Learn the core principles, concepts, and workflow of data-driven AI.
Keywords: Classifier, supervised machine learning
Teachable Machine is a browser-based tool that allows learners to train a machine to recognize images and to build their own AI-driven mobile apps without programming. Learners train a classifier using their own data (for example, photos they take themselves or images selected from the web) and define what the application does in different recognition cases. Finished applications can be shared with other learners.
By using the tool, learners discover that machine learning is based on data, examples, and probabilities instead of rules or human-like understanding. They can concretely experience how the amount and quality of training data affect recognition performance, why models can make mistakes, and how incorrect behavior can be improved by adding or modifying data.
Teachable Machine makes the full model training and deployment workflow visible:
From an initial idea and data collection, through model training and testing, to publishing a working application for others to use. AI is thus presented as a system designed and shaped by people, rather than as an opaque black box.
The tool runs entirely in the browser, locally on the user's device, and does not collect or store personal data. Applications can be shared directly between devices.
The tool runs entirely in the browser,
locally on the user's device, and does not collect or store personal data. Applications can be shared directly between devices.
Key concepts: Data, training data, class, label, classifier, data curation, training, input, output, confidence, bias
Released in Spring 2023