Navigating the Divide: Open vs. Proprietary AI
summary
Panel states enterprises multisource models are normal—over 80% use more than one model—and gives concrete tactics: use ChatGPT, Perplexity Pro, Notion AI and coding tools like Cursor for prototyping; run weekly triage on thumbs‑down feedback; build evals before touching models because benchmarks lie; prefer small high‑quality fine‑tuning sets; expect post‑deployment monitoring, data flywheels and function‑calling APIs; demand model release notes and prepare for fast iteration toward multimodality and agentic AI within five years.
AI-written summary · about
chapters (10)
- 0:00Introduction and panelists
- 1:03AI tools in daily life
- 4:37Where AI adds value today
- 10:32Choosing open vs proprietary models
- 17:56Production and operational challenges
- 26:22Best practices for AI data lifecycle
- 34:11API-first in the AI world
- 38:43Safety and guardrails in AI
- 42:34Predictions for AI in 2029
- 46:29Advice for developers and Postman
related talks
58mPanel Discussion: AI- Friend or Foe?
56mPANEL: Unleashing the Future: The Impact of Gen AI on Development Productivity
11mOpen source GenAI vs. enterprise GenAI: Navigating the future of AI innovation
14mReliable Agentic AI, the Open Way
46mIntelligence as Infrastructure: What Developers Must Learn to Build with AI
44mKeynote: Tracking the Pulse of Generative AI
47mAI and the Future of Software Development
51mAssessing AI's progress