Though still in its relative infancy, artificial intelligence (AI) is already impacting many business functions. Recent applications of AI capabilities demonstrate how machines are able to display human-like decision-making skills when adapting to unknown environments or situations, and these skills are being used for real-world business applications. As machines become smarter and more efficient, AI's influence will continue to grow.
Key insights
-
AI and machine learning are still in their infancy, but already able to perform impressive tasks posing enormous and complicated consequences on multiple industries, including financial services
-
Current applications of machine learning and AI are able to adapt and learn based on real world environments, without requiring explicit instructions and human inputs
-
AI technologies are learning to make predictions and decisions that go beyond the limitations of the information they are explicitly provided
-
While some jobs will be displaced as a result of AI and automation, there will be many new ones and existing roles will likely become more efficient and provide greater value when paired with these new technologies
Artificial intelligence and machine learning intersect
Once the sole province of science fiction, AI has grown into a field with incredible potential. Cameron Schuler, executive director of the Alberta Machine Intelligence Institute (Amii), provided a preview of the AI revolution during a presentation at RBC Investor & Treasury Services' recent Investor Forum.
Schuler explained that unlike traditional systems that are designed to execute on specific commands within a closed ecosystem, recent developments are enabling machines to adapt to real world circumstances. “Simply put, our goal is to enable computers to make good decisions in ambiguous and uncertain environments," said Schuler. The director added that Amii aims to achieve this objective through a combination of machine learning, which he defines as a machine's ability to learn from and adapt to its environment, and AI, or “advanced computer systems that can perform tasks that normally require human intelligence to achieve."
Forecasting through machine intelligence
There are a wide variety of applications for machine intelligence, but one of the most significant for business leaders is its ability to help them make better decisions through advanced forecasting. “You need historical data if you're going to do forecasting, and you need to be able to understand how that forecast changes as it's influenced by the future," explained Schuler. He described reinforcement learning, which enables computers to learn from experience and which underpins a next-generation forecasting technology called "temporal difference learning: looking at the future and using that as a data source." Schuler noted that intelligent systems are now able to adapt their forecasts based on future potential outcomes.
As an example, Schuler asked the audience to imagine autonomous vehicles driving in poor weather conditions through a school zone. “When you start thinking about predicting surprise, thinking about risk management, there's a potential to say, 'the system should have heightened awareness in this type of environment', because something unexpected could happen here," he said. “It may be a false positive, but nonetheless the ability for a computer to recognize that it should be aware that something is changing is important.“ It is this heightened awareness in potentially precarious environments that gives machine intelligence the ability to improve on traditional forecasting practices.
Supervised to unsupervised learning
One of the most significant innovations in machine intelligence is the evolution from supervised to unsupervised learning. Supervised learning, explained Schuler, is when a machine is trained on labelled data, but there are challenges with introducing new data the system has not encountered during training. Machines that are capable of unsupervised learning, however, can collect data independently, and adapt over time. As an example, Schuler played a video of an automated system designed by the Institute's researchers, Michael Bowling and team. The automated system controls the joystick movements of a classic Atari video game, and its goal is to improve its score over time without explicit directions or a strategy on how to do so. The video shows the system's incremental improvements until approximately 600 games later, when it has optimized a gameplay strategy.
One of the most significant innovations in machine intelligence is the evolution from supervised to unsupervised learning
“What's exciting is that this system is actually discovering and learning," said Schuler. “It can't make you coffee or mow your lawn, it's not general intelligence, or something you have trained, but it's learning on its own. That's unsupervised learning and, specifically reinforcement learning." Other reinforcement learning systems can make predictions between 40,000 and 60,000 times per second, said Schuler, and many can run off of a standard laptop. He adds that similar technologies, which are capable of predicting, evolving and learning, can be applied to traditionally closed computer systems as well.
Towards an AI future
While there are concerns about the social impact of these new technologies, particularly regarding the potential for automation and AI to displace jobs, Schuler envisions positive changes for the future. He likens the AI revolution to the agrarian technologies that enabled farms to run more efficiently and safely. He believes that traditional roles will either evolve into new opportunities or become more dynamic with the help of AI. “Think about all the mundane things you don't like to do, and the things you could automate easily, like auditing," he said. While some will be displaced by AI, quality of life, Schuler said, will increase overall, and many in attendance agreed. According to an audience poll, 72 percent believe that AI will improve their quality of life over the next five years, 22 percent disagreed and six percent said it was too early to tell.
Traditional roles will either evolve into new opportunities or become more dynamic with the help of AI
Computer systems have historically operated within a closed system, taking commands and performing tasks based on information they have been explicitly provided. Work like Amii's shows machines will be able to adapt to unknown environments in the not-so- distant future. These systems will not only enable organizations to better forecast future potential outcomes, but also allow them to scale and evolve tools and applications in real time, according to real world conditions. Industries currently sit at the dawn of the age of AI, and before long these systems will become integral to business operations.
Source
RBC Investor & Treasury Services' Investor Forum (May 26, 2017): Artificial intelligence and the future of business