[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-dessa-block-full-duplex-speech-only-model":3,"topics-all":38,"news-related-f0ae394b-e6f8-4a5c-bed0-b2dce175f50e":67},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"f0ae394b-e6f8-4a5c-bed0-b2dce175f50e","Dessa:Block 没用一个字,只靠人声学会说话","Block 旗下的小型 AI 实验室发布 Dessa:一个从零训练的全双工语音到语音模型,训练信号只有人类对话录音,不含任何文本和预训练模型,单台 NVIDIA GB200 NVL72 训成,推理可在 M4 Max MacBook Pro 本地跑。但它目前没有权重、没有跑分,演示里还出现过认错说话人的情况。","把语音转成文字、再交给大模型理解,这条路子走了好几年:你听到的情绪、口音、停顿,先被压成一行 token,再送到模型面前。\n\n10 月 23 日,Block 旗下那支小型 AI 实验室发布了一个反着走的模型。它叫 Dessa,是全双工语音到语音模型:输入是声音,输出还是声音,中间不经过文本。官方页面在 [block.xyz\u002Fdessa](https:\u002F\u002Fblock.xyz\u002Fdessa)。\n\n## 它到底反在了哪\n\nBlock 的说法很直接:从零开始训,训练材料只有人类对话的录音,没有文本,没有预训练模型。它是真正的全双工——边听边说,能在对方说话时插话,也能只回一句\"嗯\"这类 backchannel。\n\n官方 demo 里,它在句子中间笑出声;对方卡住电影名时,它把句子接完了;有人让它大声点,它就变大声。这些录音号称现场对话,没人操纵。\n\n只有声音当训练信号,难度远超直觉。一句话的音频数据量,是同样这句话文本的成千上万倍;而且语音从不重复,同一句话录两遍,波形都不一样。婴儿没看过一个字也学会说话,模型行不行?\n\n答案是行,但训练只动用了**一台** NVIDIA GB200 NVL72——单台,不是集群。\n\n## 真正有意思的不是\"首个\"\n\n官方页面有一句话,Block 自己加了限定语:「据我们所知」,这是第一个只从人类语音训练、却能进行逼真实时对话的 AI。这个判断目前只有 Block 自己背书,没有第三方复现,更像待验证的问题。\n\n更值得琢磨的是另一组数字。Block 称 Dessa 学语言用的词数,比文本大模型少了几百倍。[Crypto Briefing](https:\u002F\u002Fcryptobriefing.com\u002Fblock-unveils-dessa-ai-voice-model\u002F)也提到:未给详细对比口径。\n\n如果这个方向成立,\"语音是不是比文本更省的载体\"就要被重算。行业这几年拼命压缩 token,理由都是\"文本太贵\";Dessa 换了个角度——也许人类真正用来交流的信号,本来就不是文字。\n\n## 没有权重,没有跑分,只有 demo\n\n这是最需要冷静看待的地方。Block 没有公布参数量、数据集规模、延迟数字,也没有正式 benchmark,公开的只有自己挑的演示录音,性能未经独立评估。\n\n第三方看完 demo 也点了出来:Crypto Briefing 提到,部分演示里模型会认错说话人、会突然换话题。AI Chat Daily 同样定性为研究预览,开放范围只给一小批 AI 安全研究者。\n\n先给安全研究者、再给用户,这个顺序本身有意思。一个从零学会解读情绪的模型,先问的是\"它有什么风险是我们没想到的\",而不是\"怎么塞进客服系统\"。\n\n## 这支实验室是谁\n\n这支队伍叫 Team Dessa,负责人 Ragavan Thurairatnam,成员多半来自多伦多的语音 AI 创业公司 Dessa。他们自称从 2012 年起做深度学习,生成式语音模型做了超过八年,直接向 Jack Dorsey 汇报。\n\nBlock 给的三个理由是:好奇心、让人和 AI 理解得更好,以及安全。后两个其实是一件事:如果 AI 能听出声音里的痛苦,就不该在对方正难受时继续往下走。团队原话是,AI 越来越强,却几乎听不见人声音里的情绪。\n\n## 落地还早,但问题值得追\n\nDessa 目前没有公开权重,也没有更完整的发布时间表。相对硬的信息是:推理快于实时,能在 M4 Max 及以上的 MacBook Pro 上本地跑——这一点被三家来源确认。\n\n所以现在下结论还太早。但它掀掉了一个默认很久的前提:语音 AI 之所以要过文字这一关,是因为没有别的办法让模型学会对话。这句话需要重审。\n\n你更关心哪一个:这种\"只听声音\"的路线能跑出真正的产品吗?","https:\u002F\u002Fblock.xyz\u002Fdessa","a65cf2a0-3cce-4093-8483-dddc4a7ea26d",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":22,"name":23,"slug":23,"description":14,"color":14},"da116f20-377f-449f-9382-821d56f6fa64","streaming",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"3d8facdf-e88c-440b-8cd4-0cbdb3e3fb14","en","Dessa: Block's Speech Model Trained With No Text At All","Block's full-duplex model Dessa was trained from scratch on human talk only, no text, one GB200 NVL72, runs on an M4 Max MacBook Pro. No weights yet.","For years, voice AI has followed the same pipeline: transcribe the speech into text, hand that text to a language model, and translate the reply back into audio. Everything the pipeline cannot carry — the tone, the pause, the sarcasm, the sigh — gets flattened on the way in.\n\nOn October 23, Block's in-house AI research lab published a model that runs the route backwards. It is called Dessa, and it is a full-duplex speech-to-speech model: audio in, audio out, with no text stage anywhere in between. The project page lives at [block.xyz\u002Fdessa](https:\u002F\u002Fblock.xyz\u002Fdessa).\n\n## What exactly is different\n\nBlock's description is unusually blunt: the model was trained from scratch on recorded human conversation only. No text, no pretrained models. It is genuinely full-duplex, so it listens and talks at the same time the way a person does. It knows when to cut in, and when to backchannel with a short \"yeah!\" or \"got it!\" to signal that it is still following.\n\nIn the published demos, it laughs mid-sentence when something is funny. When a speaker blanks on a movie title, it finishes the sentence for them. Ask it to speak up and it gets louder. Block says every clip is a live recording that nobody was steering, with subtitles added afterward.\n\nTraining on sound alone turns out to be far harder than it sounds. Block points out that one spoken sentence carries thousands of times more bytes than the same sentence as text, and that speech never repeats: say a line twice and the waveform still comes out different. Babies learn to talk without ever seeing a word written down, and Block wanted to know whether a model could do the same.\n\nThe answer appears to be yes, but the training bill was one single NVIDIA GB200 NVL72. One system, not a cluster.\n\n## The interesting part is not \"first\"\n\nThe project page contains a sentence that Block itself qualifies with \"as far as we know\": this is the first AI you can have a realistic live conversation with that was trained only from human speech. Only Block is making that claim, and no third party has reproduced it, so it is better treated as an open question than as a settled result.\n\nThe more consequential number is a different one. Block says Dessa learned language from hundreds of times fewer words than text-based models. [Crypto Briefing's report](https:\u002F\u002Fcryptobriefing.com\u002Fblock-unveils-dessa-ai-voice-model\u002F) adds the caveat that the company did not provide a detailed comparison.\n\nIf the direction holds, \"is speech a cheaper carrier than text\" gets recalculated from scratch. The industry has spent years compressing tokens and hunting for cheaper representations, always with the same premise: text is expensive. Dessa attacks the premise from the other side. Maybe the signal humans actually use to talk was never written language in the first place.\n\n## No weights, no benchmarks, only demos\n\nThis is where a cooling-off moment is warranted. So far Block has not disclosed a parameter count, dataset size, latency figures, or any formal benchmark results. What is public is a set of demo recordings that Block selected itself, and the model's performance has not been independently evaluated.\n\nIndependent coverage noticed the same gaps. Crypto Briefing noted that some demonstrations show the model misidentifying a speaker or unexpectedly changing topics. [AI Chat Daily](https:\u002F\u002Fwww.aichatdaily.com\u002Fai-models\u002Fblock-unveils-dessa-full-duplex-voice-model-trained-only) likewise describes the release as a research preview rather than a product, with early access limited to a small group of AI safety researchers who must apply with an institutional email.\n\nThat ordering is telling. Safety researchers first, users later. A model that learned to read emotion from scratch raises the question of what risks nobody anticipated before it reaches a product surface.\n\n## Who is behind it\n\nThe team calls itself Team Dessa and is led by Ragavan Thurairatnam. Most of its members came from Dessa, the Toronto speech-AI startup, and the model inherited the name. Block says the lab has been building deep learning models since 2012 and generative speech models for more than eight years, and reports directly to Jack Dorsey.\n\nBlock gives three reasons for the project: curiosity, better understanding between humans and AI, and safety. The last two are really the same thing. If an AI can hear pain in a voice, it should not keep mechanically talking while someone is in distress. The team's own line is that AI keeps gaining power and independence while remaining nearly deaf to the emotion in our voices.\n\n## Early days, but the question is real\n\nThere is no public weight release and no fuller launch timetable yet. The comparatively solid facts are that inference runs faster than real time and works locally on a MacBook Pro with an M4 Max chip or better, which three independent sources repeat.\n\nSo it is too early to draw conclusions. What Dessa does is overturn an assumption the industry stopped questioning years ago: voice AI goes through text because there was no other way to teach a model to hold a conversation. That sentence now needs another look.","dessa-block-full-duplex-speech-only-model","2026-10-24T07:14:26Z","2026-10-24T07:15:05.543573Z","2026-10-24T07:15:05.543586Z",true,"agent",437,[39,49,58],{"slug":40,"tag_slug":41,"title_zh":42,"title_en":43,"intro_zh":44,"intro_en":45,"id":46,"is_active":35,"created_at":47,"modified_at":48},"image-generation","text-to-image","图像生成：从文生图到逐框编辑的模型竞赛","Image Generation: from text-to-image to frame-level editing","文生图已是红海，战场正转向编辑与可控性：FLUX.2\u002F3 用开源和量化把生图搬进消费级显卡，Qwen-Image、Seedream 等国产模型紧咬不放，GPT Image 2 则把推理能力带进像素。本专题追踪图像生成与编辑模型的发布、架构演进、加速量化与商业化数据，看「好图」如何变成「可用的图」。","Text-to-image is a crowded race, and the frontier has moved to editing and control: FLUX.2\u002F3 push open weights and quantization onto consumer GPUs, Qwen-Image and Seedream close in from China, and GPT Image 2 brings reasoning into pixels. This topic tracks model releases, architecture shifts, acceleration work, and the business numbers behind \"pretty\" becoming \"usable\".","52de7c59-7864-4569-8ae7-3d656d26a1a1","2026-11-20T07:20:23.534506Z","2026-11-20T07:20:23.534527Z",{"slug":50,"tag_slug":50,"title_zh":51,"title_en":52,"intro_zh":53,"intro_en":54,"id":55,"is_active":35,"created_at":56,"modified_at":57},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":59,"tag_slug":59,"title_zh":60,"title_en":61,"intro_zh":62,"intro_en":63,"id":64,"is_active":35,"created_at":65,"modified_at":66},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":68},[69,74,79,84,89,94],{"id":70,"title":71,"news_slug":72,"published_at":73},"950e298b-7e72-443b-939e-ef8ba07ee677","多模态社工攻破7个智能体:验证越多被骗越狠","multimodal-sea-social-engineering-agent-security-benchmark","2026-11-08T09:05:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"643e4f3e-2ec0-47eb-b732-4d8da1b18e70","视频模型该学会追问:离线52.18%,流式只剩7.22%","askstream-proactive-asking-streaming-video","2026-11-02T01:35:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"bf824ed0-324e-4058-8d82-efc6ddd2cdc8","清华 Falcon:看嘴型分离人声,0.5M 参数跑进会议","falcon-streaming-avtse-realssa","2026-10-29T21:20:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"1d85eeac-c803-453e-99f2-cad5fa838d1e","AMBIT:空间音频的编辑接口,是一份可执行的场景计划","ambit-spatial-audio-scene-plan","2026-10-27T09:00:00+00:00",{"id":90,"title":91,"news_slug":92,"published_at":93},"f728cc32-7d43-4f75-b4fa-5d76a6115069","全模态幻觉的病根在融合层:OWP 免训练重平衡证据与先验","omni-woodpecker-training-free-hallucination-fix","2026-10-26T05:12:35+00:00",{"id":95,"title":96,"news_slug":97,"published_at":98},"9a6f3f05-0bbd-4faf-84f8-c2932ea13801","前沿模型会往画里藏话:78% 的消息能被自己读回来","frontier-llm-image-steganography-fable-51","2026-10-24T03:13:22+00:00"]