As Media Lab students in 2010, Karthik Dinakar SM ’12, PhD ’17 and Birago Jones SM ’12 teamed up for a category project to develop a tool that will help content moderation teams at corporations like Twitter (now X) and YouTube would support. The project caused quite a stir and the researchers were invited to reveal at a cyberbullying summit on the White House – they simply needed to make it work.

The day before the White House event, Dinakar spent hours putting together a working demo that would discover worrisome posts on Twitter. Around 11 p.m., he called Jones and told him he was giving up.

Then Jones decided to take a look at the info. It turned out that Dinakar’s model flagged the suitable forms of posts, however the posters used teenage slang terms and other indirect expressions that Dinakar didn’t understand. The problem wasn’t the model; It was the disconnect between Dinakar and the teenagers he desired to help.

“Shortly before we got to the White House, we realized that the people constructing these models shouldn’t just be machine learning engineers,” says Dinakar. “It must be individuals who understand their data best.”

The findings led researchers to develop point-and-click tools that allow non-experts to construct machine learning models. These tools became the muse for Pienso, which today helps people construct wealthy language models to detect misinformation, human trafficking, arms sales, and more without having to put in writing code.

“Applications like this are necessary to us because our roots lie in cyberbullying and understanding find out how to use AI to do things that basically help humanity,” says Jones.

As for the early version of the system shown on the White House, the founders eventually worked with students from nearby schools in Cambridge, Massachusetts, to offer them the chance to coach the models.

“The models these kids trained were so a lot better and more nuanced than anything I could have ever imagined,” says Dinakar. “Birago and I had this big ‘Aha!’ The moment we realized that empowering subject material experts – which is different from democratizing AI – was the perfect path forward.”

A project with meaning

Jones and Dinakar met as graduate students within the MIT Media Lab’s Software Agents research group. Her work on what became Pienso began in course 6.864 (Natural Language Processing) and continued until her master’s degree in 2012.

It seems that 2010 was not the last time the founders were invited to the White House to reveal their project. The work sparked great excitement, however the founders worked on Pienso part-time until Dinakar accomplished his PhD at MIT in 2016 and deep learning became increasingly popular.

“We are still connected to a number of people on campus,” Dinakar says. “The experience we had at MIT, merging human and computer interfaces, has expanded our understanding. Our philosophy at Pienso wouldn’t be possible without the vibrancy of the MIT campus.”

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

An early partner was SkyUK. The company’s customer success team used Pienso to construct models to grasp their customers’ commonest problems. Today, these models help handle half one million customer calls day-after-day, and the founders say they’ve saved the corporate over £7 million to this point by reducing the length of calls in the corporate’s call center.

“The difference between democratizing AI and empowering people through AI is “who understands the info best – you or a health care provider or a journalist or someone who works with customers day-after-day?” says Jones. “These are the individuals who must be creating the models. This is the way you gain insights out of your data.”

In 2020, just as Covid-19 outbreaks began within the US, government officials contacted the founders to make use of their tool to raised understand the emerging disease. Pienso helped virology and infectious disease experts arrange machine learning models to research 1000’s of research articles on coronaviruses. Dinakar said they later learned that the federal government’s work helped discover and strengthen critical supply chains for medicines, including the favored antiviral remdesivir.

“These connections were discovered by a team that was not acquainted with deep learning but was capable of use our platform,” says Dinakar.

Building a greater AI future

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

“The Pienso interface is a series of web apps stitched together,” explains Dinakar. “Think of it like Adobe Photoshop for big language models, but on the net. You can view and import data without writing a line of code. You can refine the info, prepare it for deep learning, analyze it, give it structure if it is just not labeled or annotated, and you’ll be able to get a fine-tuned, wealthy language model in 25 minutes.”

Earlier this yr, Pienso announced a partnership with GraphCore, providing a faster and more efficient computing platform for machine learning. The founders say the partnership will further lower the barriers to deploying AI by dramatically reducing latency.

“If you construct an interactive AI platform, users won’t drink a cup of coffee each time they click a button,” says Dinakar. “It must be fast and responsive.”

The founders consider their solution enables a future through which more practical AI models for specific use cases are developed by the people most acquainted with the issues they try to resolve.

“No model can do the whole lot,” says Dinakar. “Everyone has a unique application, their requirements are different, their data is different. It is extremely unlikely that one model will do the whole lot for you. It’s about bringing together a garden of models and giving them the power to work together and orchestrate them in a way that is smart – and the individuals who do this orchestration must be the individuals who understand the info best.”

This article was originally published at