EPSRC Announces £440m for Doctoral Training
Funding for 75 Centres for Doctoral Training (CDTs) under the umbrella of the Engineering...read more
Artificial Intelligence, or machine learning, is steadily working its way into our lives. Businesses are increasingly confident of using AI to increase customer engagement with giants Microsoft, Apple, Google and Amazon leading the charge. Amazon’s Alexa experienced failure on Christmas Day as their servers were overloaded by the surge in users plugging in their new personal assistants across the globe. While these personal assistants are increasing in the home, AI in business is experiencing a similar boom thanks to developments in machine learning.
Machine learning is an type of AI, allowing systems to learn and develop without being programmed directly with the information. For example, a computer can learn to recognise specific things in images, like pedestrians or cars. The computer scans thousands of these images of these things until it learns to recognise them with great accuracy in new pictures it hasn’t encountered before. This simplified example above is one method, but as the old saying goes – Give an AI a picture of a fish and it recognises it for a day, teach an AI to create pictures of fish and it will recognise them for a lifetime.
Imagine a driverless car that teaches itself by generating countless road-based scenarios while it’s still parked. A far better learning experience than going out on the road time and again with a human overseer. Getting AI to think rather than simply recognise is already well under development. Modelling AI after the neural networks of the human brain has led to “deep learning”, a type of machine learning encouraging AI to get creative and develop an imagination of its own. Deep learning is also becoming a sought-after skill in AI development jobs, especially for unmanned vehicles.
Machine learning and deep learning involve AIs programmed to receive and learn from data provided by human overseers. Another breakthrough in AI development is getting AIs to develop themselves. Generative Adversarial Networks (GAN) challenge two AI neural networks to duel each other. The duel can be anything from “who can generate the most accurate picture of a person” to “who can win a game of Go”. The AI have already learned how to recognise these images through machine or deep learning, and GANs have an increased ability to generate novel content. They do still get it wrong, from time to time, so human programmers aren’t out of a job yet.
A large fear with AI is job losses, with a lot of numbers being thrown around with little fact to back them up. The reality is AI will certainly do away with many jobs just as automation did at the turn of the 20th century but will undoubtedly create many more while making existing jobs simpler. Research and development is the second most popular field for the use of AI services, after IT, and we can expect the trend to continue in 2019. R&D has broad scope for AI use in software development, healthcare and scientific research.
In the UK, AI forms a pillar of the national industrial strategy and a robust support network is in place for its future development. Former science minister Lord Drayson points to the UK’s leadership in healthcare and how AI can be used to revolutionise drug discovery and assist in disease diagnosis. UKI2S portfolio companies Antiverse, Eagle Genomics and Synthace are prime examples of UK software companies using AI to assist in experimental design and data analysis. These companies have made great strides in 2018 and the year ahead looks promising for AI.