To celebrate being named one of the world’s 100 most influential data leaders, FICO’s Scott Zoldi talks about his greatest achievements, scaling AI capabilities and his team’s goals for 2020
What would you say were your greatest professional achievements of 2019?
In 2019, I was proud to have realized my dream of filing over 100 patents in the area of AI and machine learning. I found inspiration almost daily in the news, as entire industries are challenged to build and apply ethical AI technology. In 2019 I focused on creating unique algorithms and methods to support the delivery of ethical AI. These include novel machine learning algorithm architectures that are more transparent; model confidence measures as to when to trust machine learning; and finally blockchain model governance to ensure that the entire model development process is auditable and meets regulatory and governance standards for data assets, model construction, and bias testing.
How will you build on those achievements over the next 12 months?
In 2020, I will be working corporate partners who want to make ethical AI a reality in their development shops. There are numerous companies that want to use the blockchain model governance framework to ensure models are built to the same standards across data scientists and feed into their model governance functions. In this way, companies can launch innovative ethical models into production more rapidly, reliability, and confidently. In addition, I will be releasing a neural network trainer that will produce transparent neural networks, in which the relationships that drive the outcome are exposed. This transparency can be used to demonstrate palatable relationships, and those that can be tested for bias.
What are the key challenges your sector’s data and analytics leaders will face this year? And how can they be overcome?
No practitioner in data and analytics can ignore the significant responsibility they take on by working with data and building models that are unbiased. As AI and machine learning find their ways into more and more applications, we will see more focus on proper model build standards and governance around data, model development, and bias testing.
In my view, regulation begets accountability, which is required to move up the Maslovian hierarchy, as it were, from ethical AI to responsible AI. As we wait for regulation to help establish standards across practitioners, all analytics leaders should work to build a set of development standards to leverage in their data science and be accountable.
What do you think the data and analytics industry will look like at the end of 2020?
In 2020, I expect to see broad demands for responsible AI. Organizations will look to adopt AI development principles similar to those that are already the bedrock of software development to ensure a standardized, auditable model build process.
Analytic notebooks will become source of records of the model development process and derivatives of the process will directly feed model governance and regulators. This will help ease the current widespread discord between model development teams and model governance teams, who attempt to understand model builds post-mortem.
In 2020 we will see these two functions drop their longstanding beef and contribute to a single auditable process.