Traditional reliability meant consistency. Given identical inputs, systems produced identical outputs. Costs were stable and behavior predictable. Inference reliability on the other hand is shaped by nondeterminism. Outputs vary due to stochastic generation, retraining introduces drift, and token-based billing can cause cost fluctuations. The new dimension of reliability is semantic consistency, that is, the ability to deliver outputs of acceptable quality, accuracy, and predictability over time despite probabilistic behavior. In this episode of Pop Goes the Stack, #F5's Lori MacVittie and Joel Moses are joined by guests Ken Arora and Kunal Anand as they dive into the topic of reliability in #AI systems. They explore the concept of 'slop' (AI variability) as a potential feature rather than a bug, discuss the importance of contextual semantic consistency, and weigh guardrails and evals tailored to specific inference workloads. Tune in to learn how to navigate the evolving AI landscape and take note of practical tools and strategies like multi-model chaining, distillation, and prompt engineering to ensure reliability. Chapters: 00:00 Welcome to Pop Goes the Stack 00:47 Inference: Why reliability isn't what it used to be 02:00 The case for AI 'creativity' (Slop or Jazz?) 05:55 Context is key for reliability 06:45 Semantic consistency in agentic AI 07:55 The role of guardrails and context 09:16 Evals are critical to agentic composition 12:31 Defining and measuring semantic consistency 15:48 Is AI reliability subjective? 16:46 Multi-model chains and AI refinement 17:49 Staging vs. production in AI systems 19:00 Key takeaways: Know your problem, guardrails and evals, & pragmatism Find out more in the blog How AI inference changes application delivery: https://go.f5.net/cxd0imiv Learn how you can stay ahead of the curve and keep your stack whole with additional insights on #AppSecurity, multicloud, AI, and emerging tech: https://go.f5.net/h33oydn2 More about F5: https://go.f5.net/s12cpz9q Read our blog: https://go.f5.net/3dy52zk3 Follow us on LinkedIn: https://go.f5.net/s2arxbc3

 169      4