Deep learning
Deep learning is at the core of AI. It enables computer systems to learn new knowledge and improve their functionality through experience rather than by being programmed. Each new experience has the potential to be converted into significant improvements across a range of functions. Deep learning is made possible by the advanced data processing and analytics capabilities which are built into the system at the outset. What sets deep learning apart from standard incremental learning is the fact that the computer's learning ability also improves with experience. At present, American Express is using advanced machine learning algorithms for real-time fraud detection, while car manufacturers are using it to predict when vehicles will need a service and what parts will be most in need of attention.
AI chips
AI may be powerful but it is also immensely greedy for computing power. The average chip found in your phone or PC will not have nearly enough capacity to support true AI. This has led the leading manufacturers to design and develop specialised AI-enabled chips which are capable of carrying out the enormously complex mathematical tasks required to deliver AI functions such as facial and speech recognition. The next generation of smartphones, intelligent medical devices and high-spec cars will all feature these AI chips.
Chatbots
Most people who have used online helpdesk services have probably experienced a chatbot, whether they know it or not. A chatbot is an automated assistant programmed with answers to thousands of the most frequently asked customer questions. When it comes across a question it doesn’t recognise, it hands the conversation over to a human. The latest generation of AI-powered chatbots learn as they work and over time hand over fewer and fewer calls to human team members. The very latest incorporate speech recognition and voice synthesis and can carry out pretty realistic telephone conversations.
AI ethics
The famous trolleybus question is often wheeled out – pun definitely intended – in relation to autonomous vehicles. If an autonomous car’s AI-powered systems can predict it will be involved in an accident with a bus full of people and calculates that the only way to avoid it will be to crash into a wall, potentially killing its own passengers, which option does it choose? This may seem like idle dinner party stuff but it is the subject of real debate within the tech industry and at government level around the world. If AI is to replace human workers, will it be designed in a manner which reflects society’s ethical values? Already, there is talk about the development of standards in relation to legal responsibility for harm caused by AI-controlled systems and the protection of basic human rights from infringement by those same systems.
Neural networks
This may sound like something from a Star Trek android but the ability to create a computer network capable of emulating the human brain is already here. Examples of what such systems can do is read handwriting which it had never encountered before, compose music in the style of any composer or in a totally original form, predict stock market and currency fluctuations and, most usefully, perform online medical diagnoses through combining vision- and audio-processing with chatbot capabilities.