Company Represented: Deloitte Risk and Financial Advisory
Speakers: Alexey Surkov (Partner, Leader Trustworthy AI Services), Sanjana Jain
Intro to Deloitte’s AI Institute
On October 7th, students attending the Business+Tech Tech Literacy Download heard from Alexey Surkov, Partner and Leader of Trustworthy AI Services at Deloitte Consulting to gain a deeper understanding of how AI is shaping business and how Deloitte is trying to help solve artificial intelligence’s (AI) greatest compliance issues.
AI is growing rapidly with a global valuation of $4 trillion. It’s changing and innovating businesses and services across pharmaceuticals, finance, healthcare, and tech. “We are currently seeing a transformation like never before”, says Surkov. He adds, “As the face of AI is growing and prevalent, so are the risks and potential concerns and issues with its use, it can bring risk to companies and a potential source of bias and inequities.”
The event kicked off with an introduction to Deloitte’s AI Institute, a definition of AI and its capabilities, a breakdown of Deloitte’s Trustworthy AI and Responsible AI frameworks.
Surkov works closely with Deloitte’s AI Institute supporting research, innovation, and assisting clients with a range of AI topics related to AI risk management, controls, responsibility, and ethics. He touched on what Artificial Intelligence can do:
- Emulate human behavior with strong sense of comprehension and responsiveness
- Learn, at times autonomously, through iterations and using structured, unstructured data
- Utilize various learning methods: symbolist, connectionist, genetic, analogist, and Bayesian; and approaches: supervised, unsupervised, and reinforcement
- Identify patterns, categorization, anomaly detections, and regression (prediction)
- Value Generation: Intelligent Automation, Enhanced Interaction, Augmented Judgement, Fortified Trust, and Product, Service, and Market Innovation.
Trustworthy AI Framework – How Diversity Impacts AI
This framework was designed to help devise a strategy to help AI be “good” and “not take over the world”, says Surkov. This framework is an assessment to help provide confidence and reliability in what AI can do. It accounts for risks, ethics and governance and compliance to meet mission and business needs for AI. irst item on the framework, Fair and Impartial, is defined as whether AI applications include internal and external checks to help enable equitable applications across all participants. It works in alignment with the topic of diversity and inclusion, bias, and fairness. This step was designed to answer whether AI is putting certain groups at a disadvantage and ensuring equal opportunity for all groups. There are ways to build algorithms and train models that can consider risks from bias and controls and create fairness and outcomes as part of the output. There are ethical walls and philosophical nuances that the company must implement as part of the model it wants to optimize, even though it might lead to a lower R-squared.
The next element in the AI framework is Transparency and Explainability, which inhibits all participants to be able to understand how their data is being used and how AI systems make decisions, algorithms, attributes, and correlations open to inspection and transparency.
Responsibility and accountability are another important part of the ethical set of issues around AI. It has to do with temptation to point to an algorithm and say “it did that”. According to Surkov, there should always be a human overseeing the algorithm whose responsibility it is to monitors theelements of the algorithm that escalates issues and identifies bias.
Robust and Reliable:when AI systems have the ability to learn from humans and other systems and produce consistent and reliable outputs. This phase ensures that the algorithm doesn’t veer off course when it’s brought into the real world even when it has successfully passed the controlled environment. Algorithms need to be built to withstand a variety of external scenarios, especially under malicious attempts, to preserve its reliability and accountability.
One hugely important aspect is Privacy. Consumer privacy is respected and customer data is not used beyond its intended and stated use; consumers are able to opt in or out of sharing their data. An organization that is data hungry will go above and beyond temptation to train the data and avoid consumer consent to create the best possible solutions.
Safe and Secure is labeled as the traditional cyber risks. Cyber risks are present in any data system. AI tends to interact with customers in real time, AI traditionally has more risks than any other technical system. AI systems can be protected from risks (including cyber risks) that may cause physical and/or digital harm.