Embedded AI – what does it mean?
Posted on 22/06/2020
The government report “Growing the artificial intelligence industry in the UK“ noted:
Increased use of Artificial Intelligence (AI) can bring major social and economic benefits to the UK. With AI, computers can analyse and learn from information at higher accuracy and speed than humans can. AI offers massive gains in efficiency and performance to most or all industry sectors, from drug discovery to logistics. AI is software that can be integrated into existing processes, improving them, scaling them, and reducing their costs, by making or suggesting more accurate decisions through better use of information.
It has been estimated that AI could add an additional USD $814 billion (£630bn) to the UK economy by 2035, increasing the annual growth rate of GVA from 2.5 to 3.9%.
Our vision is for the UK to become the best place in the world for businesses developing and deploying AI to start, grow and thrive, to realise all the benefits the technology offers.
AI is best known for analysing data to make predictions. The algorithms are run on large server platforms, usually in the cloud, and they have been used for a wide range of applications: from predicting trends, diagnoses disease or even playing TV gameshows.
IoT represents the trend for ubiquitous connectivity between devices – from smartphones to individual sensors. This data is sent to a central point for analysis and action. It is believed that there are currently approx. 10 billion IoT devices – and Business Insider predicts that there will be 41 billion IoT devices by 2027.
However, recent advances in electronics – particularly specialist chips for processing AI – are allowing product developers & designers to combine these two technologies and embed AI processing in products and deliver a new range of capabilities.
This emerging sector of Embedded AI has been called the “Artificial Intelligence of Things” or AIoT.
Drivers for AIoT
The major drivers for this new sector are:
- Increased security and privacy: if critical or personal data doesn’t have to be transmitted to a central processor it is automatically more secure. AI on the machine also allows improved access security – such as facial recognition on the latest smartphones.
- Reduced data transfer: currently data collected by a sensor or device is sent to a central computer for analysis. The results are then sent back to the device for action. If a device can process the data locally and only send the results to the cloud it makes a dramatic reduction in data.
- Increased responsiveness: transferring data and waiting for the result takes time. There are situations – e.g. autonomous vehicles or safety systems – where this delay may be critical.
- Increased resilience: removing the requirement for the data transfer improves the resilience of a machine as it can continue to operate and send data to the main system when the link is re-established.
- Increased functionality: by building AI into the machine designers can provide new functionality. This may allow the machine to self-diagnose faults or improve control – e.g. being able to use voice or gesture control.
To deliver these benefits development engineers need to learn new skills – in embedded development and system design.
An Engineers viewpoint
XMOS – a UK AI chip designer – recently surveyed 200 electronics engineers and found that they had a good understanding of the opportunity for AIoT.
82% of electronics designers believe that AI will enable them to increase the competitive advantage in the products that they design. 44% of engineers said that AIoT is critical to improving the way we interact with technology and 40% said that AIoT will radically change technology for the better.
The survey also did an analysis of the applicable markets and found that the top four sectors to benefit from AIoT were: Manufacturing/Industry 4.0, Smart Homes, Healthcare and Autonomous Vehicles.
Deloitte published its Technology, Media, and Telecommunications Predictions 2020 included a set of predictions for the Edge AI chip sector.
They predict sales of more than 750 million edge AI chips in 2020 representing US$2.6 billion. This represents a compound annual growth rate of 36% from the 300 million devices sold in 2017.
By 2024 they expect sales to exceed at least 1.4 billion devices. This would represent a growth of at least 20% – more than double the predicted 9% growth in the overall semiconductor market.
The UK’s position
The UK has strength in AI algorithms supported by investment from government.
The UK also has a significant history in micro-electronics and chip design – as exemplified by ARM and Imagination. Both specialise in power-efficient devices – an essential requirement for AIoT. UK companies are producing specialist chip designs targeted at processing AI e.g. Graphcore and XMOS.
This combination of both hardware and software expertise gives the UK an opportunity to establish a strong industry sector and a resilient supply chain.
As demonstrated by the XMOS survey and the Deloitte report using embedded AI will enable increase efficiency, security and resilience in all industry sectors.
What is KTN doing to assist?
We are running a series of webinars for people responsible for R&D from CTOs to engineers. To ensure that we bring industry and academia together we are working with the EPSRC funded eFutures network – who focus on electronics research in the UK.
The first webinar covered the Hardware Issues of implementing Embedded AI – from academic challenges through to industry needs. The second webinar focussed on “Running AI at the Edge”. These webinars have had over 400 delegates – demonstrating the strong interest in this technology. An interesting aspect is that using webinars has allowed us to form international engagements – with both the speakers and delegates.
Our third webinar will focus on Vision Systems, you can find out more here.
In parallel with these webinars we are running technical workshops – strongly focussed on development engineers – to cover how to implement embedded AI.