Leading global technology companies appear to be in a race to secure their spot, particularly in the world’s future technology space. Thanks to their keen interest in research and development of AI and machine learning technologies, they can create groundbreaking products.
For instance, Apple launched Create ML and Core ML 2 during the Worldwide Developers Conference (WWDC) 2018 that was recently held in San Jose, California. The new tool was designed for facilitating basic, efficient machine learning on the MacBook.
Create ML is the new GPU-driven tool intended for native artificial intelligence (AI) training on Apple’s Mac Books. According to Craig Federighi, the senior VP of software engineering at Apple, this tool supports the training of natural language processing and vision through your own custom data.
Thanks to its built-in Swift, Create ML allows you to utilize drag and drop programming interfaces such as Xcode Playgrounds in training models.
On stage, Federighi expounded by saying previously it would take a single developer, Memrise, about 24 hours simply to train a model using 20,000 images.
However, Apple’s new tool, Crete ML minimized the training time for a similar model to 48 minutes and 18 minutes, especially on a MacBook Pro and an iMac Pro respectively. Furthermore, the tool managed to reduce the model’s size from 90 MB to 3 MB.
Aside from the announcement of Create ML, Apple also unveiled Core ML 2, which is said to offer 30% faster on-device processing compared to its predecessor, Core ML. The company attributed this ability to a method called batch prediction.
What’s more, Apple told the audience at the Worldwide Developers Conference (WWDC) that the toolkit would allow developers to minimize the size of their trained machine learning models by utmost 75 per cent via quantization.
Apple unveiled Core ML back in June 2017 with the introduction of iOS 11. This tool enables developers to load on-device machine learning (ML) models, particularly onto an iPad or iPhone.
It also allows the conversion of models from frameworks such as Scikit-learn, XGBoost, LibSVM, Keras as well as Facebook’s Caffe and Caffe2. In addition, Core ML is created to augment models for power efficiency and can leverage the benefits accompanied by machine learning models without the need for an Internet connection.
The announcement of Core ML’s upgrade closely follows the ML Kit, which is a machine learning development kit that is designed for both iOS and Android that Google launched recently in May during the I/O 2018 developer conference.
Previously in December 2017, Google had introduced a tool that can convert artificial intelligence (AI) models made through TensorFlow Lite, its well-known machine learning (ML) framework, into a file type that works harmoniously with Apple’s Core ML.
Core ML is anticipated to serve an essential role in Apple’s upcoming hardware products. Currently, the technology guru is creating a chip dubbed the Apple Neural Engine or ANE in a bid to expedite speech recognition, computer vision, facial recognition among other artificial intelligence forms.
Apple also intends to include the tool in its future devices. According to Bloomberg, third-party developers will also gain access to the chip in a bid to operate their AI.