Google DeepMind 发布 Genie 4 世界模型，文字提示可生成可探索的交互世界
🌎 Google’s Genie 4 turns short text prompts into interactive worlds you can explore with your keyboard. It generates visuals and sound together, while Gemini helps guide events and maintain continuity. Beyond creative experiments, Google is researching whether these worlds could provide more varied training environments for robots and AI agents. Robots already learn in simulations, but generating new scenarios could expand where they practice. The challenge is reliability, however, as objects can drift, details can change, and complex physics can fail. Creating a convincing world is one important step. Making its lessons useful in the real world is the bigger test ahead.
Google DeepMind 发布 Genie 4 世界模型，输入短文字提示即可生成可用键盘探索的交互环境，并同步生成环境音、音乐和随场景变化的角色对白。Gemini 3.8 Flash 会逐帧检查生成画面、把目标拆成步骤并随场景变化调整计划，Genie 4 还改进了人物动作以及碰撞、空气动力学等物理交互。
粉丝 8.6万
The latest in AI and robotics, explained Founder & CEO of @The Rundown AI Zurich 🇨🇭 / NYC 🇺🇸
- 浏览
- 594
- 点赞
- 14
- 评论
- 1
- 分享
- 0
- 收藏
- 4
平台标注：AI 生成内容
数字来自平台，为最近一次抓取时的数据。
同一新闻，精选展示《Google DeepMind 发布世界模型 Genie 4，支持实时交互与原生音频生成》
以下是视频口播的自动转写，可能有错字。
So watch what happens when i type an underwater city and then click this button. I've been testing out google deepmind's new genie 4 world model, which generates environments.
You can explore using your keyboard and while you're inside, you can do things like change the weather, drop in a character or object or well, pretty much anything just by typing.
What's cool is that it also generates sound alongside the visuals including ambient noise music and characters whose dialogue adapts to what's actually happening.
The next train will leave in two minutes. But if we take a step back, here's how google tries to keep this world from falling apart.
Basically, a second ai called gemini 3.8 flash watches the incoming frames as genie creates them. It breaks goals into steps, checks whether each step happened, and adjusts its plan when you actually change the scene.
Google says that genie 4 also improves human movement and physical interactions, including collisions and aerodynamics. But to be clear, it's still very experimental.
Objects can change unexpectedly, visual details drift, and complicated physics can break down. But overall, this is very important for google's bigger ambition.
Which is creating environments where robots and ai agents can practice before acting in the real world. If you want to stay in the loop with the latest in ai and robotics, follow for more.
来源：TikTok · Rowan Cheung · tiktok.com