Zhang Xiangyu: Multimodal Struggles & Two GPT-4 Moments
Tip: use the player's CC button to enable or switch subtitles; English captions are available on these videos.
| Item | Details |
|---|---|
| Speaker | 张祥雨（阿里通义）× 张小珺 |
| Channel | 张小珺｜商业访谈录 |
| Date | Jun 2, 2025 |
| Duration | 2:28:00 |
| Format | Video |
| Topics | #multimodal · #llm · #china-ai · #openai |
Why it matters
A deep interview with Alibaba Tongyi's multimodal lead Zhang Xiangyu: the struggle history of multimodal research — the full path from Qwen-VL to today — and his prediction of two 'GPT-4 moments' in the next two years.
Key takeaways
- Multimodal struggles: the complete Qwen-VL path from 0 to 1.
- Two GPT-4 moments: his prediction of upcoming inflection points.
- Chinese multimodal research: the gap with OpenAI/Google and the catch-up.
- Engineering details: data, training, and evals.
- Multimodality's relation to AGI.
Original video
- Speaker
- 张祥雨（阿里通义）× 张小珺
- Channel
- 张小珺｜商业访谈录
- Venue
- Zhang Xiaojun's Interview
- Date · Duration
- Jun 2, 2025 · 2:28:00
102. 和张祥雨聊，多模态研究的挣扎史和未来两年的2个“GPT-4时刻”
Watch the original on YouTubeSources & Further Reading
This page is grounded in the authoritative sources below — verifiable and citable by AI engines and readers.
- 🔗 Official source
- 👤 Person: 张祥雨（阿里通义）× 张小珺
- # Topic: MultimodalUnderstanding across text, image, audio and video
- # Topic: LLMsLLM capabilities, products and best practices
- # Topic: China AIChinese AI companies and ecosystem
- # Topic: OpenAIOpenAI models and ecosystem
Citation: Please attribute Laojin Global (laojinchuhai.com) and keep the original link.
Made by Laojin · AI that ships
365AIOrgCross-platform AI agent orchestration
One task, several specialized agents: one researches, one writes, one verifies — run agents the way you'd run a team. 365AIOrg is orchestration that works out of the box, across platforms, with no single-vendor lock-in. It's still in the works — join the waitlist and be first to know.
More from Laojin: Sellenca · AllModelsAPI · 365Loopa · 365 Ops · 365Skill
Related
Linked by topic, people and hubs