Kristen Smith
Wes Green, senior vice president of global public sector for Invisible Technologies, has encouraged public sector agencies to adopt a hybrid artificial intelligence solution comprising large language models and small language models.
In an opinion piece published on Federal News Network on Tuesday, Green called the AI model combination “AI mullet”–derived from a hairstyle characterized by short hair on the sides and top, with longer hair at the back and often described as business in the front, party in the back.
The AI mullet concept identifies SLMs as the business in the front, dealing with structured, mission-critical tasks, and LLMs as the party in the back, where broad, complex challenges are addressed, the executive explained.
Hybrid AI: Key to Efficiency, Flexibility and Innovation
According to Green, SLMs are optimized for efficiency and precision and could help public sector agencies ensure compliance, security and cost efficiency, while LLMs, with their capability to handle dynamic, open-ended inquiries and strategic analysis, could enable agencies to innovate and engage with citizens in new ways.
AI Mullet Approach Advantages
The AI mullet approach could enable agencies to protect sensitive data in compliance with strict security regulations.
“SLMs ensure that sensitive data stays locked down in a controlled environment,” Green said. “Meanwhile, LLMs can handle large-scale analytics, broad research inquiries and citizen engagement efforts without exposing secure data.”
The combined technology is also expected to bring cost optimization by offloading routine, structured tasks to SLMs and keeping LLMs available for complex problem-solving and strategic initiatives. According to the executive, using LLMs’ compute power even for simple tasks can be expensive.
Best Practices for AI Mullet Adoption
Green provided best practices to maximize the impact of the AI mullet implementation, including ensuring ethical use, bias mitigation and compliance; investing in adaptable infrastructure to support both SLMs and LLMs without creating bottlenecks; and prioritizing interoperability to enable seamless integration across departments and use cases.
