Amazon Web Services has a reputation for taking a lot from the open-source-software community without giving much back.
Now, the cloud giant is taking another baby step away from that image by making some of its own artificial-intelligence code available for anyone to use for free. It's only the second open-source project out of AWS, which is largely considered the No. 1 player in the cloud-computing market.
Last November, AWS announced a machine-learning feature called SageMaker Neo, which allows users to train and run artificial-intelligence programs on Amazon's cloud. Now, AWS is making much of the SageMaker Neo code available as open source under the name Neo-AI.
The new project will help developers program hardware platforms, like home-security systems or perhaps even self-driving cars, to use machine-learning models like TensorFlow, the mega-popular AI technology originated at Google. Since Neo-AI is available as open source, anyone can use, download, or modify the code for free.
An olive branch from Amazon
Beyond the technology itself, Neo-AI could be an important effort from AWS to mend bridges with the open-source world.
In recent months, AWS has been criticized for taking open-source code and reselling it to customers as a paid service. Doing so is completely legal — open-source software, by its nature, can be used for any purpose, even commercial use. But Amazon's reputation for not contributing back to open-source projects has worked against it, as it's perceived as happy to profit from the software but not to contribute to making it.
Some studies have suggested that Amazon contributes very little code to open-source projects compared to fellow tech giants such as Microsoft, Google, Red Hat, and IBM.
Neo-AI isn't the first wave in Amazon's open-source charm offensive. Firecracker, announced in November 2018, was taken by developers as a sign that AWS was finally ready to contribute significant projects to the open-source world. With Neo-AI, Amazon is releasing even more of its internally developed technology as open source.
Neo-AI helps make hardware smarter
Normally, developers may need to spend weeks or months manually adjusting the program so that it works on whatever hardware device they're using — different types of gadgets have different levels of computing power and even battery life, making for a lot of variables that need to be fine-tuned.
Not only that, but the software on the device might also be a mismatch with the software the developer is using. Neo-AI eliminates these compatibility issues by converting these programs into a common format, and it also makes these programs run more efficiently on the hardware.
Neo-AI supports hardware platforms from Intel, Nvidia, and Arm, and in the future will support Xilinx, Cadence, and Qualcomm.
