NECPUC 2017: Recap

Last month, I was honored to be a panelist at the New England Conference of Public Utilities Commissioners (NECPUC). Thanks to our moderator, Massachusetts Commissioner Karen Charles Peterson, we had a great discussion about the opportunities and challenges facing artificial intelligence (AI) applications in regulated industries, like energy and telecommunications.

By Colin Gounden

Often AI is thought of as science fiction. But, the truth is, AI is already pervasive (well beyond speech recognition, like Apple’s Siri and Netflix’s recommendation system). In fact, you likely already use some kind of AI every day whether you are aware of it or not.

Beyond the consumer level, high-stakes organizations in regulated industries are increasingly incorporating AI into regular operations. For example, a utility company that wants to plan long-term investments in their critical infrastructure may use AI applications to determine which assets are most vulnerable.

Critical decisions, like those that affect the lives of a utility’s customers, require trust of the AI application’s predictions and recommendations. Building this trust and credibility is perhaps the biggest hurdle facing AI developers as organizations incorporate these new technologies into their operations. I recommend that developers start by prioritizing client data security. We know that AI requires data to work, but organizations are often reluctant to share their proprietary information. Providing a secure, efficient way to transfer data that serves the needs of both developers and clients is absolutely essential: enter blockchain. Blockchain has the potential to change the landscape in ways we are only starting to imagine – similar to the internet’s impact on daily life. Its ability to securely share and store vulnerable information makes it a significant innovation for the continued development of AI.

Transparency, or the “why” behind AI predictions and recommendations is another important component of building trust. Via Science works specifically to address this concern with its explainable AI systems. Most deep learning has an input and an output, but the in-between steps are a mystery. One of our key differentiators is to make the logic behind AI recommendations more transparent to people so that companies and the regulators that oversee the industry have greater trust in the decisions being made.

While developers can secure client data and provide transparency, utilities and other high-stakes, regulated organizations may still be hesitant about integrating AI. Pilots are the best way for these organizations to familiarize themselves with the potential impact of AI applications and value its potential. In fact, pilots are a growing trend across utilities in particular, as more and more consider turning to AI for modernization initiatives.

To view my full presentation from NECPUC, please visit their website.