谷歌怒发十几款AI新品,但最出圈的专题还是「翻车」
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">一年一度的Google I/O<span style="color: black;">研发</span>者大会如期而至,在本场发布会中,谷歌一口气交出数个AI新品,轮番轰炸<span style="color: black;">咱们</span>的视觉神经。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">首要</span>是真正<span style="color: black;">道理</span>上的全新<span style="color: black;">制品</span>——Google AI Overviews,这是一项基于大模型技术<span style="color: black;">研发</span>的全新搜索引擎,旨在以聊天的方式为用户<span style="color: black;">供给</span><span style="color: black;">精细</span>、<span style="color: black;">有效</span>的搜索结果。没错,传闻中OpenAI正集全力要打造的<span style="color: black;">便是</span>这玩意儿。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-axegupay5k/b2a2a3b63df748f69f75fe18c75333a8~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=YjNcKvHbONkCL26v8tOhuUgK%2Fzs%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">一样</span>与搜索<span style="color: black;">相关</span>的还有「Ask Photos」,相比起AI Overviews,它更专注于图像理解与图像信息<span style="color: black;">捉捕</span>,这<span style="color: black;">寓意</span>着<span style="color: black;">此刻</span>你<span style="color: black;">能够</span><span style="color: black;">经过</span>文字描述的方式,找到藏在相册里的<span style="color: black;">哪些</span>被遗忘已久的照片。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">还记得前天<span style="color: black;">夜晚</span>OpenAI发布的GPT-4o在视觉与听觉上的巨大<span style="color: black;">提高</span>吗?谷歌<span style="color: black;">亦</span>整了一个<span style="color: black;">类似</span>的AI工具——Project Astra。定位上,Project Astra和GPT-4o都是多模态AI项目,用户<span style="color: black;">能够</span>利用手机摄像头和麦克风与现实世界进行交互,例如<span style="color: black;">帮忙</span><span style="color: black;">眼瞎</span>人士识别路上的风景等。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/83fa52b79b9e408c951c27f848a039ee~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=7PDGCh%2BL3UDu2rYS9Z6oaxg6pI0%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">以上<span style="color: black;">说到</span>的,只是本场I/O<span style="color: black;">研发</span>者大会上谷歌<span style="color: black;">颁布</span>众多AI新技术的冰山一角,还有<span style="color: black;">更加多</span>新玩意值得<span style="color: black;">咱们</span>深入探讨。难怪,在这场发布会后,</span><strong style="color: blue;"><span style="color: black;">不少<span style="color: black;">媒介</span>都<span style="color: black;">暗示</span>认为谷歌像是要用海量新品「围剿」OpenAI,重树谷歌在AI市场的地位。</span></strong></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">对轰GPT,谷歌怒发数款AI新品</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">很显然,Google I/O 2024最大主角<span style="color: black;">便是</span>「AI」,从硬件到软件,从服务到功能,几乎<span style="color: black;">无</span>任何一部分离得了AI这个关键词。据不完全统计,这场发布会上,谷歌<span style="color: black;">最少</span>提了121次「AI」。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">既然聊到AI大模型,那还是先<span style="color: black;">瞧瞧</span>Gemini又有了<span style="color: black;">那些</span>新变化。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">Gemini Pro从前代的100万Tokens升级到了200万Tokens,</span></strong><span style="color: black;">与月之暗面的Kimi Chat长度相近,但这个模式并非面向所有用户开放,需要单独申请。Gemini 1.5推出了Flash版本,支持100万Tokens,主打一个便宜量大,1M Tokens输入仅0.35美元、1M Tokens输出<span style="color: black;">亦</span>仅需0.53美元。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/1f365e994a624fe59b99f55bdd3e83b7~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=94%2FdN5nSOTXpSvLMTe3SsUUG0TA%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">另一</span>,谷歌还宣布Gemini Nano进入手机端,<span style="color: black;">日前</span>它能实现的功能是<span style="color: black;">帮忙</span>用户接打<span style="color: black;">tel</span>、识别<span style="color: black;">tel</span>诈骗和骚扰电话。实话说,Gemini Nano的手机端实用性还是略显寒酸,<span style="color: black;">乃至</span>连文字处理都不支持,还不如小爱<span style="color: black;">朋友</span>来得简单粗暴。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">假如你对Gemini Nano的功能表现不太满意,<span style="color: black;">亦</span>能试试谷歌即将推出的Gemini手机客户端,它与前天发布的GPT-4o<span style="color: black;">同样</span>,都是AI多模态应用,能听、会读,还能<span style="color: black;">供给</span><span style="color: black;">心情</span>价值。只是从演示视频来看,Gemini还不是太「拟人化」。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">Gemma2<span style="color: black;">亦</span>在这场发布会上正式登场,<span style="color: black;">做为</span>谷歌下一代开源模型,它升级到了27B规模,和Meta的Llama 3相近,但体积更小些。值得<span style="color: black;">重视</span>的是,Gemma2能够在NVIDIA的GPU或Vertex AI的单个TPU主机上<span style="color: black;">有效</span>运行。Gemma还迎来了一位新成员:PaliGemma,这是一个开源的图像输入模型。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">除了Gemini和Gemma的升级之外,基于大模型技术,谷歌还推出了三款全新的AI大模型应用:</span><strong style="color: blue;"><span style="color: black;">Imagen 3、Music AI Sandbox、Veo。</span></strong></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/a6ae0996b50f4636918438e20b5fbd6b~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=mm7I3AYoE23C469TSmSwgxcgnSg%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">Imagen 3是谷歌最新的画图模型,你<span style="color: black;">能够</span>理解为谷歌版本的stable diffusion,即文生图模型。<span style="color: black;">根据</span>谷歌的说法,Imagen 3相比起前代,在生成速度、生成质量和理解能力上均有<span style="color: black;">很强</span>的<span style="color: black;">提高</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">Music AI Sandbox是一款音乐创作大模型,和之前火遍全网的suno差不多,而它的<span style="color: black;">优良</span>在于创作完成后能够一键上传至Youtube,这何尝不是一种生态<span style="color: black;">优良</span>呢?至于Veo,这是谷歌首款文生视频模型,对标GPT的Sora,</span><strong style="color: blue;"><span style="color: black;">但它时长支持到最高1分钟、分辨率<span style="color: black;">亦</span>支持到1080P,还支持<span style="color: black;">更加多</span>滤镜和电影风格,无论是哪一方面,看起来都比Sora<span style="color: black;">可靠</span>得多。</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">不难看出,谷歌<span style="color: black;">日前</span>在AI<span style="color: black;">行业</span>的战略还是相对稳健,例如模型性能升级,仅从100万Tokens<span style="color: black;">提高</span>到200万Tokens,和早前<span style="color: black;">公众</span>的预期有些差距;而新的AI大模型应用部分,无论是新升级的文生图模型,还是全新的音乐创作模型、文生视频模型,都是「守擂」型<span style="color: black;">制品</span>,少了些创意和想象力。但谷歌天然的<span style="color: black;">优良</span>在于生态,这才是其与OpenAI叫板的底气。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">AI加入谷歌<span style="color: black;">整家</span>桶</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">做为</span>当前的互联网巨头之一,谷歌的软件生态、服务生态自然是相当全面的,<span style="color: black;">例如</span>谷歌<span style="color: black;">持有</span>当前最大用户数量的浏览器Chrome,<span style="color: black;">亦</span><span style="color: black;">持有</span>最完善的Google办公套件,以及<span style="color: black;">日前</span>最大的移动操作系统生态。现如今,谷歌<span style="color: black;">亦</span>正式将AI引入到「谷歌<span style="color: black;">整家</span>桶」中,彻底梭哈AI。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">首要</span>,</span><strong style="color: blue;"><span style="color: black;">谷歌发布了全新侧栏应用Side Panel,</span></strong><span style="color: black;">这是一项综合了谷歌旗下服务的「侧边栏」,当你在Gmail中收到重要信息时,<span style="color: black;">能够</span>直接在Side Panel中呼出谷歌云盘存储信息,又或是<span style="color: black;">起步</span>Google Map进行导航,还<span style="color: black;">能够</span><span style="color: black;">运用</span>谷歌日历记录日程等。在谷歌的计划中,Gmail<span style="color: black;">火速</span>会<span style="color: black;">持有</span>自动处理重要信息的能力。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">Gmail还加入了智能对话的特性,</span></strong><span style="color: black;">简单<span style="color: black;">来讲</span>,<span style="color: black;">此刻</span>你能够以对话的方式在邮箱中找到所需要的信息及<span style="color: black;">关联</span>邮件,还能让Gemini帮你总结这些邮件说了什么,<span style="color: black;">乃至</span>还<span style="color: black;">能够</span>让它帮你智能回复对方,并保持邮件所需的正式用语和语气。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/be5f15ed4cbe431681d80a0773916773~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=vvazWZG6t9mfaAYrzQ3Kf15NOjk%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">其次,前面<span style="color: black;">说到</span>了Gemini将推出手机客户端,除了能够和用户进行普通的对话、<span style="color: black;">文案</span>总结、文字生成等,</span><strong style="color: blue;"><span style="color: black;">Gemini还<span style="color: black;">持有</span>Gemini Live功能,即<span style="color: black;">经过</span>摄像头和你直接对话。</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">最后,<span style="color: black;">亦</span>是最重磅的——</span><strong style="color: blue;"><span style="color: black;">AI Overviews。</span></strong><span style="color: black;"><span style="color: black;">做为</span>搜索引擎巨头,谷歌要<span style="color: black;">怎样</span>将大模型融入到搜索上,这<span style="color: black;">始终</span>是<span style="color: black;">咱们</span>所好奇的,就在这场发布会上,谷歌<span style="color: black;">最终</span>推出了首款AI搜索大模型AI Overviews。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">与Perplexity或Arc Search相比,AI Overviews的<span style="color: black;">优良</span>在于综合搜索能力有了长足的进步,且得益于Gemini的推理能力<span style="color: black;">提高</span>,其<span style="color: black;">得到</span>的搜索结果<span style="color: black;">亦</span>会更符合用户的<span style="color: black;">需要</span>。AI Overviews还支持Plan Ahead,即「为你计划」,尽管AI搜索只能帮用户做搜索结果的总结,但基于推理能力和决策能力的升级,Plan Ahead能够为用户生成<span style="color: black;">各样</span>计划,例如<span style="color: black;">膳食</span>计划、健身计划、旅行计划等。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/a930006ba4ea463b9ed900974de87ae9~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=WjXm6SLenoqaHwERaX%2FkmXvPz2E%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">AI Overviews不仅支持文字搜索,还能<span style="color: black;">经过</span>语音和<span style="color: black;">照片</span>进行搜索,例如当你遇到不认识的植物,<span style="color: black;">那样</span>仅需拍摄、上传,就能让谷歌帮你找出这种植物的<span style="color: black;">关联</span>资料。而这项功能<span style="color: black;">亦</span>会与Pixel上的「划圈即搜」功能相结合,在<span style="color: black;">将来</span>几个月内上线。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">能够</span>说,AI<span style="color: black;">已然</span><span style="color: black;">作为</span>谷歌几乎所有业务的核心,除了<span style="color: black;">咱们</span><span style="color: black;">熟练</span>的大模型应用之外,还在办公、娱乐<span style="color: black;">行业</span><span style="color: black;">供给</span>基于Gemini大模型的AI功能,并互相之间有所串联,<span style="color: black;">提高</span>综合工作的效率。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">One More Thing:<span style="color: black;">奥秘</span>AI眼镜<span style="color: black;">揭发</span></h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">在演示Project Astra的过程中,除了像OpenAI<span style="color: black;">同样</span><span style="color: black;">运用</span>iPhone进行功能展示,还<span style="color: black;">运用</span>了一款智能眼镜。但与<span style="color: black;">咱们</span>之前见到的Google Project Galass<span style="color: black;">区别</span>,这可能是一款全新的智能眼镜<span style="color: black;">制品</span>。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/cb4b54980ce84eec9f88b5fef5744c3f~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=fe6q%2BNneVmYY3O%2BUneHU%2Fev3pqM%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">初代Google Project Galass诞生于2012年,在那个智能手机还不算普及的年代里,谷歌就将智能穿戴设备打<span style="color: black;">导致</span>平民化的消费级<span style="color: black;">制品</span>。但事实上,Google Project Galass受限于<span style="color: black;">制品</span>形态和<span style="color: black;">制品</span>性能,以及相对<span style="color: black;">昂贵</span>的售价,<span style="color: black;">始终</span>都不受市场的欢迎,而谷歌<span style="color: black;">亦</span>在前几年宣告这个项目被取消。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">Project Astra可能是最适合Google Project Galass的AI形态,一方面,它的交互简单,无需太多传感器进行辅助识别;另一方面,它利用大模型的学习、理解、推理能力,能够简单地<span style="color: black;">帮忙</span>用户判断现实世界的物体、景色以及突发事件。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">当然,谷歌并<span style="color: black;">无</span>真正发布这款<span style="color: black;">制品</span>,但结合Apple Vision Pro重新带动虚拟现实(空间计算)市场的<span style="color: black;">热榜</span>来看,谷歌极有可能抢先于苹果,将AI大模型带入到穿戴设备,以快速抢占AR/VR市场。</span></p>
<h1 style="color: black; text-align: left; margin-bottom: 10px;">眼花缭乱的I/O,谷歌真慌了神?</h1>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">整场发布会下来,谷歌发布了非常多的AI新品,<span style="color: black;">触及</span>到大模型技术的迭代、新AI应用的落地以及开源模型再一次进化。但长达数小时的发布会加体验环节,谷歌缺少了一款真正<span style="color: black;">道理</span>上的「爆品」来<span style="color: black;">导致</span>市场的高度关注。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">例如</span>,谷歌发布了一款对标OpenAI Sora的Veo,无论是输入内容的支持、生成视频时长还是生成视频的清晰度,都<span style="color: black;">能够</span>说是「吊打」Sora的存在,但Sora<span style="color: black;">已然</span>抢先于谷歌,<span style="color: black;">诱发</span>了市场对文生视频应用的讨论,这就<span style="color: black;">引起</span>尽管Veo很牛,可讨论度<span style="color: black;">显著</span>不足。</span></p>
<div style="color: black; text-align: left; margin-bottom: 10px;"><img src="https://p3-sign.toutiaoimg.com/tos-cn-i-6w9my0ksvp/bff104501f47483d9b0f00a04d4631be~noop.image?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1722928959&x-signature=qO1waVKDnRPxjTgS1uEywFr8XB8%3D" style="width: 50%; margin-bottom: 20px;"></div>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(图源:Google)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">又<span style="color: black;">或</span>说,谷歌似乎不太懂<span style="color: black;">怎样</span>抓住<span style="color: black;">大众</span>的眼球,这从其频频翻车的演示<span style="color: black;">亦</span>能看出些端倪。还记得Bard首次<span style="color: black;">显现</span>时回答问题错误的<span style="color: black;">状况</span>吗?是的,在这一次AI Overviews的演示上,又一次给用户错误的<span style="color: black;">意见</span>,<span style="color: black;">导致</span>不小的争议。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">而谷歌在会上强调了Gemini的「优惠价格」,试图想要以低价与头部企业们竞争。但现实<span style="color: black;">情况</span>是,百度文心一言、阿里通义千问早就开放了长文本阅读、月之暗面的Kimi<span style="color: black;">亦</span>加入了200万Tokens的免费大战,就连有些晚来的豆包,<span style="color: black;">亦</span>在今日举行的发布会上公开了超低价的策略。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">创意欠缺、低价拼<span style="color: black;">不外</span>,这<span style="color: black;">便是</span>谷歌在I/O<span style="color: black;">研发</span>者大会上给<span style="color: black;">大众</span>的印象。<span style="color: black;">不外</span>,谷歌最重要的杀手锏仍然是它的AI搜索,而这项功能<span style="color: black;">是不是</span>会让谷歌实现逆风翻盘,或许还要等到AI Overviews正式上线后<span style="color: black;">才可</span>解答。</span></p>
楼主的文章非常有意义,提升了我的知识水平。 楼主的文章深得我心,表示由衷的感谢! 外链发布社区 http://www.fok120.com/
页:
[1]