How Open‑Source AI Is Shaping the Future: Insights from Xiaohongshu’s AMA
In a 1024 Programmer's Day AMA, Xiaohongshu’s tech leaders and Hugging Face co‑founder discussed AI coding as a core skill, open‑source as a catalyst for rapid advancement, and the challenges of achieving human‑like AGI, offering practical takeaways for developers and product teams.
On the 1024 Programmer's Day, Xiaohongshu REDacademy and the Tech Potato community hosted a live AMA featuring Xiaohongshu’s Vice President of Technology KAIQI, voice‑model lead Feng Long, and Hugging Face co‑founder Thomas Wolf.
AI coding is becoming a foundational capability.
Open source is the core engine accelerating technology development.
The path to AGI with a "human‑like" feel is clear but still faces many hurdles.
Open Source vs. Closed Source : All three guests agreed that open source drives rapid progress. Xiaohongshu is contributing models—from text and multimodal to the FireRed series of voice models—to let developers and small‑to‑medium enterprises share AI benefits.
AGI and the Future : They noted that general AI will arrive, yet challenges remain in creativity, complex environment perception, and data. Wolf warned that current LLMs lack the ability to formulate novel problems, a key bottleneck for true AGI.
Making AI More Human : KAIQI introduced the concept of a "human trainer" to give AI a more natural, empathetic interaction style, aiming for a community with a "living‑person" feel.
AI as a Personal Skill : The discussion highlighted a shift in core competencies for technologists—defining problems, framing them, and using AI as a tool rather than merely coding.
Product Management with AI : AI lowers product barriers, enabling non‑technical colleagues to build products alone. Understanding AI limits, prompt engineering, and fine‑tuning become essential skills for product teams.
AI Coding in Practice : KAIQI shared that Xiaohongshu open‑sourced a 1.0 text model (its first MoE model), a multimodal model comparable to Gemini and Seed, and an OCR model ranking fourth on Hugging Face. In voice, they released FireRedASR (top‑ranked on 50+ benchmarks) and FireRedTTS, achieving SOTA in podcast generation. These releases aim to democratize large‑model usage.
Future Outlook : Wolf emphasized that AGI will not appear instantly but will bring multi‑functional improvements, including robotics and new scientific discoveries. He urged users to critically evaluate AI outputs and consider error rates. Both Wolf and KAIQI stressed that AI will continue to evolve as a powerful assistant, but human direction remains crucial.
Xiaohongshu Tech REDtech
Official account of the Xiaohongshu tech team, sharing tech innovations and problem insights, advancing together.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.