Keynote – Alexandr Wang
Summary
Wang announced Meta’s upcoming release of omni‑lingual models covering >1 600 languages, with a focus on Indian languages. He highlighted Meta’s investment in personal superintelligence—AI assistants that understand individual goals and act proactively across devices.
Key notes:
- Multilingual expansion: New models significantly improve low‑resource language performance, aiding inclusion.
- AI as a personal assistant: Future agents will manage health plans, event organization, and personal productivity.
- Safety & transparency: Meta will publish model cards, benchmarks, and conduct extensive red‑team testing.
- Collaboration with India: Meta is providing datasets in ten major Indian languages and partnering with local developers.
Wang concluded that open, responsible AI will be the backbone of the next wave of consumer services.
Key Takeaways
- Omni‑lingual models: >1 600 languages, with significant gains for Indian dialects.
- Personal superintelligence vision: AI assistants that act as extensions of the user.
- Safety by design: Model cards, audits, and red‑team evaluations will be standard.
- India partnership: Data‑sharing and co‑development to tailor AI for local contexts.