Microsoft’s artificial intelligence application can now turn text into images

By Anirban Ghoshal

  • 19 Jan 2018
Credit: Reuters

Microsoft has developed an artificial intelligence application that turns text into images, leveraging Generative Adversarial Network.

"The network consists of two machine learning models—one that generates images from text descriptions and another, known as a discriminator, that uses text descriptions to judge the authenticity of the generated images," Microsoft said in a statement.

"The generator attempts to get fake pictures past the discriminator; the discriminator never wants to be fooled. Working together, the discriminator pushes the generator toward perfection."

The researchers fed the new system with lots of data that were essentially image and caption pairs. Post that, they created another GAN to help the AI programme turn long descriptions or captions into images. This resulted in the completion of the text-to-image system with a nearly three-fold increase in image quality compared to previous techniques, the company said.

Microsoft said that this piece of technology might be used to assist painters via a virtual assistant module comparable to Apple's Siri or its Cortana which are currently used for a variety of purposes including maintaining schedules.

In a separate development, the firm has released a new open source IoT extension that adds to the capabilities of Azure CLI 2.0.

"Azure CLI 2.0 includes commands for interacting with Azure Resource Manager and management endpoints. For example, you can use Azure CLI 2.0 to create an Azure VM or an IoT Hub. A CLI extension enables an Azure service to augment the Azure CLI, giving you access to additional service-specific capabilities. The IoT extension gives developers command line access to IoT Hub, IoT Edge and IoT Hub Device Provisioning Service capabilities," Chrissie Chi, program manager at Microsoft, wrote in a blog post.

The launch of this IoT extension means that the company will be retiring version CLI 1.0 and the IoT-hub-explorer tool.

The US-based tech major has also been very vocal about regulations in the AI and IoT space. According to a Bloomberg report, the Redmond-headquartered firm is of the opinion that companies which are making and selling products and solutions related to AI will need to abide by rules that ensure that such products don't cross a line in terms of "unreasonable practices". The company cited the example of a self-driving car which might not be tuned to maintain road safety rules.

The firm said that companies working on AI should also take a responsible approach towards employees who will lose jobs as a result of new computing methods.

“We are trying to be clear-eyed in talking about the challenges,” Microsoft president and chief legal officer Brad Smith told Bloomberg. However, the news agency quoted a professor of law saying that technology calling for rules to govern themselves is very unusual.

“There are a bunch of players in this space, and if you are Microsoft, you want to be seen as trusted,” said Ryan Calo, a professor at the University of Washington School of Law.

Both Microsoft and Calo said that new laws will take some time to come as AI trends are still emerging. Microsoft said that it will shape the company’s ethics and design rules to govern its AI work in partnership with Smith’s legal group and the AI group run by executive vice president Harry Shum.