Three perspectives on artificial intelligence
Artificial intelligence (AI) is bestowing machines with degrees of human intelligence. The technology is accelerating exponentially; businesses are swiftly applying it; the stock market is struggling to separate hype from reality. Below are three expert perspectives on AI’s prospects.
1. Technology: accelerating development
In the past six years, the technologies behind AI have rapidly progressed, allowing its development to accelerate exponentially.
The breakthrough came in 2012, when a deep neural network called AlexNet allowed computers to identify objects in real images. Computers find it difficult to recognise pictures. But since then, there have been quick improvements and machines are almost as good as humans at object recognition. They can also recognise language. “With perception solved, AI research has moved on to higher-level cognition tasks like planning and reasoning,” explains Prof. Dr. Damian Borth, Director Artificial Intelligence and Machine Learning Lab, University of St. Gallen.
A recent development, for example, is deep reinforcement learning, a neural network that interacts with its environment and gets a reward or punishment that teaches it to learn.
Notably, in March 2018 Google announced Auto ML, which allows two of the three steps in building a neural network to be done by machine. “That is exciting, as it means you can accelerate development, making the technology much cheaper,” notes Mr. Borth. “If you can industrialise the development of intelligent networks, AI will soon be a commodity technology.
Damian Borth argues that the impact will be like the internet. There will be companies that are early adopters like Amazon and Google. Those that react more slowly will lose their competitive edge.
“But we are far away from a machine that can generalise and has a consciousness,” he concludes. “These machines are self-learning systems but they are experts in their narrow domains and simple-minded compared with humans. They are very good at specific tasks but bad at others. A neural network that can recognise spoken language can’t drive cars. A neural network that can drive cars can’t play chess.”
2. Business: great potential in every sector
As the technology evolves fast, organisations from many sectors are applying it to drive productivity and insight. As a first step, many organisations are deploying chat bots to answer simple client queries, freeing staff to concentrate on giving excellent service in more complex areas. Another example might be robotics, which answer client emails or employees’ questions to HR.
More specifically, IBM’s Watson AI platform has been acting as an assistant to oncologists in more than 230 hospitals and healthcare organizations, covering 13 different types of cancer. IBM has also developed a robot called CIMON® (Crew Interactive MObile CompanioN), a joint venture with Airbus, that is assisting the astronauts on the International Space Station with their experiments.
But where is this all leading? “I think we will see AI optimise industries across the board – in automotive, banking, telecommunications, manufacturing, and especially healthcare given the wealth of information and willingness to share it,” says Matthias Hartmann, Chairman of the Board of Management, IBM Germany. “Every job will be influenced by AI. That does not mean necessarily AI takes away the job; but assists.”
Over time, Hartmann anticipates that advances in AI will be matched by improvements in quantum computing, leading to great advances in areas such as medicine discovery or investment portfolio optimisation. He also foresees AI leveraging the explosion in big data to increase knowledge exponentially.
“Today you can only Google 20 per cent of the data out there. Eighty per cent of the data is held within enterprises. You need a platform to make this data accessible. My clients see lots of business advantages by applying their own data.”
3. Investment: divorcing hype and reality
Just like the Internet 20 years ago, the AI investment story is being hyped. There is a danger that AI is already being viewed too soon as the answer to many problems, causing disappointments. In fact, many of today’s applications are relatively narrow and need higher investments and iterations of improvements than the excitement suggests.
“In the long term, we do think that AI is being under-priced in healthcare, where it is improving medical diagnostics and research outcomes,” asserts Fabiano Vallesi, Next Generation Portfolio Manager at Julius Baer. “Similarly, in transportation it’s improving safety in general, enabling the basics for self-driving cars.”
But where are the interesting investment areas for today? “We think that the most attractive segment is the integrated cloud computing providers, which are building platforms facilitating access to basic AI tools,” explains Vallesi. “They will commoditise AI fast and give access to these tools for everyone. But also, the cloud software providers will be far more productive and available with AI enhanced solutions. We do not favour the hardware makers and semiconductors, as we think their stock valuations already discount their prospects.”