Meta at SXSW: How Llama and open source AI is leveling the playing field and enabling innovation

Open source artificial intelligence, including models like Llama, is enabling people to create new opportunities that are transforming the way we work and live. As leaders in this field, Meta wants to ensure that everyone has the potential to reap the economic rewards of AI-enabled innovation.

This month, we were excited to continue that conversation at SXSW, an annual technology conference in Austin, Texas, where we engaged with policy leaders, researchers, developers, and our partners across the industry.

We kicked off the week with a panel featuring Rebekkah Hogan, Director of GenAI Program Management at Meta, and our AI Alliance partners from IBM and the Aspen Institute. Hogan and our partners discussed the newly published “Guide to Essential Competencies for AI,” where—together—we emphasize the unique chance we have to foster alignment and coordination across various sectors, particularly in aligning educational pathways with the evolving needs of employers. As AI technologies become even more integral to various sectors, open source AI will play a crucial role in workforce development and improving productivity.

“There’s a huge opportunity not only to change the way in which technical skills are being deployed in the university, but also across companies and workforces,” Hogan said. “There’s a big gap that we see between what is available and what is already in mandated curricula and where we feel people could have an impact when it comes to future training and future skills. Having a pillar—a working group—based on skills and education seemed like a natural place where we could align on what should be the standards we should push for.”

Later in the week, we sponsored a full day of AI policy programming and demos at SeedAI House. SeedAI is a nonprofit, nonpartisan AI policy organization working to ensure AI policy meets the real needs of communities across the country while advancing US leadership on the world stage. At the event, we welcomed elected officials and other policy leaders, our industry peers, developers, researchers, journalists, and other interested parties who drove the discussion about the economic, scientific, and creative opportunities open source AI is creating.

Attendees heard from Polina Zvyagina, Director of AI Policy at Meta, and Michael Richards, Director of Policy at the U.S. Chamber of Commerce, about how we can democratize access to AI so everyone has the opportunity to benefit.

“The higher the adoption rate, the cheaper it becomes to run the models. The more efficient they are, the easier they're going to be to deploy,” Zvyagina said. “Whoever releases the best and the most efficient model, and releases it open source, that’s going to drive the highest adoption rate.”

This message was brought to life by businesses, startups, and developers who showed how Llama, Meta’s open source AI model, is being used to drive innovation, productivity, and societal value. We were joined by:

• WriteSea, an Oklahoma-based tech company that uses Llama to support Job Search Genius, an AI career coach designed to help candidates stand out and enhance the job search experience. Job Search Genius provides an essential resource for candidates who need help reaching better outcomes in a competitive job market.
• Authorship, a North Carolina-based company that uses Llama for AI legal and compliance agents to give users end-to-end control of sensitive information. From private on-device AI to compliance control over AI datasets, Authorship uses Llama to provide essential data controls for regulated industries.
• Zauron Labs, a Texas-based company focused on improving healthcare outcomes. Their proprietary tool, Guardian Pro, uses Llama to identify likely false negatives in medical exams and flags these for second opinions and collaborative learning, like a spell-checker for radiologists.
• Cofactor, a company based in Chicago and Austin that uses Llama to help healthcare providers recover lost revenue by automating the appeals process for insurance denials.
• HiiiWAV, an incubator based in Oakland, Calif., that teaches artists and musicians how to build startups and new creative tools using Llama and other open source AI models.

SeedAI House also hosted a discussion about the power of personalization in AI. Lisa Titus, AI Policy Manager at Meta, joined other industry experts to discuss how future AI experiences will be highly personalized and people will be empowered to create the AI products and services they need to suit their personal needs and goals. Titus shared how personalization has always been a core part of Meta’s products.

“In the social media space, how people interact shapes their further interactions so they can have experiences that are more tailored to their interests,” Titus said. She explained how Meta is continuing this with our AI experiences, including AI Studio, where everyone can build their own AIs.

“We really believe a one-size-fits-all approach doesn’t work best for most people,” Titus said. “Providing that personalized experience means that people are going to be more engaged and want to use the technology and that it is going to suit their needs better.”

During a fireside chat at the conference, Ella Irwin, Head of AI Trust at Meta, shared more details about how we’re building AI in ways people can benefit from the technology, while making sure we have the right protections in place to build consumer trust in a new technology. This involves continuous testing and assessment of our models and features.

“We’re learning and constantly evolving our trust process as we look at what people are creating and how those assistants are being used,” she said. “By far, the use cases are positive. We see lots of great things happening, how users are interacting with these assistants.”

At the UK House, Meta participated in a discussion about the role governments can take to support AI innovation. Zvyagina represented Meta on a panel that discussed actionable steps governments can take to get involved, including helping to develop datasets, adopting open source models to solve policy goals, and investing in AI infrastructure to enable AI globally.