded device. And STM32CubeMx.ai solves all those gaps in one-stop solution. For example, use a client server application that helps any customer to automatically convert any pre-trained neural network in a memory-efficient, fast library that can run on any STM32 microcontrollers. And then it runs as a simpler page in a browser. So all the conversions can be controlled with simplicity because all the machinery and complication is hidden in a server that we prepared. And we support the community of developers and 75% accordingly to measures on 2017. We're using exactly the off-the-shelf tool we interface and we interoperate with. And we are showing a different kind of application, audio processing, inertial processing running on different type of STM32 with different mix of memories and complexity in terms of operating frequencies. So can we see the tool chain in action? Yes, Chris. So I loaded the neural network on STM32CubeMx.ai. I generated an implementation that fit on the resources. I downloaded the network and put it on the STM32. Now I'm feeding the STM32 with an audio file, indoor audio sources. It will flash one because the network recognized that we are indoor. In a while it will be outdoor sound. Here we are and so flashing two times because it recognizes we are outdoor. In a second we'll go in car and so we'll flash three times. Here we are. That means that it's able to recognize indoor-outdoor and in car. Another example is human activity recognition. So let me connect the sensor tile and the network I have here with an interface, a graphical interface. In a while it will be connected and we'll show that here we are. I'm walking and now I'm running and here we are. Okay, take care. Thanks for the demo dinner. Thank you so much.