As enterprises work to gain maximum value from artificial intelligence (AI), there are important questions being asked: Can humans trust AI to make decisions? If so, how can we know if the intelligent agent made the correct one?
A classic example of this conundrum was brought to the world’s attention a couple of years ago when one of the best players of the complex ‘Go’ board game competed against AlphaGo, an AI-based computing system built by Google. At one point during the game, the AI system made a move that made no sense to the player, the expert commentators in the room or the millions of people watching a livestream of the game.
But ‘Move 37’ – which all the complex carbon-based life forms thought was a mistake –became the turning point in the game that AlphaGo eventually won.
Google’s researchers had no prior knowledge the system would choose this path. They had put the system through extensive “training” to refine its play, though, including competing against different versions of itself. With that accumulated data, AlphaGo was able to look ahead to the potential results of each move, estimating the probability that this particular maneuver would stump its opponent and result in a win.
AI’s Role in Mission Critical Decision Making
But what happens when you apply AI’s predictive capabilities to mission-critical situations such as deciding how an autonomous vehicle should react to a sudden road hazard? Or less urgent but still significant business questions such as deciding which company to acquire?
If an AI system makes a counter-intuitive recommendation like the AlphaGo example, it is imperative to help people understand how it came to that conclusion. People need a way to understand and validate the system’s decision-making process. This is necessary to ensure peace of mind for the company as well as meet the requirements of industry regulators. Fortunately, there are several ways to ensure recommendations from AI systems are vetted and delivered by humans.
Human Decision-Making Backed by AI = Optimal Business Results
At mCloud, we’ve already implemented procedures to get the most out of AI systems while maintaining human decision-making for optimal, verifiable business outcomes in our key verticals: energy efficiency for small commercial spaces, blade inspection for wind farms, and enhanced operations and maintenance for oil & gas and other process facilities.
To start, we maintain a continuous history of all the data that's connected to the system. When we set out to solve issues using AI, we're very clear about how the system arrived at a specific recommendation. If the answer is not immediately apparent, we take a step back and review the data, especially when it comes to conclusions that appear remotely questionable.
There is always human review and control in all our AssetCare solutions. The AI system simply provides recommendations supported by data to help drive business decisions and put the right processes in place that deliver measurable benefit. The rest is up to the decision-making capabilities of those complex carbon-based life forms we know as people.
For example, while we apply AI to assist with the day-to-day management of HVAC systems at quick-service restaurants, our Live Operations team always has human supervisors monitoring the decisions that are made regarding energy use.
We tap into the power of AI to help us manage complex webs of data, understanding underlying patterns and making meaningful predictions. For mission-critical situations, we maintain a human in the driver’s seat, with AI playing the role of digital advisor. This is how we harness the potential of AI without displacing the very important role human context and judgment plays at oil and gas facilities and more.
To learn more about how mCloud integrates AI in its AssetCare solutions to ensure enhanced performance in smart buildings, wind farms and oil & gas facilities, contact us today.