Like MIT’s campus computing environment, Athena, a pre-cloud solution that lets files and apps follow their users, Dropbox’s Drew Houston ’05 takes his alma mater with him wherever he goes.
After earning a bachelor’s degree in electrical engineering and computer science, Houston’s frustration with the cumbersome need to carry around portable USB drives led him to team up with fellow MIT student Arash Ferdowsi to develop a web-based solution—which later became Dropbox.
Dropbox, which now has more than 500 million users, is still adapting. The file-sharing company recently exceeded $1 billion annual subscription revenue threshold. It expands its business model by selling at the enterprise level – employees at companies that operate Dropbox can essentially operate one huge box.
True to his company’s goal of using technology to connect people (and files), Houston is eager to share his wisdom with others, especially those at MIT. Houston delivered the 2013 graduation speech, saying, “The hardest-working people don’t work hard because they’re disciplined. They work hard because working on an exciting problem is fun.”
He has also been a guest speaker at “The Founder’s Journey” course to demystify entrepreneurship and at the MIT Enterprise Forum in Cambridge; regular and vigorous participant Begin froman entrepreneurship workshop held during the Independent Activities Period (IAP) and a regular part of the MIT Better World Tour, an alumni event held in cities around the world.
Houston visited MIT in February to attend the latest version of StartMIT to have a “fireside chat” about the early days of Dropbox, when it was run with some of his friends from Course 6, and discussed the company’s current challenges: managing scale. His company now employs more than 1,000 people.
“With thousands of employees in a company, you need coordination, and it can become total chaos. Ultimately, all vectors must point in the same direction,” he told the students. It turns out that Dropbox’s recent online collaboration suite, Paper, could play a key role in bringing these vectors into line, offering lessons for both those just getting started in startups and more seasoned entrepreneurs.
The Department of Electrical Engineering and Science (EECS) sat down with Houston to ask him about his perspective on Paper, a recent tool created by Dropbox out of necessity, and the potential of artificial intelligence to improve the way people and teams organize their work online .
Q. Why did you decide to create a recent form of cooperation?
AND. File sync is a way to keep teams in sync – sharing information keeps teams on the same page. But teams need more than just file sharing; they need a way to organize their knowledge, which may even mean taking bits of useful information from different files and combining them in recent ways.
In Dropbox, we used tools like Google Docs. And we found that they worked really well in certain situations: creating documents in real time was incredibly straightforward, collaborating and editing – it was a great solution. However, the problem is that after using them, we ended up with a inert list of 100 documents and 20 different projects that was not very organized. The documents were stand-alone and unconnected.
The other half of the time we used wikis—which are good because they’re like the web: they’re connected, public by default, so they become the home of your knowledge. But the editing experience is pretty narrow. There’s not a lot of formatting. You can’t store more than text.
Either way, we say, “God! We’re always compromising.” But what if we had the best of both worlds?
It’s one thing for tools to be simply functional. But what if they were pretty too?
Look at this building [] We spend a lot of time, money, and thought designing our physical spaces – what about a collaborative work environment?
Q What exactly is pretty about paper?
AND. [] In fact, this is how we run our company. [] It is very straightforward. Look at what is not there; there are no 10,000 buttons. [Beautiful by default.] I can add a few more things. []: “@Drew, remember to eat. []… See you tomorrow.” [] You can add a picture. You can share a document or send it. There are stickers. You have the formatting that you would expect, but also emojis, code, LaTeX, tasks, tables. They are pretty by default. You can embed things — Google Docs, GIFs, Dropbox files with previews, Spotify files, Soundcloud files, tweets, you can paste a URL. It’s really amazing. And it’s on your phone. There are updates, tasks to do — so it gives me a feed of what’s going on in my world.
Q Paper also includes machine learning tricks. What does this mean in practice? And where could it go?
AND. When you zoom out wider [than a single document]There’s just a lot of context that helps people figure out what’s significant or how to prioritize or organize things. For example, you might be working on a Paper doc and based on keywords or various other signals, we can say, “Oh, wait, here are some suggestions for where you could put this, like a public folder.” I think there are a lot of ways we can assist; instead of everyone having to be a librarian and sorting things out or tagging things and manually making those connections, algorithms can do a lot of the work of ranking things and so on, much better. And that’s throughout the product. It’s a fundamental part of the experience.
Another area that I’m really interested in is natural language processing. A lot of work, like when you’re emailing or texting someone or leaving a comment, is just unstructured text. But from that, you can see that there’s actually structure, intent, meaning, and association. For example, if I type “document scanner,” it’s not just a bunch of words; it’s a feature that triggers associations. You should also be able to test your team’s knowledge, and you don’t have to worry about the explicit question, “Where are the specs for the document scanner?”
Natural language processing can essentially ask, “Is this what you mean?” When you try to type a query, it can assist you with automatic suggestions: “Oh, are you looking for (this)?” Or “are you looking for something because of (this person)?” Or “do you have something in mind that is part of this project?” Or “anything before that date?” These are just a few ways [machine learning and automated assistance] can be used, but there are potentially thousands of them. We are not very far in understanding true natural language yet. But I think the first step is to just get everything together in one place – do some plumbing so you can do some basic research.