“Brands can get more relevant placements, and publishers and platforms can get more monetization and allocate their inventory more efficiently” with Hive’s contextual AI, Calpin said.
For example, an airline might want to air its 15-second or 30-second spot after a scene in a TV show that features a vacation, Calpin said.
Hive’s AI classifies content using the IAB’s taxonomy. Media and tech companies can connect to Hive’s API to use its contextual segments.
While many legacy contextual companies excel at analyzing written text, Calpin said, fewer have invested in images and video – where Hive focuses. Some of those other contextual companies are working with Hive to add image recognition to their toolset.
“Especially with the urgency in the market to solve for a cookieless world, building this new type of visual solution within each company is not the fastest or more efficient way to go to market,” Calpin said.
Hive, which has 175 employees, plans to use its new funds to add 100 more, mostly in sales and marketing, though it will hire tech staffers as well.
Hive was initially founded as a consumer app, but pivoted as content moderation took a disproportionate amount of the founders’ energy. After helping others moderate content, it branched out into other applications – like identifying logos and context for brands and media companies.
AI is only as good as the data it uses to build models, and Hive holds an “unfair advantage because it creates models with a tremendous amount of training,” Calpin said. The company created an app to distribute micropayments to 2 million people who have labeled more than 1 billion pieces of content. That data serves as the basis for its models.
The use of AI often raises issues of ethics and bias. Hive said it sidesteps those issues by working on use cases in media and attribution, and by avoiding AI contracts with governments or using facial recognition.
However, as 2020’s dramatic news year showed, brands’ use of contextual-based blocklists harmed news publications’ ability to monetize their content, even when it covered topics that advanced public health, democracy and social justice. Creating accurate, highly granular metadata is one way to avoid creating harm through blocklists, Calpin said: “For us, the responsibility is to be the most accurate and the most surgical metadata.”