This year, Facebook CEO Mark Zuckerberg made a New Year’s resolution to build an artificial intelligence that helps out at work and home. We haven’t heard much about Zuckerberg’s progress since then, but his goal inspired a Seattle startup to build an AI helper of its own.
Algorithm marketplace Algorithmia built their AI front desk assistant in just 24 hours during an internal hackathon, heavily relying on the company’s own algorithm library. The AI uses facial recognition to quickly identify people, announce them on Slack and perform some other neat tricks.
The user-facing component of the AI is a $50 Amazon Fire tablet taped up to the wall outside their Seattle headquarters, which captures the visitor’s face with its front-facing camera and presents them with further directions. If a visitor hasn’t been to the office before, the system asks them to answer a few questions and make a few faces.
When the AI does recognize the user, it will start their favorite song via Spotify and send, via a Slack private message, all the commits to the company’s GitHub that were posted while the user was away.
Algorithmia CEO Diego Oppenheimer said that the team could have added plenty more integrations, but wanted to finish the project during the company’s internal hack day.
“With over 2,000 algorithms in our library the possibilities were limitless,” he said. “[The team] had to work harder on narrowing down the system than expanding it.”
Other options the team considered included connecting the system to WiFi to unlock the office door or turn on the lights. The team also opted for a web-based system, but spending a little more time to build into an Android app could have allowed them to run the entire thing almost entirely within the $50 Amazon Fire tablet.
While the system is built off Algorithmia’s library of algorithms, the team did a lot of custom (and messy) coding to get the AI off the ground, so it is not quite ready for a public release. But Algorithmia did provide a list of components used in their system, along with an in-depth look at how they built it, over to the blog.
While it may not live up to Zuckerberg’s standards, it is still impressive that 24 hours of coding and a handful of modern algorithms can build a front desk robot for a startup.