Keynote Panel: Open Innovation in AI - Present an...
summary
Research groups demonstrate estimating full human pose from sparse sensors such as laser reflections, event cameras, and even acoustic-only inputs, and use simulation to create niche datasets. A labeling company says half its workload is LLM data, employs over 1,200 annotators with large speedups versus a comparator and manages end-to-end workflows while training refugees and disabled workers. An engineering team open-sources models and training code (OPT, Llama) to boost reproducibility and hiring, but faces bottlenecks from compute costs, dataset licensing, and privacy.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 10:42Sparse measurement for state estimation
- 20:22High-quality datasets and labeling
- 30:35Open source and AI research
- 41:45Correctness in generative AI
- 45:18Popular open source projects and best practices
- 47:54Roadmap for openness in AI
- 50:48Meta's open source strategy
related talks
33mKeynote: Fine-tuning our way towards openness in AI
39mKeynote Panel State of AI in the Enterprise
40mKeynote: Open-source Multimodal AI in the Wild
16mKeynote: Open Source AI Everywhere
13mKeynote: State of Open GenAI
20mKeynote: Open-Sourced Inspiration - The Present and Future of Tech - AI
31mKeynote Panel: Gen AI: Opportunites & risks
14mKeynote: Building AI with Open Source & Hugging Face