Alibaba Cloud Artificial Intelligence Strategy is Taking Shape

In the past, I’ve written extensible about the astonishing trajectory of Alibaba Cloud to go from an unknown platform to a force to be reckon with in the cloud ecosystem. In those articles, I repeatedly mentioned the pivotal role that artificial intelligence(AI) should play in the future of Alibaba Cloud as it competes more frequently with PaaS incumbents such as AWS, Azure, Google Cloud or Bluemix. Recently, we’ve seen a few announcements from Alibaba Cloud that share some light into their AI strategy in order to bridge the gap with the market leaders.

A few days ago, Alibaba Cloud announced the general availability of its machine learning platform (PAI) as well as the release of AI APIs for scenarios in healthcare and manufacturing. At this point, I believe Alibaba Cloud’s AI strategy is based on three fundamental pillars:

1 — An Open cloud AI platform

2 — A suite of AI platform AO APIs

3 — A suite of vertical specific AI APIs

Open AI Cloud Platform

The PAI machine learning service is at the core of Alibaba Cloud’s AI strategy. With the release of PAI 2.0, Alibaba Cloud announced support for multiple deep learning frameworks and a catalog of over a hundred AI algorithms that can be used by applications built on the PAI platform.

PAI 2.0 contrasts with the strategy followed lead AI cloud technologies such as AWS ML, Azure ML or Google Cloud ML. By supporting the execution of AI programs written in open source deep learning frameworks, Alibaba Cloud can allow organizations to leverage the existing developer talent in deep learning communities such as TensorFlow, Torch, Caffe, Chainer, Theano, DeepLearning4J and many others while also providing a robust runtime and management experience for those applications. In contrast, stacks such as Azure ML or AWS ML only support their proprietary machine learning language with some limited extensions in languages such as R or Python. While Google Cloud ML is a great platform for TensorFlow models, it doesn’t support the rest of the popular deep learning stacks.

PAI 2.0 also allows organizations to pilot AI-deep learning applications on their own premises using their favorite deep learning stack while leveraging Alibaba Cloud for running and managing the solution at scale. That level of hybrid interaction is missing from most of the cloud AI stacks. PAI 2.0's flexibility seems to have directly paid off with a large catalog of AI algorithms that can help to bootstrap applications built on the platform.

Vertical AI Services

Another clever move by Alibaba Cloud was to start offering industry-specific AI services. Specifically, the new version includes ET Medical Brain and ET Industrial Brain which abstract core AI capabilities such as image analysis, energy optimization or predictive maintenance which are relevant in the healthcare and manufacturing industry respectively. The launch of those AI industry services should make Alibaba Cloud more competitive with IBM Watson and Google DeepMind which also provide vertical AI services.

Platform AI Services

AI APIs that abstract key cognitive capabilities in areas such as vision, speech, language or knowledge are also a key component of Alibaba Cloud’s strategy. These capabilities should make Alibaba Cloud more competitive with technologies such as Watson Developer Cloud or Microsoft Cognitive Services.