Software Engineering Radio - the podcast for professional software developers SE Radio 661: Sunil Mallya on Small Language Models
37 snips
Mar 25, 2025 Sunil Mallya, Co-founder and CTO of Flip AI, shares his expertise on small language models (SLMs) versus large language models (LLMs). He delves into their differences, revealing how SLMs can be more efficient and accurate for specific tasks. Sunil highlights the importance of domain-specific training datasets and discusses recent advancements like the DeepSeek R1 that show smaller models outperforming larger ones in particular contexts. He also touches on the evolving landscape of model deployment and how organizations can optimize performance while managing costs.
AI Snips
Chapters
Transcript
Episode notes
SLM Definition
- Small language models (SLMs) are defined by practicality, not size.
- They're easily accessible and don't need top-tier hardware, unlike large language models (LLMs).
Mixture of Experts
- Mixture of expert models use a subset of their parameters.
- Not all parameters are active during inference, unlike in models like GPT-3.
Parameters Explained
- Language model parameters are analogous to neurons.
- More parameters generally mean larger memory requirements, impacting computational footprint.