In early March, Google made a perplexing announcement, stating that only the Pixel 8 Pro phone would be able to run its latest artificial intelligence model, "GoogleGemini," while the Pixel 8 was excluded due to alleged "hardware limitations."
Despite the similar configurations of the two phones, Google claimed that the hardware of the Pixel 8 couldn't support the operation of this model, sparking widespread skepticism.
A few weeks later, Google seems to have changed its stance, stating that it will provide the "Gemini Nano" feature for the Pixel 8 in the June quarter update, but with limitations: the Pixel 8 Pro can use it directly, while the Pixel 8 will require enabling it through hidden developer options, making it difficult for most users to experience this feature easily.
SeangChau, Vice President of Google's Devices and Services Software, explained on the "MadebyGoogle" podcast, "The Pixel 8 Pro has 12GB of memory, which is well-suited for running 'Gemini Nano,' but the Pixel 8 only has 8GB of memory, which may affect the user experience."
He further elaborated on the impact of running large language models on phones, stating that Google wants some AI models to be "resident in memory" for instant access.
Chau pointed out that "Smart Reply" requires models to be resident in memory to ensure they are always available, leading to the Gemini-driven "Smart Reply" feature being hidden in developer options on both the Pixel 8 and Pixel 8 Pro. He also noted that Gemini Nano may occupy a significant amount of system memory, making it difficult to release easily.
However, it is worth noting that even the base version of the Samsung Galaxy S24 series has only 8GB of memory but can still run the Gemini Nano model. Google's initial claim of "hardware limitations" for the Pixel 8 seems to be unfounded.
编辑点评：
现在大型AI模型越来越受欢迎，但作为普通用户，我们是否真的需要这样的功能呢？虽然ChatGPT有其用处，但实际可用的操作系统级生成式人工智能功能却寥寥可数。
自动生成回复功能并非对大多数用户必需，如果消耗大量内存、影响流畅度和体验的话，用户可能更希望将其关闭。
