Small Language Models Offer Public Sector AI Solution Amid Security and Infrastructure Constraints
Public sector entities encounter significant hurdles in implementing AI, primarily due to security obligations, governance issues, and operational challenges. According to a Capgemini report, 79% of executives in the public sector express worries about the security of AI-related data. Han Xiao, the AI vice president at Elastic, points out difficulties related to data management and network accessibility. Many public organizations lack familiarity with GPU hardware necessary for AI models. Small language models (SLMs), which utilize billions of parameters, present a less resource-intensive option for local use. Research indicates that SLMs can perform equally well or even surpass large language models (LLMs). Gartner forecasts that by 2027, SLMs will be utilized three times more than LLMs, improving transparency, auditability, and adherence to regulations like GDPR.
Key facts
- 79% of public sector executives globally express concerns about AI data security according to Capgemini study
- Small language models (SLMs) use billions rather than hundreds of billions of parameters
- 65% of public sector leaders struggle to use data continuously in real time at scale per Elastic survey
- Gartner predicts by 2027 small specialized AI models will be used three times more than LLMs
- Government agencies often operate in environments with limited or unreliable internet connectivity
- Public sector organizations typically lack experience managing GPU infrastructure
- SLMs can be housed locally offering greater security and control over sensitive data
- AI can search across unstructured data including PDFs, images, spreadsheets and recordings in multiple languages
Entities
Artists
- Han Xiao
Institutions
- Capgemini
- Elastic
- Gartner
- MIT Technology Review
Locations
- Europe