Karthik Dinakar and Birago Jones, both MIT graduates, developed a tool facilitating social media content moderation as part of a class project in 2010. The project generated significant interest and was demonstrated at a White House cyberbullying summit. However, they initially struggled with their demo due to their unfamiliarity with teenage slang used in harmful posts. This led them to realize that machine-learning engineers might not always be the best people to build these models, as understanding the targeted content is key to developing effective filters.
This realization resulted in the development of Pienso, a solution that enables non-experts to create AI models without writing any code. With Pienso, users can build machine-learning models for various purposes including misinformation detection, human trafficking prevention, and weapon sales monitoring. The early version of the system was improved by collaboration with local students, forming a significant aspect of the founders’ commitment to empowering domain experts.
Both founders were part of the Software Agents research group at MIT’s Media Lab. The initial work on Pienso started in a Natural Language Processing course and continued during their graduate studies. The project attracted further interest from the White House, demonstrating the enthusiasm it created.
Partnerships aided in the development of Pienso. SkyUK used the tool to understand customer issues more effectively, leading to productivity improvements and cost reductions of over £7 million at their call center. Furthermore, amidst the COVID-19 pandemic, governmental authorities reached out to leverage Pienso in understanding the new disease.
Pienso’s versatility, as it can function on both internal servers and cloud infrastructure, positions it as an alternative for companies having to share their data with external AI service providers. Pienso is like an Adobe Photoshop for large language models, enabling users to import and refine data, arrange it for deep learning, and walk away with custom-tailored machine-learning models in under half an hour.
Pienso announced a partnership with GraphCore earlier this year, which is expected to lower AI usage barriers by reducing latency. The founders believe their tool empowers individuals most familiar with specific problems to create unique AI models to address these issues, challenging the notion of a single. They envision a future where a “garden of models” work in collaboration, orchestrated by those with a deep understanding of the data they are dealing with.
