Introducing Georgian Partners Epsilon v1.0
We’re excited to announce the release of Georgian Partners Epsilon v1.0, our machine learning (ML) product that helps bring privacy guarantees to our portfolio companies’ AI solutions. We believe that Epsilon will enable our companies to:
- Build trust with customers by minimizing data security and privacy risk
- Reduce the cost of complying with regulations, such as the General Data Protection Regulation (GDPR) in Europe
- Acquire broad data rights and develop a competitive advantage
In light of the recent Facebook / Cambridge Analytica affair, we seem to be entering a new era, where trust is evolving into a key factor in buying and adopting a software solution. Legislation like GDPR is only serving to further move us in this direction. Privacy guarantees are an important aspect of building customer trust.
Privacy guarantees also allow our companies to aggregate data and ML models across customers. Aggregation, in turn, improves predictive outcomes for everyone. Furthermore, our companies can accelerate onboarding and time to value for new customers, doing so in days instead of months, by leveraging existing customer data and models.
R&D at Georgian
Our mission is to accelerate our companies’ success through the adoption of key technology trends. We’re doing applied research and developing software to help our portfolio companies further differentiate themselves by giving them access to breakthrough research and technology.
With Epsilon, that breakthrough research is differential privacy. Differential privacy is a mathematical definition for the privacy loss that results to individual data records when private information is used to create an AI product. Specifically, it measures how effective a particular privacy technique — such as inserting random noise into a dataset — is at protecting the privacy of individual data records within that dataset.
Epsilon v1.0 is the first in a series of software products based on breakthrough research that we will be rolling out in the months and years ahead. In this initial release, we’re supporting two common ML techniques — Logistic Regression and Support Vector Machines. We will continue to expand Epsilon to include other differentially private ML techniques.
In addition, we expect to release software products in two other areas:
Accelerated AI learning — while Epsilon brings privacy guarantees and enables data and ML model aggregation, benefiting from existing data or ML models requires incorporating research from the area of transfer learning, i.e., transferring information from one ML model to another. We plan to release software in this area later this year. We are also exploring data generative approaches, to reduce the dependency on real data when training ML models.
AI transparency — in most industries, customers require some form of explanation as to how ML model predictions and recommendations are generated. Regulation is also a driver. We are investigating areas such as interpretability, fairness and bias, to help our companies increase the adoption of their AI solutions and meet regulatory requirements.
Stay Tuned for More
Releasing Epsilon v1.0 is an important step for us to help our portfolio companies accelerate their adoption of key technology trends. Particularly, as trust is becoming a differentiator. To learn more about our software offering, check out our product page. You can also sign up to our newsletter for updates as they become available by clicking here.