谷歌提前发布人工智能模型 Gemini

软餐获悉，和传闻的推迟不同，谷歌提前发布了其最新的人工智能模型 Gemini。谷歌声称，在大多数大型语言模型基准测试中，Gemini Ultra 的性能优于 OpenAI 的 GPT-4（也是微软 Windows Copilot 的基础），并且 “其功能几乎在每个领域都是最先进的”。Gemini 可理解文本、图像、音频和视频源，也能创建和理解 Python、Java、C++ 和 Go 等代码。

Gemini 分为三个版本：

• Gemini Ultra“用于高度复杂的任务”；
• Gemini Pro 将提供 “跨广泛任务” 的性能；
• Gemini Nano 是 “为了设备上的任务” 而设计的。

Google 的 Bard 聊天机器人已通过 Gemini Pro 进行了更新，这将允许 Bard 用英语提供 “更高级的推理、规划、理解等”。未来，Google 将推出 Bard Advanced，使用 Gemini Ultra 型号提供更高级的功能。

谷歌 Pixel 8 Pro 的用户今天将获得功能更新，该设备将获得 Gemini Nano。谷歌表示，它将允许 Pixel 8 Pro 在其 Recorder 应用程序上总结 “录制的对话、采访、演示等”，并在 Gboard 中提供对 WhatsApp 等应用消息提供回复建议。

Contrary to rumors of delay, Google has released its latest artificial intelligence model Gemini ahead of schedule. Google claims that in most large-scale language model benchmarks, the performance of Gemini Ultra surpasses that of OpenAI’s GPT-4 (which is also the foundation for Microsoft’s Windows Copilot), and it is “state-of-the-art in almost every field.” Gemini can understand text, images, audio, and video sources, as well as create and comprehend code in Python, Java, C++, Go, etc.

Gemini is divided into three versions:

• Gemini Ultra for “highly complex tasks”;
• Gemini Pro will provide performance across a wide range of tasks;
• Gemini Nano is designed for tasks on devices.

Google’s chatbot Bard has been updated with Gemini Pro, which will allow Bard to provide “advanced reasoning, planning, understanding” in English. In the future, Google will introduce Bard Advanced with advanced features using the Gemini Ultra model.

Users of Google Pixel 8 Pro will receive feature updates today as their device will be equipped with Gemini Nano. According to Google, this will enable Pixel 8 Pro to summarize “recorded conversations interviews demonstrations,” etc., in its Recorder application and provide reply suggestions for messaging apps like WhatsApp on Gboard.