Google is helping people to start exploring how machine learning works.
Google’s got a new little way for users to teach themselves about artificial intelligence from the comfort of their own home/office/anywhere you can load a browser from.
The in-browser experiment, called Teachable Machine is a two-minute summary of what artificial intelligence can and can’t do – the can’t do bit is important.
Teachable Machine works through the users webcam so that they can teach an AI program.
The user can train three sections – green, purple, or orange by showing the AI through the webcam different things. The technology learns and then outputs a GIF, a sound effect, or some speech once it shown the object it was trained with.
Whilst clearly fun and entertaining, Teachable Machine is also teaching some valuable lessons to those that want to see some of the fundamental aspects of machine learning in process.
The program is designed to learn from example, it follows what is being shown to it by others and “learns” to recognise patterns and then remembers. Programs like this require a lot of examples to learn from and these systems can be easily confused.
What the Teachable Machine is able to show users is that AI does not see things like a person does. It does not see a cup of tea as a cup of tea, instead it sees pixels and only pixels.
There is no extra knowledge or understanding aligned with cup other than it is a cup. The liquid inside doesn’t matter to it.
Artificial Intelligence has advanced rapidly over the past few years but the reality about the technology is that it’s still the stuff of science fiction.
Pretty much everything AI is doing now is programmed into it. Algorithms are getting smarter and more complex, people are learning how to make AI systems smarter, but don’t expect it to strike up a conversation with you any time soon.
The Google Teachable Machine can be found here.