The Future of Visual Recognition Technology – An Interview with Clarifai’s Keeyon Ebrahimi
Clarifai is an industry leader in visual recognition technology. Last year, Extensis and Clarifai teamed up to bring the power of visual recognition to digital asset management, enabling companies to accelerate image keywording. As keywords are the underpinning of effective digital asset management, the technology integration represents an exciting advancement for DAM.
We sat down with Software Engineer, Keeyon Ebrahimi to discuss how Clarifai makes AI accessible to everyone, gets smarter over time, and what the future of image recognition technology looks like.
We also ended up talking about illegal contraband, hackathons, how to avoid bathroom lines at work, and preparing for the imminent robot apocalypse! Enjoy!
Joscelyn: Hi, Keeyon. Let’s kick off with the obvious question- can you tell us a little bit about yourself and your role at Clarifai?
Keeyon: Hello! I’m a Software Engineer on Clarifai’s data team. I focus on acquiring and cleaning data for use for our models. I also work on the infrastructure pertaining to this process, making acquiring and cleaning data easier and faster. Before Clarifai, I went to school for my undergraduate degree in Computer Science at the University of Utah and my Masters of Computer Science at NYU. I also worked in the industry for a couple of years at L3 Communications as both a Network Engineer and also GUI developer.
Joscelyn: How is Clarifai different from other visual recognition companies?
Keeyon: Clarifai is really focused on making AI accessible to everyone, including developers and non-technical users. We have the best technology on the market and the most experience in the market building real visual recognition solutions for real customers. You can read more about what makes us different from other visual recognition companies in our blog!
Joscelyn: Who uses Clarifai the most?
Keeyon: Companies in a wide variety of industries use the Clarifai API to power their businesses. Online marketplaces use Clarifai to moderate and filter user-generated content to prevent users from listing illegal or contraband items like guns, drugs, and live animals. Media companies like Buzzfeed use Clarifai to automatically tag images and video and make content searchable to their audience. Marketing technology companies like Curalate use Clarifai to build image recognition into their platforms to deliver better social media and sentiment analytics to their clients. And, last but not least, individual developers use Clarifai in many creative ways to power their own smart apps and projects!
Joscelyn: Can you discuss how Clarifai gets smarter over time?
Keeyon: Clarifai gets “smarter” as you show it more examples of what it’s trying to learn. However, more data is not necessarily better unless its good data—from day one, we focus on collecting the right kind of data. There comes a point of diminishing returns with data: Ten examples are obviously better than one example, but ten million examples are only incrementally and better than one million. What matters more is the type of data you have—you can’t just train any random image from a web crawl to improve accuracy; you must train images that are labeled well and suit a specific task. Clarifai has access to specific datasets from certain parts of the web, data partners, and user feedback. Clarifai’s API is unique in the market because of the ability for the end user to send feedback to our algorithm, making it smarter with every use.
Joscelyn: What’s one of the most interesting applications of the technology you have seen?
Keeyon: I’m blown away by the creative hacks that students come up with when we host hackathons. One of my favorites was called Classifi.me where they took Instagram photos and used them to figure out people’s Meyers-Briggs personality type.
Of our business users, I love what i-Nside has done with our tech. They incorporated our visual recognition technology into their mobile endoscopic diagnosis tool and have been using it to help diagnose ear diseases in regions of the world where people lack access to proper medical care!
Joscelyn: Any funny applications?
Keeyon: We hosted an AI hackathon recently where the theme was, “Build something to accelerate the robot apocalypse.” Students came up with some pretty devious and hilarious hacks. But the one that really impressed me was a hack our CEO Matt Zeiler built during an internal hack day. We only have two bathrooms in our office and there is often a wait for them. Because Matt’s desk is farthest away from the bathrooms, he got sick of walking all the way over there just to find them occupied, so he built a hack where he taught our tech to recognize if the bathroom door was open or shut, and let him know when to go.
Joscelyn: That’s hilarious! So… let’s end with a crystal ball moment. What does the future of image recognition technology look like?
Keeyon: AI is the cornerstone of the next generation of smart applications and is already on its way to infiltrating every part of our daily lives. I think image recognition technology will be ubiquitous in everything from self-driving cars to online shopping to healthcare to education. And it will change the way people interact with both content and computers. Exciting times ahead!
To learn more about Extensis’ application of Clarifai’s visual recognition technology in its digital asset manager, check out the new Smart Keywords module.
To accelerate your DAM knowledge, check out our DAM Best Practices Guide: