Google’s Nano AI Model Excluded from Pixel 8 Handsets Due to Hardware Limitations
Google’s latest AI model, Nano, will not be available on the Pixel 8 handsets due to some hardware limitations. Nano, the smallest AI model in Google’s generative Gemini series, was announced in December and is designed to run on high-end Android smartphones such as the Pixel 8 Pro and Samsung’s Galaxy S24 line.
Nano offers various AI features, including the ability to summarize audio recordings as text and Magic Compose, which generates quick replies to text messages. However, these capabilities will not be available on the Pixel 8, despite being released in October. Terence Zhang, an engineer working on generative AI at Google Android, revealed during a panel discussion on The Android Show that this limitation is due to hardware constraints.
It is unclear what exactly these limitations are since both the Pixel 8 and Pixel 8 Pro devices are equipped with the same Google Tensor G3 processor and CPU and GPU cores. The key difference lies in the amount of memory, with the Pixel 8 Pro having 12GB of RAM compared to the Pixel 8’s 8GB. This disparity in memory is likely to affect the performance of the Nano model, resulting in its exclusion from the Pixel 8.
Google has chosen not to comment on this matter officially. Large language models like Nano typically require significant memory and computing power, making it challenging to run them on smartphones, even those with 8GB of DRAM.
To address this issue, Google developed AICore, an Android system service that enables applications to utilize the Google Nano AI model for machine learning inference tasks directly on the device. AICore leverages any available hardware acceleration to enhance performance. However, since there is no Nano model on devices like the Pixel 8, there is no AICore available for use.
AICore offers several advantages, including the ability for apps to access Gemini Nano on the local device without an internet connection. This ensures that all generated text remains private and reduces latency, resulting in faster performance. Additionally, AICore is free to use since it utilizes the device’s hardware, unlike accessing a large language model from the cloud.
It is important to note that AICore is currently only available on Google Pixel 8 Pro and Samsung S24 Series devices, as stated in the developer documentation.
Looking ahead, Google plans to introduce more AI features for its voice-controlled Assistant, powered by the larger Gemini Pro and Ultra models, on its Pixel devices next year.
In conclusion, while Google’s Nano AI model offers advanced AI capabilities, it will not be available on the Pixel 8 handsets due to hardware limitations. Despite both the Pixel 8 and Pixel 8 Pro having similar processors, the difference in memory capacity affects the performance of the Nano model. However, Google continues to innovate in the AI space and plans to introduce more features in the future.
