Date: 23/02/2025

Run Intel AI assistants locally on AI PCs instead on on cloud-based instances

At CES Las Vegas in January, Intel unveiled Intel AI Assistant Builder (code-named Project SuperBuilder) to enable computer makers and software vendors to create tailored AI assistants for any use case in minutes.

AI Assistant Builder takes the nitty gritty guesswork out of building these models. It breaks down the tasks to just three easy steps: choose a ready-made AI assistant model, download and install it, and then launch the program.

“We’re providing the industry the building blocks to create their own AI-centric content. That enables our customers to reduce their overall development time and speed up the rollout of their own smart solutions,” explained Olena Zhu, senior principal engineer in Intel’s Client Computing Group.

This is how it works:

The app features a chat window where users can ask questions and upload or delete documents across numerous formats (including PDFs, spreadsheets, presentations and text files).
Each time the user asks a question, the AI assistant responds based on the query. It learns from the conversation history, related documents and the inherent knowledge of the large language model (LLM).
It also includes advanced AI safety guardrails that include profanity filters and a secondary safety system that guides the model to respond safely without bias.
It has ever-growing rich feature sets to allow users to pick and choose different advanced features and custom-build their own AI assistant.

Intel’s key partners are already using AI Assistant Builder to craft customized models for targeted needs:

Acer built an AI sales assistant proof of concept (POC) integrating its camera vision technology in just three days.
In South Korea, Intel and Samsung Electronics introduced a generative AI chatbot at Kyobo Bookstore that runs on-device – not the cloud. It’s based on smart LLMs and enables customers to search for books and receive personalized information.
Asus leveraged Intel’s technology to create a technical-focused Q&A assistant for its NUC customers.

AI Assistant Builder will run on any Intel-powered AI PC that meets specific hardware requirements: At a minimum, it requires a PC running on an Intel® Core™ Ultra Processor (Series 1) with 16 gigabytes of RAM and Integrated Intel® Graphics. (In September, Intel introduced Intel® Core™ Ultra Processors (Series 2) with a 3x increase in AI performance at up to 50% lower power than its predecessor.)

Zhu points out another key benefit of the software: It runs entirely on the local machine, with no cloud connectivity required to process queries (an Internet connection is needed initially to download the toolkit). Running AI assistants locally on the PC also assuages security concerns.

“By running AI assistants locally on their AI PCs versus depending on cloud-based instances, users benefit from peace of mind and are in complete control over the security of their personal data,” explained Zhu.

Source: Intel

Tweet Follow @ecewire