Say hello to the not-so-new AutoML!
An AI announced by Google brain in May 2017, and like the name (Auto-machine learning) suggests, it is capable of generating its own AI, or should I say ‘child models,’ that can be further trained and evaluated by the parent — AutoML.
The child AI, in this case, is called NasNet.
Recently, NasNet was given the task of recognizing objects in a video and in real time. After the task, it was discovered that NasNet outperformed all other computer vision systems that were human-made — it was 82.7% accurate and 1.2% better than any other results from human-made systems.
How can this be applied to everyday use?
For one, an AI like AutoML could make tasks easier, there is also the fact that NasNet’s computer vision algorithms could be used to address vision problems for both humans (In this case, the blind) and machines (it could help autonomous vehicles recognize objects), the possibilities are endless. However, there is the fear of an AI apocalypse, where AutoML could pass unwanted biases to NasNet and, in turn, create a system that could be difficult to control, but like the original article stated, in the near future, regulations would be put in place to control such from happening.