In December, Chinese technology company Tencent released a report that estimated the number of artificial intelligence researchers worldwide at 300,000. And last year, an independent research laboratory estimated that only about 10,000 people have the necessary skills to carry out serious artificial intelligence research.
Both conclusions highlight the relatively small pool of software engineers that understand artificial intelligence and the shortlist of corporations that can afford their skyrocketing salaries. But the complexity of the technology has lit fires underneath more companies to create standard tools for artificial intelligence techniques like machine learning and deep learning.
That is the reason behind Acumos, a new project launched by the Linux Foundation to create an out-of-the-box environment for building and sharing artificial intelligence software. The foundation, which oversees the open source Linux operating system, is trying to reduce the quirks and complexities of artificial intelligence software, ultimately making it more widespread.
By lowering the bar for software engineers to share artificial intelligence software, the foundation could help them create new and better versions of technology that can identify faces in photographs or automatically wireless networking. Acumos supports TensorFlow and other machine learning libraries as well as popular software languages including Java and Python.
Users can browse a marketplace stocked with machine learning models trained on vast amounts of information, such as images streamed from security cameras, by a community of software engineers using Acumos. These models can be trained, evaluated and chained together into complete solutions with little additional code development, the Linux Foundation said.
The plan is for software engineers to pluck applications from the marketplace, like location tracking and facial recognition, and combine them automatically so that they function as a single application. If it takes off, Acumos could be used, for example, to create a video analytics application to determine where the video was taken based on background landmarks and to identify the people in it.
That would be the ideal outcome for the Deep Learning Foundation, an organization that the Linux Foundation launched to oversee artificial intelligence projects like Acumos. Members include several of the largest internet and hardware companies in China, including Tencent, Baidu and Huawei. Others include Nokia and AT&T, which began working on Acumos in October.
“Our goal with open sourcing the Acumos platform is to make building and deploying A.I. applications as easy as creating a website,” said Mazin Gilbert, vice president of Advanced Technology at AT&T Labs, in an October statement. “The platform will be available to anyone and the more users who adopt it, the better it will get.”
Google has used the same playbook. Nearly three years ago, it open sourced TensorFlow, now one of the most popular machine learning software libraries. On Friday, the company announced TensorFlow Hub, a new repository of machine learning models that have already been trained but which engineers can share and reuse for different tasks. The library contains what Google calls modules, or tiny snippets of TensorFlow code.
