How Machine Learning is Useful in Mobile App Development
TechnologyBit News Today1 min read
Machine learning reached everyday phone apps by two routes: services that learn from user behaviour, such as playlists and dating profiles, and developer toolkits that run trained models directly on the handset. The table lists the launches that set this pattern between 2015 and 2020, together with the figures each company reported.
| Date | App or toolkit | Machine learning feature | Reported result or scale |
|---|---|---|---|
| Jul 2015 | Spotify Discover Weekly | Personal weekly playlist driven partly by collaborative filtering of listening patterns | 40 million users and nearly 5 billion streamed tracks by May 2016 |
| 2015-2016 | Spotify back end | Weekly regeneration of every user's playlist | 100 million playlists rebuilt each Sunday night from about a terabyte of fresh data |
| 15 Sep 2015 | Snapchat Lenses | Real-time face tracking taken from Ukrainian startup Looksery | Purchase price of $150 million later confirmed in Snap's IPO filing |
| Oct 2016 | Tinder Smart Photos | Rotates profile pictures and reorders them by how other users swipe | Up to 12% more matches in Tinder's own tests |
| Jun 2017 | Apple Core ML | Framework for running trained models on Apple devices, used by Siri, Camera and QuickType | Converts models from Keras, Caffe, scikit-learn, libsvm and XGBoost |
| 12 Sep 2017 | Apple A11 Bionic | Dual-core Neural Engine for Face ID and Animoji | Apple claimed 600 billion operations per second; closed to third-party apps at launch |
| 8 May 2018 | Google ML Kit for Firebase | Ready-made APIs for Android and iOS: text, faces, barcodes, image labels, landmarks | Released in beta; face and barcode detection on device, landmark recognition in the cloud |
| Jun 2020 | Standalone ML Kit | On-device APIs split from Firebase into their own SDK | Apps no longer need a Firebase project to use them |
Sources: TechCrunch, Engadget, IEEE Spectrum, MIT Technology Review, Kyiv Post, Android Authority, InfoQ, CNBC, Firebase Blog, VentureBeat, Android Developers Blog.
