Jordan Nanos, HPE & Enayet Shabon, Peraton Inc. | HPE Discover 2024
summary
Claims that data gravity still forces models to move to data and that HPE’s GreenLake/Flex-style stacks enable private AI deployments with GPUs at the edge. Demonstrates a prototype pipeline that chains six models (ASR, translation, retrieval, chat, TTS) to run on ships and vehicles, argues 2024 is the year of inference, and insists GPUs are needed broadly. Notes GPU supply pressure (H100 lead times quoted ~4–10 weeks) and a debaters-and-judge prototype to reduce hallucinations.
AI-written summary · about
chapters (14)
- 0:00Welcome and introductions
- 0:49About Peraton Inc
- 2:16AI use cases and GenAI
- 3:21Private vs public AI
- 5:26Hybrid AI needs
- 7:18Year of inference
- 9:44GPUs for inference
- 11:32GPU supply challenges
- 13:17Workforce skills for AI
- 15:45Internal productivity use cases
- 16:48Combining multiple AI models
- 18:13Hybrid AI is real
- 20:02Future of AI reasoning
- 21:28Human role in AI
