The New Stack Podcast AI Agents are Dumb Robots, Calling LLMs
Mar 20, 2025
Mark Hinkle, CEO of Peripety Labs and a veteran in serverless technologies, discusses the intriguing role of AI agents in software development. He describes these agents as 'dumb robots' that automate tasks but rely on large language models for real intelligence. Hinkle highlights the transformative potential of AI in creating bespoke tools on-the-fly and emphasizes the necessity of effective management practices. The conversation also spans his three-decade journey through technological evolution and the shift towards innovative GPU utilization in AI.
AI Snips
Chapters
Transcript
Episode notes
Shifting Computational Entities
- Serverless functions and AI agents represent a shift in how we view computation.
- Traditional computation is bound by persistence and state, but serverless and AI agents are not.
Serverless and AI Agents
- Serverless computing reduces resource consumption by only using resources during task execution.
- AI agents, as goal-oriented code, can leverage serverless functions for tasks like querying APIs or exchanging data.
Memory, GPUs, and LLMs
- Large language models (LLMs) use a context window as short-term memory, and its size is crucial.
- GPUs execute LLM algorithms, and memory bandwidth is becoming a bottleneck.