#Asia When AI eyes ads


AI eye

Photo credit: Alfred Payne.

“Nobody wants to click,” says Amos Huang, speaking of those ads that appear before you watch a video. And that’s a major concern for a guy who’s building technology for advertisers.

His answer is artificial intelligence and image recognition. With that combo, ads can be smarter – and that means ads that you and I might actually want to interact with.

Amos, who hails from Taiwan, did his BA in advertising. He then took quite a leap to do his MA in artificial intelligence. He studied AI before it was cool – seven years ago. And now, his startup, Viscovery, is basically blending his two majors.

The idea is that Amos’s “deep learning” tech scans the movie or TV series that you’re streaming and then brings up an ad that’s in some way relevant to what you’re watching. So if the characters in the movie walk past a Starbucks, the AI-powered ad machine will pause the stream and play an ad for Starbucks’ latest Frappuccino creation.

Amos Huang. Photo credit: Viscovery.

Amos Huang. Photo credit: Viscovery.

Or you’re watching football online and Paul Pogba appears on screen. The image recognition bots get to work and pull up the star’s newest Adidas ad.

Amos is betting that because it ties into what you’re watching, you’re more likely to click it than those ads for random things that appear before the stream begins. They’re called “pre-roll ads”. “They’re not efficient,” Amos says of them. Indeed, only 0.8 percent of viewers actually click them; it can barely get any lower than that.

His startup, after partnering with both streaming sites and ad agencies, will instead put the ads “in-stream” by pausing the feed for about 30 seconds, or “out-stream” by sticking a banner or text over part of the video.

Running on empty

Amos and the team last week wrapped up US$10 million in funding to get this system out of the lab and onto the web.

The lead investor was China Development Industrial Bank (CDIB) – which ties into his focus on the mainland China market, where streaming is massive. It’s also hugely competitive as most of the nation’s multiple tech giants throw money at their video sites as part of their growing web empires. Even online shopping giant Alibaba is doing so, as evidenced by its multi-billion-dollar acquisition last year of Youku, China’s top online video firm.

The startup has no revenue now – and has had none from its advertising tech for the past two or three years.

China now has just over half a billion regular online video viewers – from virtually nothing just a decade ago. That has fueled ad revenue from streaming, which was worth US$5.9 billion in 2015. It’s projected to nearly triple to US$16.3 billion by 2018.

“We’ve just started to collaborate with a giant,” says the founder and CEO, but he’s keeping tight-lipped on which Chinese site that is.

Amos admits that the startup has no revenue now – and has had none from its advertising tech for the past two or three years. So a lot hinges on its first few online deals to get the deep learning-powered ads into action. “We’re glad we got funding, but there are still a lot of things to do,” he concedes.

It’s described as a “series A-plus” round that adds to the US$5 million it nabbed last year.

His startup recently made the difficult decision to kill to a mobile app it had built for shoppers. It allowed people to snap a photo of a product in a shop and then get a link for that at an online store – where it would probably be cheaper. But consumers like you or I are fickle beasts who are overwhelmed with free app choices – which made the app a big struggle.

“Finally we see it’s the wrong direction, because there’s not enough money to educate users,” explains Amos.

All that image recognition tech is now ploughed into its video ad product, called FITAMOS – an acronym of face, image, text, audio, motion, object, and scene. Marketing to agencies is easier than getting you or I to install an app we’ve never heard of, he says.

“Before the past two or three years, nobody wanted to talk about AI [because] the tech was not ready” to do most of the things that people had in mind for it, says Amos, even though academics had been working in the field for about 30 years. The AI brain simply needed too much computing power to hash through all that info and make decisions.

Now he believes deep learning and AI are ready to roll in millions of ways. “They’re not just a theory,” he says – just look at what Google’s AlphaGo can do.

Converted from Chinese yuan. Rate: US$1 = RMB 6.77.

This post When AI eyes ads appeared first on Tech in Asia.

from Startups – Tech in Asia http://ift.tt/2eARcjn

This entry was posted in #Asia by Startup365. Bookmark the permalink.

About Startup365

Chaque jour nous vous présenterons une nouvelle Startup française ! Notre pays regorge de talents et d'entrepreneurs brillants ! Alors partons à la découverte des meilleures startup françaises ! Certaines d'entre elles sont dans une étape essentielle dans la vie d'une startup : la recherche de financement, notamment par le financement participatif (ou crowdfunding en anglais). Alors participez à cette grande aventure en leur faisant une petite donation ! Les startups françaises ont besoin de vous !