Follow Datanami:
September 21, 2017

Gaining Control Over AI, Machine Learning

As artificial intelligence and machine learning technologies make their way into advanced data management platforms, the emphasis for developers and data scientists is broadening to include not just deployment but “control” of the data accessed by these automation tools.

Along with the goal of completing often-tedious data science tasks in minutes and even, some claim, seconds, new enterprise tools also emphasizing greater visibility and control of the development processes underpinning AI and machine learning applications.

One such data management platform released this week by analytics startup Immuta of College Park, Md., seeks to ride the wave of AI and cognitive systems algorithms increasingly used by business to augment decision-making. The company argues that relying on these algorithms make companies more susceptible to risks ranging from critical errors to fraud.

To gain control of algorithm-driven business models, the company argues: “Organizations require greater control over the data being used by machine learning and AI models.”

“As organizations rely more heavily on algorithms to make critical business decisions, they will be required to demonstrate exactly how and why these decisions are being made, especially when it’s concerning important consumer choices, like credit, loans, healthcare and more,” explained Immuta CEO Matthew Carroll. “The key to explaining decisions made by machine learning and AI models starts by gaining a deeper understanding of the underlying data being used by the systems.”

Immuta’s data management platform is designed to provide greater control of the data fed into algorithms, speeding deployment as well as increasing visibility into how automation tools are functioning.

Among the tasks burdening data scientists are compliance with complex data security regulations and information governance policies such as rules for accessing personal data. Immuta said its platform includes audit and reporting features designed for dynamic policy enforcement, yielding what the company promotes as “an understanding of the risks associated with the use of the data….”

Among the controls on data used in algorithms is a “differential privacy” tool pioneers by hyper-scalers such as Apple NASDAQ: AAPL) and Google (NASDAQ: GOOGL). The concept works by allowing users to squeeze maximum value from large data sets while “mathematically guaranteeing” the privacy of personal information contained within.

Immuta claims its virtual data access and management layer is the first to apply differential privacy to a corporate database tool, giving data scientists access to anonymized data while at the same time complying with strict data governance rules. Those include the European Union’s General Data Protection Regulation, which enters into force in May 2018.

Immuta was founded in 2014 by a group of former U.S. intelligence agencies contractors. The company’s technology is based on its common “distributed data framework” designed to provides high-level security for processing sensitive data.

The company said its data management platform for data scientists is available now.

Recent items:

Ex-Spy Contractor Immuta Eyes Enterprise Data

The Complexities of Governing Machine Learning