An ever-increasing variety of companies are incorporating machine learning into their products and services. Machine learning provides the ability to quickly and accurately perform, in parallel, a large number of well-defined tasks. The accuracy will improve over time as additional data is obtained and the machine learning model continues to “learn”. Many companies, however, are struggling with the best way to protect machine learning and artificial intelligence innovation.
In machine learning, statistical models (ie, neural networks) are trained using a set of classified data. Once trained, the model can analyse unclassified data, such as images representing unidentified objects, and classify or generate observations for that data. A significant issue slowing widespread adoption of machine learning is the inability to access or determine the internal relationships or mechanisms by which machine learning generates these observations. Information about the initial configuration and training might be known, but trained models cannot “explain” in easily understandable terms how specific decisions were made.
Read the full article in IP Magazine which delves into the legal complexities of:
- protecting the technology
- data privacy
- and bias – where the specific data used to train a machine learning model may result in unintended bias in the model’s decision making
For further information on the implications of Artificial Intelligence across industry sectors please visit our topic center