Evaluating online content for quality and civility is a significant challenge. Seekr addresses this issue with advanced AI solutions that help businesses ensure their content meets high standards.
The Code Together podcasts highlights Intel® Liftoff for AI Startups, and discusses the latest advances in artificial intelligence. The show is hosted by Tony Mongkolsmai, a Software Architect and Evangelist at Intel, who leads discussions with experts in the field, including program participants.
In the latest episode, Chief of AI Research and Development at Seekr, Stefanos Poulis explains how to build LLMs that assess civility and other content attributes. He also shares how Seekr’s tools can be adapted to meet the unique requirements of various businesses.
What exactly does Seekr offer?
Seekr's mission centers on fostering trust in the web, adapting its products to achieve this across various domains. One key product, Seekr Align, evaluates and analyzes diverse content types—text, images, videos, and audio—using AI.
It then scores content based on quality metrics like adherence to journalistic principles and absence of clickbait or biased language. This scoring system enhances trust, helping businesses and advertisers align with content that reflects their values, ensuring safe and principled content associations.
They take on the immense challenge of essentially cleaning up web content, including combating deep fakes and low-quality generated material.
In this episode, Tony addresses some of the challenges the Seekr team had to overcome in order to transform their internal tool into an external product.
Stefanos explains how they initially focused on ensuring AI could respect journalistic principles for news articles, so they defined these principles and built a system around them. They then gradually expanded to include broader content types like web documents, assessing quality by looking for profanity, toxicity, and mentions of sensitive topics.
This iterative approach involved solving specific problems and generalizing solutions, increasing complexity and capabilities over time. Finally, they refined the product for broader application and reliability, emphasizing careful data curation for fine-tuning.
“The same principles from computer science apply: you start with something simple, then expand and increase complexity slowly until you get to something powerful and general”, says Stefanos.
Another interesting discussion point from the podcast is how each business or application has its own set of principles, or 'Constitution,' that should govern its operations.
When you bring these principles to Seekr Flow, it will produce AI that respects them. “Whether you have a lot of documents or very few, Seekr Flow can handle it. If the system needs more information, it will prompt you for assistance. This approach not only ensures trust but also helps improve data collection and curation processes”, says Stefanos.
If you want to know more about the Seekr’s offering, and the role Intel® Liftoff for AI Startups has played, have a listen here: https://codetogether.podbean.com/e/seekr-building-and-enabling-trustworthy-ai-for-businesses
As a Software Tools Ecosystem Specialist at Intel, I’ve had the privilege of working on the dynamic GenAI initiative. My focus is on driving engagement with the software developer audience. I'm a proud team member of the Intel® Liftoff for Startups and the Developer Engagement, Relations, and Studio team.