IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Veritone Releases OS for Artificial Intelligence on Azure

Veritone’s aiWARE, marketed as “the first operating system for artificial intelligence,” lays a foundation for future AI programs aimed at public safety, large-scale video analysis and beyond.

When Veritone, an artificial intelligence company in California, launched its cloud-based operating system for AI in 2015, its sole customers were media and entertainment companies. Four years later, dozens of police departments are using aiWARE, and Veritone is releasing it on the Microsoft Azure Government cloud, optimistic that state and local governments have barely scratched the surface of its potential.

According to the company’s website, Veritone built aiWARE to help customers process a volume and complexity of data that was becoming unmanageable for human staff. For companies like ESPN, that meant indexing, labeling and analyzing an enormous catalog of audio/video content without asking interns and staff to watch it all frame by frame.

Veritone President Ryan Steelberg said the resulting metadata from those activities was valuable, but it became more practical when customers could weigh it against other metrics, like audience engagement, for new insights.

“If I’m trying to [figure out] why ratings for this show are so low, historically, you wouldn’t be able to look inside the content and say, ‘The reason is, you’re showing this one host way too much,’” he said. “If it’s a digital player, I know exactly when people are abandoning a video. If it’s a five-minute video and I leave at minute two, why? Your ability to look inside the content, even on a frame-by-frame basis, allows you to start to make those correlations.”

Veritone made aiWARE available in AWS GovGloud in March 2018, and  Federal Risk and Authorization Management Program (FedRAMP) authorized it a year later. Steelberg said more than 35 police departments now use two of its AI-powered applications, Veritone IDentify and Veritone Redact, to identify people or redact sensitive information in video, audio and documents. The release of aiWARE and its applications in the Microsoft Azure Gov cloud and Azure Marketplace are part of Veritone’s push to make AI more accessible to the public sector.

Steelberg said part of aiWARE’s versatility is that, unlike most other AI models, it decouples the applications from the operating system’s underlying AI models. The OS sells separately from the applications and can function without them. 

“We’ve broken apart the components of the platform of AI to make it interchangeable and interoperable, to make it fast and efficient ... One thing we’re learning is that this bleeding-edge curve of AI is happening so fast that we don’t want to have to disrupt or retrain the investigator who’s using our software, even though we may have already upgraded … and improved the AI model 10 times in the back end," Steelberg said. " We call it an operating system because you can build your applications and you’re not tied to any specific AI models. So in the back end … as a new or enhanced model comes to the market … you can seamlessly, through the operating system, provision and deploy that new model without having to disrupt or impact the application layer at all.”

Veritone is betting that there will be a lot of use cases for this. Steelberg said making police departments more efficient will be the company’s initial focus, because that’s where the budgets are. He also sees real-time surveillance on the horizon, for all the security and privacy concerns that go with it.

But with an AI platform that can accommodate new applications available to the public sector on both AWS and Azure government clouds, what else will be possible? Steelberg likened emerging AI to mankind first harnessing fire — a dangerous weapon in the wrong hands, but with too much potential for innovators to ignore.

“Once you do enough scale deployment of these solutions, the bigger opportunity emerges,” he said. “I think AI, machine learning and next-generation data analytics will have a major impact for good, if we embrace it. … For me, personally, there’s no way that with all these cameras and technology and audio-capturing equipment, somebody should be able to walk onto a school campus who shouldn’t be there, carrying a bag with them. We can solve these things.”

Before the floodgates can open on some of these innovations, Steelberg said, people need more streamlined, less bureaucratic ways to test them. AI tools need data to run, and sometimes policies and systems meant to protect sensitive data also restrict the ability to test those tools. Steelberg hopes governments and AI companies will be able to negotiate ways to make proofs-of-concept more accessible.

“If you look at China and some other countries, there’s a lot less red tape for those groups to leverage and benchmark and test the use of AI. Just to get to that phase of a trial takes months, if not years, here in the United States, and I think that’s very risky for us as a society,” he said. “We need to come up with an easier method for any government agency to test these solutions, not necessarily with dummy data, but with similar data. If I’m analyzing satellite photos, to prove efficacy, you don’t need to use the actual satellite photo. But how do I let the geospatial agency have access to the platform so they can use it, test it and see if it works for them, without having to waste time and millions of dollars to find out it’s obsolete before you even turn it on?”

At the federal level, several initiatives are underway. President Donald Trump issued an executive order in February launching the American AI Initiative, for the purpose of investing in AI research and development, making new data and models open to researchers, establishing training and educational programs and setting governance standards.

In May 2019, at a forum in Paris hosted by the Organisation for Economic Co-operation and Development, Michael Kratsios, the U.S. chief technology officer, emphasized the unknown potential of the next era of data processing.

“When the cellphone was first developed, we had no idea that years later it would be used to do so much more than simply make a phone call. The unforeseen potential benefits of AI — and the new technologies to be built upon it — cannot be [over]stated,” he said. “The United States wants to work with you to build a future where artificial intelligence helps patients receive more accurate and comprehensive disease diagnosis; where senior citizens remain mobile and independent thanks to self-driving cars; and where drones conduct successful search and rescue missions using AI sensors. We already see the incredible benefits to society from AI, and so many more possibilities are within our reach.”
Andrew Westrope is managing editor of the Center for Digital Education. Before that, he was a staff writer for Government Technology, and previously was a reporter and editor at community newspapers. He has a bachelor’s degree in physiology from Michigan State University and lives in Northern California.