Machine Intelligence Quality Characteristics for Better AI Solutions
From predictive maintenance and logistics optimization to customer service management, the applications and economic potential of Machine Learning, Natural Language Processing and Robotics makes Artificial Intelligence one of the most important advancements of the digital era.
Indeed, the AI market is expected to grow from USD 21.46 Billion in 2018 to USD 190.61 Billion by 2025, at a CAGR of 36.62%. Interestingly, the largest proportion of AI revenue comes from the AI for enterprise applications market, and businesses that are creating and using high quality AI solutions are marching ahead of their competitors.
Measure, Predict, Control
Whilst the business and customer benefits of Machine Intelligence and AI are clear, building these advanced products and software solutions presents significant challenges in terms of risks and costs. In the medical, construction or emergency sectors, for example, failure of a Machine Intelligence solution could, of course, even have fatal consequences. So, how do we instil the highest level of quality assurance into these Artificial Intelligence systems from the outset? How can we measure, predict, and control the costs of software development and assure the performance, usability, security and safety of the end AI product?
Traditional Quality Characteristics
Machine Intelligence and AI Quality characteristics need to be defined and utilised at the earliest opportunity during the design and engineering of an AI solution, to enable engineers to assess the systems strengths and weaknesses. Of course, the ISO 25010 already provides us with a standard set of existing software quality characteristics for conventional IT. These include Functionality; Performance; Usability; Reliability; Security; Maintainability; and Portability, and these are still relevant and essential to success in an AI product. However, at Sogeti, we believe that the complex architecture of Artificial Intelligence solutions requires additional quality characteristics that are specific to Machine Intelligence. We have also developed new test approaches and techniques to test these characteristics, and we are both testing AI and testing with AI.
Behaviour, Morality and Personality
We’ve identified 3 new main Machine Intelligence characteristics: Intelligent Behaviour; Morality and Personality. We’ve also created 12 MI sub-characteristics including the ability to learn; collaboration; ethics; empathy; and humour. These enable designers and engineers to prioritise AI product and solution attributes; predict issues before they arise, and create accurate, bias-free requirements and system design criteria. There are huge business benefits to applying these MI characteristics, including reducing long and short term risks; improving reputation management and enhancing the overall quality, usability and performance of the AI system.
AI Quality & ROI
In a new Report “Machine Intelligence Quality Characteristics - How to Measure the Quality of Artificial Intelligence and Robotics” Sogeti AI experts Rik Marselis and Humayun Shaukat, take an in-depth look at these MI Quality Characteristics and examine how they are different from, but connected to, those for conventional IT.
The Report defines what quality means in a world of Machine Intelligence; examines Testing of and with AI; demonstrates how to map your quality characteristics to the 6 Angles of Quality for AI and Robotics; and examines the risks and challenges of creating AI products and solutions and how to overcome them.
If you’re keen to discover how you can incorporate quality into your AI solutions from the very start and get a good return on your AI and MI investments, then download your free copy of the Report here.
 marketsandmarkets 14/02/2018 Artificial Intelligence Market by Offering (Hardware, Software, Services), Technology (Machine Learning, Natural Language Processing, Context-Aware Computing, Computer Vision), End-User Industry, and Geography - Global Forecast to 2025