Google has just upgraded its Teachable Machine artificial intelligence training tool. As a function that works in a browser, it stems from a web experiment launched by the company in 2017 to make it easier for people to understand some of the fundamentals of machine learning. The original version of the app is just an interesting little experiment that can classify visual data based on a regular webcam.
(Image via TheVerge)
Now, however, Google has added new models to the system and can export interesting models for training, making Teachable Machine 2.0 a more powerful system for building real-world AI tools.
In addition to image data, Teachable Machine can also be used for audio input and body attitude recognition. Users can upload their own set of pre-collected data and divide it into more than three categories. Then download and deploy the model locally, or host it in the cloud.
This means that users can train basic AI systems with Teachable Machine and run them on websites or apps. Google currently offers an AI training tool called Cloud AutoML that does not require coding.
Teachable Machine 2.0 – Making AI Easier for everyone (via)
Even so, it’s still quite professional, supporting a wide range of customizations, extensions, and customer support. But the new Teachable Machine 2.0 is even more powerful than it is. As an entry-level assistant for newosofer machine learning, it enables users to quickly prototype AI solutions.
Google points out that The Teachable Machine runs entirely on the user’s computer, meaning that the training data does not escape from local computing devices. Clearly, the company wants to dispel privacy concerns.