Viewdle, which recently won the Le Web pitch competition in early December, was in the spotlight for this week's Vator Box. Viewdle is a facial-recognition technology startup that was founded in 2006. Who better to evaluate technology to help us find things across the Web in an automated fashion than Marissa Mayer, Google's VP of search experience.
Overall, Mayer, Ezra Roizen (Vator Box regular and digital media investment banker and me) found Viewdle to be very promising.
Here are some highlights raised by the three of us in this segment:
- What percent of the time will the technology offer up images of Sarah Palin, when someone is searching for Tina Fey? In other words, discerning between look-alikes (and even those who look nothing alike) appear to be a challenge for Viewdle. In fact, search for Angelina Jolie on Viewdle in Reuters Labs, and the top search result is a video of Christie's auctioning off Marilyn Monroe photos
- The problem is changing as Viewdle rolls out the solution, meaning 10 to 20 years ago, there were a set or celebrities to identify. With the advent of YouTube and Flickr, there are many more faces online, and therefore many more opportunities to mistake one person for another
- Based on Viewdle's application on Reuters, the library of videos appears extremely limited
- There must be some speech-to-text technology integrated into the Viewdle system because the top search result was spot on when searching for John McCain and his retort to President-elect Barack Obama, "If you wanted to run against President Bush, you should have run four years ago"
- Licensing to big media companies, that clearly need an automated way to tag and identify content, would be lucrative as is licensing to government agencies
- Facial-recognition technology seems to be far less advanced than speech-to-text, meaning this technology may need more time to actually be monetizable
- Being a stand-alone video destination site probably doesn't make sense as consumers probably prefer a more general search engine where an end-user could find text, images, and video on one topic
- Google is looking at some speech-to-text and facial-recognition technologies. If Viewdle could help automate the world's information by using facial-recognition... well, you never know.