As Media Lab students in 2010, Karthik Dinakar SM ’12, PhD ’17 and Birago Jones SM ’12 teamed up for a category project to construct a tool that will help content moderation teams at corporations like Twitter (now X) and YouTube. The project generated an enormous amount of pleasure, and the researchers were invited to provide an indication at a cyberbullying summit on the White House — they only needed to get the thing working.

The day before the White House event, Dinakar spent hours attempting to put together a working demo that might discover concerning posts on Twitter. Around 11 p.m., he called Jones to say he was giving up.

Then Jones decided to take a look at the info. It turned out Dinakar’s model was flagging the correct kinds of posts, however the posters were using teenage slang terms and other indirect language that Dinakar didn’t pick up on. The issue wasn’t the model; it was the disconnect between Dinakar and the kids he was attempting to help.

“We realized then, right before we got to the White House, that the people constructing these models shouldn’t be folks who are only machine-learning engineers,” Dinakar says. “They must be individuals who best understand their data.”

The insight led the researchers to develop point-and-click tools that allow nonexperts to construct machine-learning models. Those tools became the premise for Pienso, which today helps people construct large language models for detecting misinformation, human trafficking, weapons sales, and more, without writing any code.

“These sorts of applications are necessary to us because our roots are in cyberbullying and understanding use AI for things that actually help humanity,” says Jones.

As for the early version of the system shown on the White House, the founders ended up collaborating with students at nearby schools in Cambridge, Massachusetts, to allow them to train the models.

“The models those kids trained were so significantly better and nuanced than anything I could’ve ever give you,” Dinakar says. “Birago and I had this big ‘Aha!’ moment where we realized empowering domain experts — which is different from democratizing AI — was the most effective path forward.”

A project with purpose

Jones and Dinakar met as graduate students within the Software Agents research group of the MIT Media Lab. Their work on what became Pienso began in Course 6.864 (Natural Language Processing) and continued until they earned their master’s degrees in 2012.

It turned out 2010 wasn’t the last time the founders were invited to the White House to demo their project. The work generated a whole lot of enthusiasm, however the founders worked on Pienso part time until 2016, when Dinakar finished his PhD at MIT and deep learning began to blow up in popularity.

“We’re still connected to many individuals around campus,” Dinakar says. “The exposure we had at MIT, the melding of human and computer interfaces, widened our understanding. Our philosophy at Pienso couldn’t be possible without the vibrancy of MIT’s campus.”

The founders also credit MIT’s Industrial Liaison Program (ILP) and Startup Accelerator (STEX) for connecting them to early partners.

One early partner was SkyUK. The corporate’s customer success team used Pienso to construct models to grasp their customer’s commonest problems. Today those models are helping to process half 1,000,000 customer calls a day, and the founders say they’ve saved the corporate over £7 million kilos thus far by shortening the length of calls into the corporate’s call center.

“The difference between democratizing AI and empowering individuals with AI comes all the way down to who understands the info best — you or a health care provider or a journalist or someone who works with customers day-after-day?” Jones says. “Those are the individuals who must be creating the models. That’s the way you get insights out of your data.”

In 2020, just as Covid-19 outbreaks began within the U.S., government officials contacted the founders to make use of their tool to raised understand the emerging disease. Pienso helped experts in virology and infectious disease arrange machine-learning models to mine hundreds of research articles about coronaviruses. Dinakar says they later learned the work helped the federal government discover and strengthen critical supply chains for drugs, including the favored antiviral remdesivir.

“Those compounds were surfaced by a team that didn’t know deep learning but was capable of use our platform,” Dinakar says.

Constructing a greater AI future

Because Pienso can run on internal servers and cloud infrastructure, the founders say it offers an alternate for businesses being forced to donate their data through the use of services offered by other AI corporations.

“The Pienso interface is a series of web apps stitched together,” Dinakar explains. “You may consider it like an Adobe Photoshop for big language models, but in the net. You may point and import data without writing a line of code. You may refine the info, prepare it for deep learning, analyze it, give it structure if it’s not labeled or annotated, and you possibly can walk away with fine-tuned, large language model in a matter of 25 minutes.”

Earlier this 12 months, Pienso announced a partnership with GraphCore, which provides a faster, more efficient computing platform for machine learning. The founders say the partnership will further lower barriers to leveraging AI by dramatically reducing latency.

“If you happen to’re constructing an interactive AI platform, users aren’t going to have a cup of coffee each time they click a button,” Dinakar says. “It must be fast and responsive.”

The founders imagine their solution is enabling a future where more practical AI models are developed for specific use cases by the people who find themselves most aware of the issues they try to unravel.

“Nobody model can do every part,” Dinakar says. “Everyone’s application is different, their needs are different, their data is different. It’s highly unlikely that one model will do every part for you. It’s about bringing a garden of models together and allowing them to collaborate with one another and orchestrating them in a way that is smart — and the people doing that orchestration must be the individuals who understand the info best.”