Ottawa, Ontario (CAN)-based AI governance firm NuEnergy.ai has secured a U.S. patent on its Machine Trust Index (MTI) methodology, a standardized measurement for artificial intelligence (AI) oversight.
The milestone comes as competition intensifies in the complex, quickly evolving field of AI visibility and explainability, giving NuEnergy an additional bragging right and legal moat. Its MTI framework was developed in 2018, even before the current rush of investment in generative AI.
As Harry Major, NuEnergy’s VP of Software explained in a release last week, “AI is transforming myriad industries and enhancing efficiency through innovative solutions. However, as these systems become more integrated into our daily lives, ensuring their ethical and responsible use is paramount.”
Quantifying accountability at the highest levels
Founded in 2017 by Niraj Bhargava and a team of technical and policy experts, NuEnergy recognized emerging challenges around artificial intelligence required practical, cross-disciplinary solutions.
As Bhargava recounted in an exclusive interview with VentureBeat: “We decided to be focused completely on governance and be an independent third party when it comes to governance, because many companies claim they understand ethics and responsible AI, but they’re governing their own AI and marketing their own AI. So they have their own biases in that respect.”
This approach led to NuEnergy’s research and the MTI system.
Furthemore, NuEnergy’s new patent, ‘Methods and Systems for the Measurement of Relative Trustworthiness for Technology Enhanced with AI Learning Algorithms’ helps to accomplish what Bhargava expressed as his company’s stated mission: “communicate to non technical people what they need to know as far as guardrails for AI.”
“We want to measure the trustworthiness of an AI algorithm…in a transparent and auditable way,” he added.
Hence, the creation of an “intricate analyses into a simple zero to 100 score,” which Bhargava says “supports comprehension across technical and non-technical stakeholders alike. “
MTI is also geared towards keeping enterprise leadership informed of the trustworthiness of their company’s AI tools, because, in Bhargava’s words: “governance belongs at the board level.”
MTI assesses a number of parameters of AI models and applications, including privacy, fairness, bias, security and more.
Bhargava noted that MTI can also be customized “to our client organization…it’s not one size fits all.”
To support this effort, MTI can accommodate case-specific parameters for unique industries such as healthcare, transportation, government agencies and beyond — ensuring relevance and actionable insights for each industry and company within it.
Measuring what matters most
NuEnergy established methods for indirect evaluation, helpful with the rise of embedded and opaque “black box” AI models.
“We spend as much time on methodologies for black boxes as white boxes,” Bhargava told VentureBeat.
The MTI provides a means of abstracted measurement to address issues like bias, privacy and transparency, even when direct examination of an AI system is not possible due to its inscrutability.
“We have methodologies for measuring inputs and outputs of models that you may not have access to the training data of. We essentially generate test data to get to the machine trust index,” said Bhargava.
Additional measures monitor model drift over time and conformance to standards such as Canada’s Algorithmic Impact Assessment integrated into their platform.
As autonomous systems continue infiltrating every aspect of modern life, NuEnergy’s work establishes an invaluable paradigm guiding ethical, inclusive development.
This article first appeared in VentureBeat