There is no question that artificial intelligence is going to shape the future. But who shapes AI?
Right now, there is an enormous knowledge imbalance in the field of AI. Massive tech companies (mostly American) are monopolizing research and innovation. Emerging communities that could truly benefit from open-source, society-bolstering AI projects are lacking in know-how and educational resources.
Jan Carbonell, a Barcelona-based developer, co-founded Saturdays.AI specifically to combat this information gap. Their signature offering, AI Saturdays, is a 15-Saturdays intensive course designed to teach anyone with a motivation to code how to build AI-powered projects. “If you look at Latin American and Africa,” says Carbonell, “by population size, they should have a lot of impact on artificial intelligence, but if you look at the big companies, they have almost no presence there. In top-tier journals of artificial intelligence, there's barely any publications from the universities in those areas.”
Standardizing the quality of these educational programs across 25 cities and 11 countries (and growing) came with its own set of challenges.
The Challenge of Keeping Courses Standardized While Scaling
Carbonell’s beginner AI course ran well in Barcelona, but when Saturdays.AI started opening training sessions in other cities in Spain and abroad, they ran into issues trying to systemize and share teaching materials, as well as issues with time management, trying to evaluate student projects.
In the beginning, Saturdays.AI took a haphazard approach to standardizing their curriculum. Each location created course content however they saw fit using Google Docs, Slack, or other tools. This made it difficult to guarantee that students were getting a uniform course experience. “We had no visibility of the changes that other cities were making,” says Carbonell. “It was challenging because, even though we were one organization, we felt we had to start over with each new course offering.”
As courses started to grow in size and scope, the organization ran into a second issue: providing useful and personalized feedback on each student’s coding exercises. This came to a head during another initiative they launched, called Akademy.AI. “AI Saturdays are around 80 hours,” says Carbonell, “and the Akadamy.AI bootcamp was much more intense, maybe 400 hours over the whole course. Initially I intended to correct the exercises from the students, but that was impossible because we taught two-hour lectures every morning, and then they did six hours of exercises. Even if we only had 10 students, for me to be able to correct the 60 hours of work that were being produced by them on a daily basis was unfeasible.”
This problem was compounded by the fact that many of the instructors were volunteers, with a full-time job from Monday to Friday, who couldn’t dedicate that much of their time after work hours. AI Saturdays needed to find a way to standardize the curriculum and take some of the pressure off of the instructors.
Eduflow Helps Keep Students on Track
Carbonell came across Eduflow and immediately seized on it as a potential tool for regulating AI Saturdays' course offerings. “I started using it, just to test a couple of things, and I saw the framework was super simple. I could create a basic process for what every course instructor needs to do in a very straightforward way.”
Not only did structuring course materials with Eduflow help Saturdays.AI standardize their offerings, but it also provided a mechanism for keeping students engaged and on track with their lessons. Because the lessons take place over the course of 15 weekends, it’s not unusual for students to miss a class for personal reasons. This becomes a problem when students try to skip ahead and are overwhelmed by more complicated material. Whereas, if all the materials are widely available in a Google Docs folder, students can pick and choose what to tackle. Eduflow’s flows require students to complete each week’s lesson before moving onto the next, preventing them from skipping ahead or falling too far behind. They have no choice but to follow the learning path laid out by the curriculum.
This shifts accountability from the teachers onto the students. “We’re saying ‘if you want to finish this course, come to the demo day, or get accreditation, then you need to complete the work. There is no pass or fail; you just have to deliver something,” says Carbonell.
Peer Review Takes the Burden off Instructors
In addition to regulating the student’s workload, Eduflow’s peer review function helps shift the work of evaluation away from the teachers and onto the students. Now students can give each other individualized feedback and code corrections on a deeper level than instructors could realistically provide.
"I came across the peer feedback function, and thought ‘wow, this is the perfect solution,’” says Carbonell.
Students aren't just checking if the answers are right or wrong; they're looking closely at how their classmates have done their work. This helps both the student who is receiving feedback and the student who is performing the assessment to develop a better understanding of the coding process.
Student-Driven Learning Helps Saturdays.AI Fulfill Their Mission
So far, Saturdays.AI has rolled out Eduflow’s features in their busiest cities: Barcelona and Madrid. They’ve seen a big jump in engagement, and Eduflow’s analytics features let them keep closer tabs on student progress.
"Every online community struggles with the fact that you keep giving your students content but they never respond to it." says Carbonell. "They just keep absorbing it and they do the work, but you never truly know who is doing it. As a teacher myself, I feel very responsible because if someone's lost, I want to know at that very moment, right? The cool thing is Eduflow enables that. There's a tracking system that tells you, listen, this person hasn't delivered on the last three sessions."
In March 2020, Saturdays.AI launched a Covid-19 AI Challenge. They are asking participants to propose a data science solution to fight Coronavirus crisis using Machine learning, Data Visualization, Deep learning, Dashboards, and much more. They are using Eduflow to onboard participants, keep them engaged, and to facilitate project submission.
Soon they will start employing Eduflow abroad as they expand their offerings beyond Latin America into Africa, India, and Southeast Asia. There are also plans to start offering courses within private organizations and schools. “We want to scale without compromising on the quality of the education of the students,” says Carbonell. “I think that Eduflow is very essential to our core mission because it enables us to use technology to scale in a meaningful way.”