The Machine Learning Canvas
A framework to guide your organization’s AI transformation
A Deep Learning researcher said that “the world is held back from the potential capabilities of Artificial Intelligence, not because we don't have a new (billion-parameter) neural network model, but instead because we just haven't hooked something as simple as logistic regression into the right place.” Doing so is actually quite challenging, for business and engineering reasons.
First, we need to figure out how to integrate Machine/Deep Learning in people’s lives and existing workflows, in ways that are useful to them, that solve their pains or their business problems, and that allow for better decision-making. Then, we need to collaborate with engineers to define how to collect and prepare data for the use of Machine Learning, how to evaluate predictive models in a way that makes sense in the application domain, and how to test models on live production data; we also need to specify time and volume constraints for predictions to make, models to create, and to monitor live performance and impact.
I will present how the Machine Learning Canvas can help with this. It is a visual chart with elements that describe the key aspects of real-world ML systems, and it bridges the gaps between engineering, science and business. I will also explain how the MLC can guide the steps of your organization's AI transformation — in particular, executing pilot projects, building team, and developing strategy.
This keynote will be part of the Data+Decisions 2019 conference (registration required).