The 11th Hour Dispatch – Wednesday, May 2, 2018

| May 2, 2018

Subscribe to The 11th Hour Dispatch, Karma’s new anti-newsletter newsletter.

BINGE WATCH

Condé Nast has announced three new OTT products for its publications Wired, GQ, and Bon Appétit. Wired’s offering will hit Amazon Fire, Roku, and Apple TV later this year while the other two will launch in 2019. Pamela Drucker Mann, CRO and CMO, said, “While consumers have more screens and more content to choose from than ever before, advertisers face the challenge of less buyable options,” which seems to be the overall theme of this year’s round of upfronts across the board. Condé Nast’s overall streaming audience rang up 12 billion video views in 2017. The media company currently has 35 returning series and 60 pilots in development across its numerous brands. 

YOUNG MONEY

Google has opened an investment program aimed at funding early-stage startups focused on travel, hospitality, or gaming that are making use of the Google Assistant. The initiative will provide financing, early access to new features and tools from Google, access to the Google Cloud Platform, and promotional help for the young companies. On top of some pretty stacked resources, participating startups will be guided through product development by Google engineers, product managers, and design experts. Four companies have already been announced as part of the program: GoMoment, the maker of concierge app Ivy; Edwin, an AI-powered English language tutor; BotSociety, a development tool for voice interfaces; and Pulse Labs, a testing app for voice services that uses real people. 

BIG BUSINESS

Using over 3.5 billion publicly shared Instagram photos and their accompanying hashtags, Facebook has built an image recognition system with 85.4% accuracy. This system, which was announced today at Facebook’s F8 conference, is the “world’s best image recognition system” according to director of applied machine learning Srinivas Narayanan. The machine can detect specific subsets of images, such as “English bulldog” rather than “dog,” which Facebook plans to use to make its properties more customized to users. Prior to this system, all of Facebook’s advancements in the space were made completely through supervised learning, which requires someone from the big FB to manually label data fed to neural nets. The system announced today utilized weakly supervised learning, which uses both labeled and unlabeled data.

NOTHIN’ BUT ‘NET

Southwest had to divert yet another flight after a window cracked mid-flight. Nope nope nope.

Want more news like this? Subscribe to The 11th Hour Dispatch to get the newsletter you want, need, and deserve in your inbox every night.