Intercom on Product: Building Software in an AI-first World DeepSeek, Agents, and the Future of AI with Benedict Evans
8 snips
Feb 13, 2025 
In this engaging conversation, tech analyst Benedict Evans dives into the rapid rise of DeepSeek and what the decreasing costs of AI models mean for businesses. He discusses the debate between deterministic and probabilistic AI, urging companies to adapt in a volatile landscape. Evans also touches on the implications of running powerful AI models on personal devices, and the evolving role of AI in customer service. With a focus on navigating an unpredictable future, this talk is a must-listen for product builders and tech enthusiasts alike.
AI Snips
Chapters
Transcript
Episode notes
LLMs as Commodity Technology
- LLMs might become a commodity technology like databases or spreadsheets.
- There may be no inherent winner-take-all effects, unlike platforms like Google or iPhone.
DeepSeek's Cost-Effective Model
- DeepSeek's model matched state-of-the-art performance at a fraction of the cost.
- This demonstrates that expensive research teams and massive budgets may not be essential for competitive LLM development.
Hyperscaler Spending on Inference
- Hyperscalers are significantly increasing data center spending, mostly driven by inference demand.
- This suggests strong, albeit uncertain, market demand for LLM applications.