Google's new browser tool helps you learn Artificial Intelligence!
Machine learning - most of us have heard about this tech trend, which refers to giving computers the ability to learn without direct programming. But how does it work? To help us understand this process, tech-giant Google created "Teachable Machine" an in-browser experiment. It makes it easier for people to explore how machine learning works, letting them create simple programs using webcams. Read more!
Coming to Artificial Intelligence (AI), machine learning is the best way to create artificial neural networks that function in a manner similar to our brain. Teachable Machine allows users to train these neural networks or simply "teach machines" in a browser, using webcams. It was developed by Google's Creative Lab and PAIR teams in collaboration with design studios Stoj and Use All Five.
According to Google, "Teachable Machine" experiment is "built with a library called deeplearn.js, which makes it easier for any web developer to get into machine learning, by training and running neural nets right in the browser."
Google says Teachable Machine "lets you teach a machine using your camera - live in the browser, no coding required." On Teachable Machine's website, users need to click on the "train green/purple/orange" buttons after which the machine records whatever it sees through the webcam. After "learning" it will output whatever the user wants (GIF/sound/speech) once it sees the objects/activities it has trained with.
Google said users must capture at least 30 images per session by pressing and releasing the button, which starts/stops the ability to capture images. However, it has clarified the photos are stored on the user's device and not on Google's servers.
Teachable Machine gives a summary of what AI can and cannot do. The experiment also explains some basic aspects of machine learning. Firstly, machines (or programs) learn from examples. They look and find patterns after which they remember them. Secondly, machines need a huge number of examples for learning. Thirdly, these programs' understanding of the world is pretty superficial and can be quickly interrupted.
In this experiment, users basically teach or train the machine to recognize an array of pixels. The program doesn't see the objects the way users do. The machine learns only about these sets of pixels; other additional information has to be programmed.