Artificial Intelligence
Navigating AI application requirements to select the ideal development board
In today’s data-rich world, AI is becoming an increasingly common tool to improve many different kinds of applications. For engineers who are not AI experts, using a development board can help accelerate the design cycle – but with many alternatives available, selecting the ideal board can prove daunting. Ankur Tomar, regional solutions marketing manager for Farnell element14 outlines some of the key considerations and introduces a new online configuration tool designed to simplify the process
A
rtificial intelligence (AI) has been a hot topic and a regular feature on most tech trends lists over the last
few years. Yet despite its rapidly growing prominence, AI hasn’t become an emergent technology trend overnight. AI has been around for more than half a century and it is growing access to vast amounts of data over the past decade that has seen AI evolve to increasingly embed itself in our everyday lives. The move away from the early goal of developing artificial general intelligence (in other words, human-level cognition, known as ‘strong’ or ‘full’ AI) towards more pragmatic forms of ‘applied’, ‘narrow’ or ‘weak’ AI is there AI is delivering real benefits to users and focus on solving specific, defined problems has seen AI applied to growing numbers of diverse use cases.
Types of AI applications Often, technology developed and matured in the course of AI research – from facial or speech recognition, and online recommendation engines, to chatbots and virtual assistants – are integrated into everyday applications without even being called AI. ‘Modern’ applied AI isn’t simply
one technology for a single specific application, but a range of techniques for many different applications, each with varying requirements. Some of the most common include: • Predictive maintenance – aggregating data from multiple sources and applying AI to anticipate equipment failure before it happens • Voice/speech/sound recognition – smart devices (including phones and speakers) can now recognise, interpret and respond to voices, speech and even other everyday sounds
• Motion/object/human recognition – as well as being deployed in security systems, this technology is critical to the development of autonomous vehicles and smart infrastructure • Machine learning – this technique fuels many different forms of AI, applying algorithms to raw, unrefined data to uncover implicit patterns, before using them to predict future outcomes
Tools for developing AI applications As AI becomes more pervasive and developers realise the benefits AI can bring to their applications, products and services, there is still a gap in terms of AI expertise and experience. To help close the gap developer tools have been designed to simplify the integration of AI functionality – from software tools and services to development boards. Using a development board can reduce the time to build an application, but it can be a complex process to select the ideal board. The following factors should be considered:
Processing power The level of processing power required will depend on your application:
36 December 2018/January 2019 Components in Electronics
• Processing from Cloud to gateway to edge – Until recently, AI applications have mainly undertaken processing in the cloud, providing access, for example to high- quality, cloud-based services, such as Amazon’s Alexa or Google Voice, for voice recognition and reducing the level of processing power required in the local device. Nevertheless, there are cases where it may be impractical to process in the cloud, for example, if there is limited bandwidth available. Other options now available mean that devices can undertake processing at a gateway or even at the edge itself. Gateway devices, such as the new Ultra96 development board from Avnet, provide huge processing power to support multiple local devices whereas devices for processing at the edge generally require less processing power. This is due to the fact that algorithms can be created in the cloud prior to implementation to allow for much simpler processing to occur for ongoing monitoring at the edge itself. • Learning or processing-intensive – machine learning technologies currently dominate AI, but not all applications need to include the learning element (for example, existing voice recognition engines have already been optimised). Where an application is new, often the system will be required to learn, as well as simply process, requiring more processing power. • Other forms of AI – Rule-based AI systems may need far less processing power than ones that use machine learning.
Sensing capabilities It is important to choose a board with the right category of sensors for your application. Since most forms of AI try to interpret the real world, applications typically
require sensors of some kind, to offer: • Environmental sensing • Motion sensing • Proximity sensing • Audio sensing • Image sensing
Connectivity
AI applications need to communicate. This might require short-range communications (such as Bluetooth or Zigbee), to link to sensors or local gateways, or longer-range communications (typically WiFi or LPWAN technology, such as LoRa), to connect to more distant gateways. Some applications will need both.
AI Configurator To help developers identify the right development boards for their AI devices, Farnell element14 has created an online tool designed to help engineers confidently select the ideal board for their artificial intelligence projects.
Users can select the application that most closely match their requirements from a list including predictive maintenance, voice recognition, object or human recognition, motion recognition and machine learning and add requirements for sensing and wireless connectivity. Once the requirements have been defined, the AI Configurator identifies boards (including add-on boards necessary to provide the required sensing capabilities) that meet the needs of the design; as well as relevant resources to help speed up the design cycle, including tutorials, case studies or technical documentation, such as Alexa API documentation for voice recognition applications.
uk.farnell.com/ai-configurator
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
