Artificial Intelligence

How GDPR will change the way we build machine learning algorithms

A new report from O'Reilly reveals that in order to keep pace with developing privacy needs, machine learning needs to evolve.

On Tuesday, a new report from O'Reilly outlined the current state of machine learning adoption in the enterprise and how recent political moves have affected it.

With the EU's recent General Data Protection Regulation (GDPR) mandates, more companies will begin to implement privacy safeguards into their machine learning practices. According to the report, some 53% of companies with extensive machine learning experience use the technology for privacy checks— 43% across all respondent companies. The new policies, however, happen to coincide with more tools and methods evolving that deal with privacy-preserving analytics and machine learning, the report noted.

SEE: EU General Data Protection Regulation (GDPR) policy (Tech Pro Research)

The report noted that GDPR pushes for "privacy by design," meaning that data protection should be included when a system is being developed, rather than be added later. Similarly, more businesses are taking interest in privacy-preserving analytic methods. These methods, the report noted, include techniques like differential privacy, homomorphic encryption, federated learning, and more.

According to the report, machine learning will need to keep pace as more companies start to develop a stronger stance on privacy. Companies, especially those with extensive dealings in the UK and EU, will need to consider all aspects of the GDPR regulations when building new machine learning algorithms or auditing those currently in use.

The report also noted that as machine learning becomes more commonplace, data professionals are becoming more interested in transparency, interpretability, and explainability.

Over half (54%) of respondents from companies with extensive machine learning experience noted that their businesses made use of machine learning for checking for biases and fairness, according to the report.

Some 51% of respondents reported that they use internal data science teams for building their machine learning models. The report noted that the use of automated machine learning solutions (like Google's AutoML) is in the single digits, and the split becomes even more apparent when it deals with "sophisticated teams." Companies with less experience in machine learning rely on external consultants, the report noted.

The big takeaways for tech leaders:

  • According to a report from O'Reilly, machine learning needs to keep pace with business's new focus on privacy, including data protections from the beginning.
  • Some 51% of businesses use internal data science teams to build machine learning algorithms.

Also see

machine-learning.jpg
Image: iStock/Everythingpossible

About Laurel Deppen

Laurel Deppen is a student at Western Kentucky University.

Editor's Picks

Free Newsletters, In your Inbox