Google’s Gemini Nano, the smallest and most efficient AI model in its generative series, is not compatible with Pixel 8 smartphones due to some hardware limitations. This means that Pixel 8 users will miss out on some of the AI features that are available on Pixel 8 Pro and Samsung Galaxy S24 devices.
What is Gemini Nano and what can it do?
Gemini Nano is a part of Google’s Gemini project, which aims to create large language models (LLMs) that can generate natural and diverse text for various tasks and domains. Gemini Nano is the smallest version of these models, designed to run on high-end Android smartphones using the AICore service, an API that allows apps to access the model on the device.
Gemini Nano can perform tasks such as summarizing audio recordings as text and generating smart replies to text messages, without requiring an internet connection or sending data to the cloud. This makes it faster, more private, and more cost-effective than using cloud-based LLMs.
Why is Pixel 8 not supported?
According to Terence Zhang, an engineer working on generative AI on Google Android, the reason why Pixel 8 cannot run Gemini Nano is due to “some hardware limitations”. He did not specify what those limitations are, but one possible factor could be the amount of memory available on the device.
Pixel 8 has 8GB of RAM, while Pixel 8 Pro has 12GB. LLMs are typically memory and compute intensive, and Gemini Nano may require more than 8GB of RAM to run smoothly. Another factor could be the availability of hardware acceleration, such as GPUs or TPUs, that can speed up the inference process.
What are the implications for Pixel 8 users?
Pixel 8 users will not be able to enjoy the benefits of Gemini Nano, such as summarizing audio files in the Recorder app, or generating quick replies in the Gboard app. They will also not be able to use any third-party apps that rely on AICore to access Gemini Nano. This could make Pixel 8 less attractive for users who value AI features and functionality.
However, Pixel 8 users may not be missing out on much, as some of the tasks that Gemini Nano can do are already possible with other methods, such as using cloud-based LLMs or speech recognition services. Moreover, some users may prefer not to use Gemini Nano, as it may raise ethical and social concerns, such as the potential for bias, misinformation, or manipulation.
What are the future plans for Gemini?
Google is not giving up on Gemini, and plans to roll out more AI features for its voice-controlled Assistant, powered by its larger Gemini Pro and Ultra models, next year on its Pixel devices. These models are expected to provide more advanced and diverse capabilities, such as generating images, music, or code, based on natural language input.
Google also claims that Gemini is beating its competitors, such as OpenAI and Facebook, in terms of quality, diversity, and efficiency of its LLMs. Google says that Gemini can generate text that is more natural, coherent, and relevant than other LLMs, and that it can do so with less data and compute resources.
