Open Source AI is On Fire, and Facebook Has the Latest Contributions
Not long ago, in an article for TechCrunch, Spark Capital's John Melas-Kyriazi weighed in on how startups can leverage artificial intelligence to advance their businesses or even give birth to brand new ones. In a subsequent post, I noted that quite a few powerful artificial intelligence tools have been tested and hardened at Google, Facebook, Microsoft and other companies, and subsequently open sourced. Some of these tools may represent business opportunities.
Since then, there has been much more action on the AI front. In the latest move, Facebook is open sourcing three tools that the company uses internally for machine vision. They're called DeepMask, SharpMask and MultiPathNet: DeepMask figures out if there's an object in the image, SharpMask delineates those objects and MultiPathNet attempts to identify what they are.
According to a Facebook post:
"When humans look at an image, they can identify objects down to the last pixel. At Facebook AI Research (FAIR) we're pushing machine vision to the next stage — our goal is to similarly understand images and objects at the pixel level...The main new algorithms driving our advances are the DeepMask segmentation framework coupled with our new SharpMask segment refinement module. Together, they have enabled FAIR's machine vision systems to detect and precisely delineate every object in an image. The final stage of our recognition pipeline uses a specialized convolutional net, which we call MultiPathNet, to label each object mask with the object type it contains (e.g. person, dog, sheep)."
"We're making the code for DeepMask+SharpMask as well as MultiPathNet — along with our research papers and demos related to them — open and accessible to all, with the hope that they'll help rapidly advance the field of machine vision."
These contributions join a flood of recent moves on the open source AI front. Here are more examples:
Nervana. Just recently, Nervana Systems, a startup focused on artificial intelligence and deep learning, announced that it had released its Neon deep learning software under an Apache open source license, allowing anyone to try it out for free. Soon after that, Intel announced that it is acquiring the company. Neon is written in Python, and includes a Machine Learning Operations (MOP) Layer, allowing other deep learning systems, like Theano and Caffe, to integrate with it.
H2O.ai. In recent interviews here on OStatic, found here and here, we have explored the efforts of H2O.ai, formerly known as Oxdata, which has steadily been carving out a niche with its open source software for big data analysis and machine learning. You can get the main H2O platform and Sparkling Water, a package that works with Apache Spark, by simply downloading them. You can run them on clusters powered by Amazon Web Services (AWS) and others for just a few hundred dollars. Find out more about the opportunity this company's tools can provide here.
From Redmond. Microsoft CEO Satya Nadella has been very enthusiastic about AI. Microsoft has open sourced the artificial intelligence framework it uses to power speech recognition in its Cortana digital assistant and Skype Translate applications. The framework is called, CNTK, and can help machines do things like understand speech and determine logical connections between photos. Microsoft released its Computational Network Toolkit (CNTK) as an open source project on GitHub, and developers are likely to leverage it to advance deep learning networks.
Facebook On Board. In early 2015, Facebook open sourced modules for the Torch deep learning toolkit. According to Facebook leaders: "Torchnet provides a collection of boilerplate code, key abstractions, and reference implementations that can be snapped together or taken apart and then later reused, substantially speeding development. It encourages a modular programming approach, reducing the chance of bugs while making it easy to use asynchronous, parallel data loading and efficient multi-GPU computations."
Meanwhile, Facebook has open sourced its machine learning system designed for artificial intelligence (AI) computing at a large scale. It's based on Nvidia hardware. Facebook's Kevin Lee and Serkan Piantino wrote in a blog post that the open sourced AI hardware more efficient than off-the-shelf options because the servers can be operated within data centers based on Open Compute Project standards.
Google's TensorFlow. In numerous recent posts, we covered Google's decision to open source a program called TensorFlow and the related platform TensorFlow Serving. These are based on the same internal toolset that Google has spent years developing to support its AI software and other predictive and analytics programs. TensorFlow is rapidly gaining momentum.
It is being leveraged by researchers who need to analyze very large sets of complex data, according to Google. According to a Google post:
"TensorFlow Serving....is a high performance, open source serving system for machine learning models, designed for production environments and optimized for TensorFlow. TensorFlow Serving is ideal for running multiple models, at large scale, that change over time based on real-world data."
Caffe. Yahoo has released its key artificial intelligence software (AI) under an open source license. The company previously developed a library called CaffeOnSpark to perform “deep learning” on the big troves of data found in its Hadoop file system. Now CaffeOnSpark has become available for community use under an open source Apache license on GitHub. CaffeOnSpark works with x86 chips or graphics processing units (GPUs). It can be run on cloud infrastructure or within data centers. Among many uses for it at Yahoo, it has helped make connections for content recommendations.
Are you hungry for even more open AI tools that can be leveraged for new ideas? InformationWeek has a good roundup of some of the other deep learning and AI tools open sourced recently. It's good to see some of the biggest tech companies contributing their deep learning and AI tools to the open source community. No doubt, these contributions will help AI advance over the next several years.