As part of TechCrunch’s ongoing Ladies in AI collection, which seeks to provide AI-focused girls teachers and others their well-deserved — and overdue — time within the highlight, TechCrunch interviewed Marissa Hummon, the chief know-how officer on the vitality firm Utilidata, the place she is working to make the electrical grid extra sustainable.
“The work that I’m doing at Utilidata is pushing in opposition to the established order of the utility trade,” Hummon instructed TechCrunch, including that AI is poised to assist the utility sector function extra successfully. “It’s going to give utilities the instruments they should function a clear, fashionable, and dependable grid that may higher serve the folks and companies linked to it.”
Hummon began her profession on the Nationwide Renewable Power Lab, the place she targeted on the best way to transfer the vitality trade away from carbon-intensive options to extra clear sources. She recalled having to take a brand new modeling strategy as a result of sure physics equations have been “unsolvable utilizing conventional analytic strategies.”
“As an alternative we would have liked to make use of numerical strategies and machine studying,” she mentioned. This was again in 2010.
Recommendation to girls
Hummon moved her method up, working on the vitality firm Tendril earlier than touchdown the CTO function at Utilidata, a place she’s held since 2018. She acquired a PhD in utilized physics from Harvard and mentioned that as she made the transition from grad faculty into the engineering area, she observed a major drop within the variety of girls working within the area. “I’ve typically felt the strain of being held to a better customary within the office in comparison with my male friends,” she mentioned.
Hummon mentioned now as a frontrunner she tries to steer by instance and create alternatives for girls to steer and be acknowledged. On the identical time, males must be accountable for creating house for extra girls to enter and keep within the area, she mentioned.
“The change we want is not only in observe; it have to be systemic and widespread,” she continued. “It begins with recruiting and hiring, continues with mentoring and training, and culminates with truthful and equitable recognition and promotion — all whereas making a protected and inclusive office.”
Her recommendation to girls getting into the AI area is to at all times keep in mind that being a lady could be a bonus. “It’s given you a perspective that’s completely different out of your male friends and breakthroughs in know-how at all times come from distinctive views.”
She says to hunt alternatives with corporations which have proven their dedication to range and which have leaders who’ve showcased their efforts to help girls and minorities. “Choose a possible employer by the way in which they present up within the interview course of, not by the statements or the stories they placed on their web site.”
Constructing moral AI
Hummon mentioned that customers ought to know that AI will not be one thing that may remedy all issues however relatively is an knowledgeable assistant to boost operations, enhance effectiveness, and assist help.
“Recognizing its limitations and guaranteeing there are correct checkpoints is vital,” she mentioned.
She mentioned it’s at all times necessary to know how generative AI was skilled and constructed with a view to perceive any potential biases it might have. “This data is a strong instrument when working with language fashions for knowledge evaluation and evaluating the feasibility of options,” she continued. “AI is just pretty much as good as the info and data it was skilled on.”
She additionally gave some recommendations on the best way to construct protected AI, explaining that Utilidata made an early resolution to maintain all of its knowledge native. “Approaching mannequin constructing utilizing distributed AI computing reduces the quantity and frequency of knowledge being transmitted to the cloud and subsequently reduces the possibilities of a safety or privateness breach.”
As for traders, she mentioned — like the entire consultants we spoke to — they should consider how an organization desires to make use of AI, particularly since accountable AI seems completely different in each trade. “The funding in accountable AI needs to be commensurate with the chance and complexity of any given firm, not an across-the-board customary.”
