the UID group: Soylent, crowds computing (from the cloud)

  1. Michael Bernstein and David Crowell at the Miller group conference and offices in the Gates Tower of the MIT Stata Center.
    Michael Bernstein and David Crowell at the Miller group conference and offices in the Gates Tower of the MIT Stata Center.
  2. Soylent homepage: http://projects.csail.mit.edu/soylent/
    Soylent homepage: http://projects.csail.mit.edu/soylent/
  3. Michael Bernstein and David Crowell at the Miller group conference and office area in the Gates Tower of the MIT Stata Center.
    Michael Bernstein and David Crowell at the Miller group conference and office area in the Gates Tower of the MIT Stata Center.
  4. Michael Bernstein and David Crowell discuss their project Soylent.
    Michael Bernstein and David Crowell discuss their project Soylent.

Meet Michael Bernstein, EECS PhD candidate in the Miller UID group working with David Crowell, EECS (6-3) senior (class of 2011).

Michael, what can you tell us about Soylent – what is it? and what is the goal in developing it?

Michael Bernstein:
“Today’s user interfaces are limited: they only support tasks when we know how to write matching algorithms or interface designs. Microsoft Word is good at laying out your document, but poor at understanding writing and suggesting edits to it. But, we now have the tools to embed on-demand human computation within interactive systems. Crowd workers on services like Amazon Mechanical Turk will do tasks for very small amounts of money. Soylent is a word processor with a crowd inside: an add-in to Microsoft Word that uses crowd contributions to perform interactive document shortening, proofreading, and human-language macros. Underlying Soylent is a new programming design pattern called Find-Fix-Verify that splits tasks into a series of generation and review stages to control costs and increase quality.

Our research goal is to explore what the world would be like if we embedded human cognition and intelligence directly inside of our user interfaces. How would that change your interactions with a computer? What could you do that you couldn’t do today? What kinds of algorithms and programming design patterns do we need to create to coordinate the crowd so that it produces intelligent results?”

How did this project get started?

Michael Bernstein:
“I had been working for a year or two in the broader space of social computing, which is interested more broadly in what happens when large groups of people collaborate or communicate online. As someone who spends most of my time building systems, I was spending a lot of effort designing incentives to get people online to act in particular ways. Mechanical Turk is interesting because it’s a completely different kind of beast — one where people will do tasks for money, rather than for interest — so it opened up this design space of tasks that someone might not volunteer to do for free. So, in a sense, I turned around and asked how Mechanical Turk would allow us to re-examine some of the big questions in human-computer interaction and social computing? Soylent was the result of that brainstorm.”

Are there a lot of users of Soylent already? How do you ‘market Soylent?

Michael Bernstein:
“There are about 500 people who have signed up to get the beta when it’s released soon. We haven’t even been “marketing” it very much; there’s just a web site and a YouTube video. Some influential folks blogged about the video and the paper, and I moved quickly enough to capture their interest into the beta signup. We’ll see what happens when we actually release!

Do you have competitors or are you covering a unique need (in a unique way)?

Michael Bernstein:
“There’s not a whole lot in this space. There are services like Standard Minds that try to do proofreading using crowds. Soylent is interesting because it pushes those results directly inside of a user interface that you’re already familiar with, and extends that interface in some new ways. The algorithms that coordinate the crowd work are also pretty unique to our approach.

Happily, lots of people agree that crowd-powered interfaces show potential. We’re already seeing lots of research that builds on and extends Soylent in exciting ways.”

What are the ‘perks’ you gain as members of Prof. Miller’s User Interface Design Group aside from getting support from this association? ie., do you all meet to describe, trouble shoot?

Michael Bernstein:
“Sure! Probably the most interesting choice is that we meet every day as a group. This really transforms our group meetings from out-of-date status reports to in-progress feedback and brainstorms. Twice a week we have open-ended brainstorms. The other three days are more business: they focus on work in progress. It really helps to be part of a community of scholars who think about these issues. Especially recently, the group has really been developing lots of projects that are different inflections of the same broad research goal.”

David, you have just completed your undergraduate degree (6-3, 2011). Can you give a look back on your experience with the Miller Group, ‘tea time’ and working on Soylent? Oh, and, congratulations!

David Crowell:
“Rob and the whole group are extremely supportive. Over the summer when I was working as a UROP, there were daily group-wide “Tea” meetings. Anyone was welcome to both attend and submit topics for Tea, which could be anything from papers or articles people found interesting to questions about project direction, progress reports and requests for feedback, and specific help with trouble shooting.

Tea was a great way to get help, follow other group members’ progress, and stay up to date with interesting new developments in the field as a whole. Rob was always a fantastic resource, and these meetings were a great aspect of working with the group. As a 6.UAP student, I wasn’t able to make it to Tea very consistently, but Rob had all of the 6.UAP students attend weekly meetings in small groups (organized by problem space) to provide weekly progress reports and seek out feedback and even specific debugging help. With a 6.UAP project, there’s a pretty high risk of letting the project lapse for a few weeks when other classes ramp up, but these meetings definitely helped me keep that in check.”

How did you happen to choose the Miller Group?

David Crowell:
“I met Rob and Michael while taking 6.813 (Michael was my TA). Around the time the Hackers’ Heaven open project presentation happened, as I was looking for projects that I might enjoy, I sent an email to Michael and Rob about working on Soylent. I liked the project, but by the time I was offered the position, it honestly wasn’t the project itself but rather the people involved that convinced me. Not to disparage any other groups, but I knew that there was a great deal I could learn from working with Michael and Rob, and that was ultimately what I was most looking for in a summer UROP.”

Leave a Reply

Replies which add to the content of this article in the EECS Newsletter are welcome. The Department reserves the right to moderate all comments. If you would like to provide any updated information please send an email to newsletter@eecs.mit.edu.