Artificial Intelligence is no longer the future, it is the present in our living rooms and, often, our pockets. As the technology continues to evolve, an important question emerges: what level of trust can—and should—we place in these artificial intelligence systems? The answer is simple, Microsoft-Responsible Machine Learning Capabilities are redefining trust in AI systems.
Instilling human values in AI
As artificial intelligence becomes more pervasive, so too has the concerns over how it can be trusted to reflect human values. Generally, one of the most common examples that get cited is to show how challenging it can be when an autonomous car has to avoid a collision. Let us assume that there is a speeding truck coming towards a driver who has to quickly swerve to avoid getting hit. However, there is a possibility that the car may hit a baby if it swerves on the left-hand side and a pregnant woman if it swerves on the right-hand side - what should the autonomous car do? The challenge lies in the fact that machines tend to get biased as the training data fed to them may not be a complete representation of what should be taught to them.
To trust computer decisions, ethical or otherwise, people need to be sure of how an artificial intelligence system arrives at its recommendations and suggestions. From a privacy point of view, it is always good to have a choice to tradeoff between utility and privacy so that every user can make that choice for themselves.
The societal advantages and implications of AI are enormous. Thankfully, Microsoft-Responsible Machine Learning Capabilities can easily alleviate all concerns, such as the ability of AI systems to make critical decisions in a way that are perceived as fair by humans, the capability to explain its reasoning and decision-making, and to be aware and aligned to human values that are relevant and specific to the issues on hand.
Without answers to these questions, a majority of humans will not trust artificial intelligence systems, and therefore they will not make use or neither fully adopt AI's positive capabilities. Microsoft-Responsible Machine Learning Capabilities answer all these and many more questions as the capabilities are designed around the principles of fairness, explainability, and traceability.