In July 2023 Google ran a benchmarking study asking organizations about the impact they expected generative AI to have.
“we did a study back in July 2023, a benchmarking study, and asked organizations about the impact they felt Gen AI would have.”
Person
Behnaz Kibria heads the Product Initiatives team at Google, within its Government Affairs and Public Policy organization, and appeared on the Barefoot Innovation Podcast to discuss how banks should manage risk from generative AI. She defined generative AI as a type of artificial intelligence that can create new content or data rather than only analyzing existing information.
Kibria said the U.S. model risk management framework was updated in 2021 to give more specific guidance on AI. She also said cloud providers like Google were once outside conversations between regulators and banks, and are now part of a three-way conversation. She cited Deutsche Bank as a Google Cloud customer already using generative AI to streamline operations and automate tasks.
1 episodes · 16 quoted claims
In July 2023 Google ran a benchmarking study asking organizations about the impact they expected generative AI to have.
“we did a study back in July 2023, a benchmarking study, and asked organizations about the impact they felt Gen AI would have.”
In the study, 82 percent of responding organizations considering or using generative AI believed it would significantly change or transform their industry.
“82 % of the organizations responding that were considering or using Gen AI believed that it would either significantly change or transform their industry.”
Kibria defines generative AI as a type of artificial intelligence that can create new content or data rather than only analyzing existing information.
“generative AI is a type of artificial intelligence that can create new content or data.”
Kibria says the U.S. model risk management framework was updated in 2021 to give more specific guidance on AI.
“while that was updated in 2021 to issue more specific guidance on AI”
Kibria says most financial services laws and regulations are principles-based and technology neutral, which lets them adapt over time as technology changes.
“most are principles based and technology neutral. And what that means is that they are capable of being adapted over time as technology changes.”
The first Google and AIR paper examined what they called risk use cases, meaning traditional AI used for fraud detection and anti-money laundering.
“what we called at that time risk use cases, you know, the use of traditional AI technologies for fraud detection, anti-money laundering, etc.”
Kibria says that in practice, especially in the U.S. system, a model is defined very broadly and can be anything with inputs and outputs.
“it's really anything that's got inputs and outputs could potentially be a model.”
Model validation, which assesses a model's accuracy, reliability and limitations, is one of the categories that model risk management frameworks examine.
“things like model validation, assessment of a model's accuracy, reliability and limitations.”
Kibria cites Deutsche Bank as a Google Cloud customer already using generative AI to streamline operations and automate tasks.
“And we've got a number of customers at Google Cloud who are already using generative AI for those purposes, like Deutsche Bank.”
Behnaz Kibria heads the Product Initiatives team at Google, which sits within Google's Government Affairs and Public Policy organization.
“I am the head of a team called Product Initiatives at Google, which is part of the Government Affairs and Public Policy team.”
Kibria says traditional models could be unpacked step by step to see how they worked, which served as evidence of safety and soundness.
“in the old days when you had traditional models, you could just unpack how the model worked.”
A core proposal of the Google and AIR paper is that assessing generative AI should rely more on output testing and techniques like grounding.
“one of the core proposals of the paper, for example, is that you really have to look more at things like testing, output testing.”
The paper groups best practices for financial institutions into four areas of risk management.
“we sort of bucketed into four issues, four areas in which we think there's a variety of best practices that financial institutions should be considering”
Kibria says human-in-the-loop systems are important for critical decisions involving generative AI, to ensure sufficient human oversight.
“the really important aspect of human-in-the-loop systems for critical decision-making to make sure that you do have the necessary extent and level of human oversight”
Kibria says banks and their technology providers need clearly divided roles and a sense of shared responsibility in managing model risk.
“there's got to be a clear delineation of roles and responsibilities and a sense of sort of shared fate and shared responsibility as well.”
Kibria says cloud providers like Google were once outside financial services conversations between regulators and banks, and are now part of a three-way conversation.
“we used to be often outside of conversations, you know, in financial services, those conversations often were between the regulators and the regulated entities”