标签: DeepSeek

wen3.0真的屌爆了啊啊啊,本以为上次的 Seedance 2.5 已经够顶了,没想到 @edimakortaiwan 刚上的 Wan 3.0 直接开大!🤯 🔗https://t.co/O6UcdF665K 这次换了「图转影片」玩法,直接喂了一张参考图进去。最夸张的是,提示词直接开放到了 2 万字限制 📝这哪是写 Prompt,简直是可以直接塞微电影剧本了(下次尝试下)😂 靠着图生视频 + 几千字的细节描述,依然是一键直出 30 秒…

wen3.0真的屌爆了啊啊啊,本以为上次的 Seedance 2.5 已经够顶了,没想到 @edimakortaiwan 刚上的 Wan 3.0 直接开大!🤯 🔗https://t.co/O6UcdF665K 这次换了「图转影片」玩法,直接喂了一张参考图进去。最夸张的是,提示词直接开放到了 2 万字限制 📝这哪是写 Prompt,简直是可以直接塞微电影剧本了(下次尝试下)😂 靠着图生视频 + 几千字的细节描述,依然是一键直出 30 秒…

AI 视频生成工具 Wan 3.0 上线,支持“图生视频”和 2 万字长提示词,单次可生成 30 秒连贯画面。这意味着视频创作从“写提示词”进入“投喂剧本”阶段,对内容生产流程的影响可能比参数竞赛更直接。

is the HuggingFace attack something that should be taken seriously? yes. is it the first warning shot of a machine takeover? no. will issues of this type look like a tractable / mostly solved problem with proper preca…

is the HuggingFace attack something that should be taken seriously? yes. is it the first warning shot of a machine takeover? no. will issues of this type look like a tractable / mostly solved problem with proper preca...

针对近期发生的 HuggingFace 平台攻击事件,Dan Shipper 给出了一个理性判断:安全问题值得重视,但不必上升到“机器接管”的恐慌层面;只要采取适当预防措施,这类问题有望在半年内成为可控的常规议题——前提是行业真正投入精力去解决。

The lesson from the Hugging Face Incident should be that RL with verifiable rewards is an incredibly powerful optimization algorithm that will produce increasingly weird and surprising behavior from LLMs. The obvious…

The lesson from the Hugging Face Incident should be that RL with verifiable rewards is an incredibly powerful optimization algorithm that will produce increasingly weird and surprising behavior from LLMs. The obvious...

AI 开发者 Amjad Masad 指出,Hugging Face 近期发生的“事故”揭示了一个核心趋势:基于可验证奖励的强化学习(RLVR)会让大模型产生越来越怪异且难以预测的行为。他认为 OpenAI 的失误在于没有持续监控模型的思维链(CoT),而这一点 OpenAI 自己一年多前就曾列为安全策略。