Samsung Electronics Co. Ltd today announced plans to integrate Google LLC’s Gemini language model series into its upcoming Ballie home robot.

The integration expands an existing artificial intelligence partnership between the companies. Earlier this month, Samsung rolled out a new Gemini integration to its flagship Galaxy S24 smartphone series. The feature allows users to point their device’s camera at an object and have Google’s AI surface relevant information.

Ballie, the Samsung home robot set to receive the Gemini integration, was first previewed in 2020. The company has since updated the machine’s design multiple times. The latest iteration, which is set to launch this summer, is about the size of a basketball and uses wheels to navigate the user’s home.

Consumers can interact with Ballie via voice commands or virtual buttons that the robot displays on the floor using a built-in projector. The buttons can be pressed by stepping on them.

Ballie’s projector doubles as a video streaming tool. At this year’s CES consumer electronics conference, Samsung demonstrated how a user could ask the robot to project a movie on a wall. The built-in AI models can automatically position the robot in a way that maximizes the projection quality.

Ballie lends itself to other tasks as well. It can greet users at the door, set reminders and adjust the settings of smart home appliances. Using Gemini, Ballie can generate wellness advice based on information retrieved through Google Search. 

The Gemini model series includes multiple algorithms with different capabilities and pricing. Samsung didn’t specify which specific algorithm it’s integrating into Ballie. The most likely candidate is the entry-level Gemini 2.0 Flash-Lite, the fastest and most cost-efficient model in the series. The pricier Gemini models offer more advanced reasoning capabilities that aren’t necessarily needed for a home robot.

“With Gemini on Google Cloud, Samsung is demonstrating how to deploy generative AI at scale, integrating it directly into the heart of their popular products,” said Google Cloud Chief Executive Officer Thomas Kurian. 

Samsung detailed that Ballie will also use unnamed “proprietary Samsung language models.” That could be a reference to the internally developed Gauss2 model series the company detailed last December. One of the algorithms in the series, Gauss2 Compact, is specifically optimized to run in “limited computing environments” such as smart home devices.

Samsung’s models are multimodal and understand up to 14 languages. The company has equipped them with a custom tokenizer, the part of a large language that turns input into tokens. Breaking up user instructions into tokens, small units of data that each contain a few numbers or letters, makes the information easier to process.

The company is no stranger to the home robotics market: it’s among the world’s largest suppliers of robot vacuums. It also makes a wide range of other home devices.

Given that Ballie includes an extensive set of AI features, Samsung might power it using a chip from its Exynos series of smartphone processors. The most capable chip in the lineup, the Exynos 2400, features an eight-core graphics processing unit based on Advanced Micro Devices Inc.’s RDNA 3 architecture. Samsung claims the chip can run AI models more than 14 times faster than previous-generation silicon.

Image: Samsung