Innovation
Introducing Peloton’s Artificial Intelligence Principles
A close-up image of the Peloton Guide, a compact, connected fitness device.
We are so excited to have Peloton Guide available to help the community get stronger. Guide represents many firsts for Peloton: our first connected strength product, our first device that can fit on a table, and of course, our first AI-powered product!
As with all our products, we spent a lot of time researching and talking to Members and non-Members. We looked at the most cutting-edge human-AI interaction guidelines based on two decades of academic and industry research to make sure we understood what people expect from similar products. One of the main ethos is that we want to really find the intersection between AI strengths and user needs.
Underpinning that work is a set of AI principles that help us make sure that as we continue to learn and develop AI at Peloton, we stay true to our mission of using technology and design to connect the world through fitness, empowering people to be the best version of themselves anywhere, anytime. These AI Principles represent the foundation of Peloton’s responsible approach to product development and the use of AI as it becomes more prevalent in the products and services used in connected fitness. These AI Principles supplement Peloton’s core values.
A man working out and executing a reverse fly in his living room, using the Peloton Guide.
Peloton commits to the design, development, and use of AI in accordance with the following principles:
Diverse and Inclusive. We source and generate diverse datasets that represent a variety of human attributes such as skin tone, weight, height, body type, and fitness levels to train computer vision AI models used in our products. For our voice AI models, we use different languages, locales, and accents from various regions across the globe to create the best experience and responsiveness for all Members. In addition, we leverage tools, techniques, and processes to help identify and mitigate bias in our datasets and machine learning algorithms, including, but not limited to, human inspection of the datasets.
Validated Process Map. We use internal review boards to ensure that our AI operates reliably and consistently under normal circumstances and conditions by performing extensive field-testing trials, test performance, data collection, and error analysis.
Responsible Engineering. We implement training protocols for responsible engineering, including continuing education on the proper implementation of AI datasets, algorithms, and models. Furthermore, we are dedicated to evolving our processes and techniques as incremental iteration and optimization of AI models form over time.
Privacy and Security. We employ AI in a manner that complies with applicable laws, regulations, and policies including relevant data protection and privacy laws. We also utilize industry best practices for maximizing the privacy, security, and resiliency of AI design, development, and use. All datasets obtained and used to power AI within our products are safeguarded by our privacy and security policies, ensuring de-identification of personal data where required.
Expect us to go in-depth on each of these pillars throughout the year and to provide updates on our development. AI is an exciting technology and we’re looking forward to taking you on this journey with us. Together we go far!