Public Funding for Mignon Technologies Ltd

The Digital Economy will ultimately see billions (potentially trillions) of cameras and sensors- so-called 'edge' devices- deployed to record images for security and collect data about: people's health; industrial processes; weather; pollution; traffic levels, amongst many others. Processing this data into information to provide useful insights will require ever-increasing amounts of computing power or network bandwidth to connect edge devices to data processing centres. New computing architectures are thus needed for edge devices to enable the digital economy to deliver its potential improvements to human health and wellbeing, the environment and the economy. Mignon, a Newcastle University spin out, are commercialising such an architecture, called a Tsetlin Machine. This is a machine learning approach driven by propositional logic and characterised by low complexity. Compared to conventional machine learning, it uses up to 10,000 times less power, with 1,000 times less latency. Mignon have demonstrated this using SRAM chips fabricated at 65nm node, trained to recognise characters (MNIST database). Alongside this, Pragmatic has developed a revolutionary semiconductor technology based on metal-oxide thin film transistors (TFTs) to create Flexible Integrated Circuits (FlexICs) fabricated on flexible substrates (e.g., polyimide). This approach cuts production times from months to days, typically at a fraction of the cost of Silicon ICs, and with 1,000 times lower environmental footprint than conventional Silicon fabs. In this project, Mignon will develop tools to automate hardware design, incorporating their Tsetlin Machine. This will automatically design hardware, based on end users with health, environmental or industrial manufacturing datasets. End users will be able to specify if they require processor speed or low power consumption. The project will first demonstrate training and inference capability using in-memory architecture, fabricated at 28nm node as an embedded silicon application, trained using the CIFAR-10 image classification dataset, at Newcastle University. We will then demonstrate that the tool can design TM inference hardware trained by a dataset of ECG samples and recognise atrial fibrillation events, as a representative example. The hardware will be fabricated as a FlexIC to serve as a useful demonstrator to engage end users and customers across economic sectors.

Many electronic sensors that monitor health, security and industrial processes (amongst other applications) are now linked to computer networks. This is collectively termed the Internet-of-Things, IoT. It is leading to better healthcare, greater security and more efficient industries by collecting and analysing data continuously. Experts predict that by 2030 there could be \>1trillion devices in the internet of things. However, current network bandwidth and computing power could soon limit development. This will first become a problem when sensors need to do a complicated task, like deciding what is in an image. Doing this within a device, rather than transmitting lots of image data over a network, is called _edge AI_. The most common technology in edge AI is called a neural network. This approach takes a lot of computing power. To improve matters, the partners on this project have developed a new approach called a Tsetlin machine. In this, different groups of electronic components decide their outputs based upon what other linked components outputs are. This is \>1000 times faster than comparable neural network AI chips and uses 10,000 time less energy to give the same answer. Once we have developed the technology further, this will be very useful in edge AI devices in the internet of things. To make money from this, we must first start to improve how it works, in this project. Other companies will then make computer chips for us. They will supply these to companies who build devices and pay us a royalty for using our idea. We only established as a company recently and are still working out how much money we will ultimately make. We think it will take 18-30 months before we can make any revenue from the invention.

Mignon's technology enables **ultra-low power, explainable, Artificial Intelligence within edge devices.** Artificial Intelligence (AI) is transforming daily life, from how we process information, to how we keep ourselves healthy and safe. As the ubiquity of AI increases, there is an increasing need for AI models to be ran outside the cloud and on devices. Experts predict that by 2030 there could be **\>1trillion devices** connected to the internet with a majority requiring AI capabilities like image recognition. However, current network bandwidth and computing power could soon limit development. Furthermore, the black-box nature of AI limits applications. AI within a device, rather than transmitting lots of data over a network, is called _Edge AI_. The most common technology in Edge AI is called a Neural Network. This approach takes a lot of computing power. Mignon is a Newcastle University spinout, to commercialise an entirely novel, ultra energy-efficient Edge AI coprocessor, based on an architecture paradigm called Tsetlin Machine. Mignon's technology will facilitate a new generation of AI-powered edge devices. Mignon's semiconductor technology implements ultra-low-power edge inference and for the first time on-chip AI training. Uniquely, Mignon's technology enables explainability in AI allowing for detailed understanding of how decisions are made from the chip level. Mignon has demonstrated a ~**10000-fold** lower energy consumption and over **1000-fold** lower latency than existing commercial incumbents, whilst maintaining the same high levels of accuracy. To ensure the UK and its global partners benefit from Mignon, we must ensure that the technology is accessible to engineers. This project will allow engineers to experiment and build on Mignon's technology, by making it available to them through software virtually using _the_ _cloud._ Therefore, they can understand how it works for them, start using it in their own projects, and eventually licence the technology from Mignon to use in physical devices. Once developed this _emulator_ will be useful in the design and development of edge AI devices for IoT. This project will set a foundation for developing an ecosystem for this new technology with a centre-of-gravity within the UK. Mignon believes that in commercialising this technology it has the ability to revolutionise the way AI is utilised in a new generation of intelligent devices, bringing about a meaningful improvement in the UK's semiconductor industry, with significant global impact. We think this project will accelerate that, taking less than 6 months to get the technology in the hands of other engineers.