A solution in search of a problem? Intel debuts AI app that runs locally on its most recent CPUs – but will it convince users of the benefits of end-user AI?
Intel has updated its AI Playground app, initially launched in July 2024, to showcase the capabilities of running AI tasks on its latest Intel Core Ultra 200V (Lunar Lake) series processors.
Designed to run AI entirely on compatible systems, the app harnesses the chip maker’s built-in AI accelerators and Arc GPUs.
Available as a free download, AI Playground is optimized for Intel Core Ultra PCs with integrated Intel Arc GPUs or discrete Intel Arc Graphics cards featuring 8GB of memory. The latest version includes installers for the Intel Core Ultra 200V (1.21b) and Ultra-H series (1.01b), as well as Intel Arc dGPUs.
New features
For many Windows 11 users, Microsoft Copilot will be their primary – maybe only – AI experience, but privacy concerns have emerged over its controversial Recall feature and potential data sharing.
AI Playground keeps all processing on the device itself, offering a more private and secure alternative for users wary of cloud-based AI.
The app is designed to help users explore and understand the benefits of AI and Tom’s Hardware likens it to how Microsoft added Solitaire to early versions of Windows to help users get accustomed to the mouse. Whether there’s a need for such an app boils down to whether Intel can demonstrate real-world value beyond just showcasing its technology and the jury is out on that at the moment.
The app’s primary modes – Create, Enhance, and Answer – offer a number of tools. Create mode generates images from text prompts using SD1.5 or SDXL models, while Enhance mode supports upscaling, inpainting, and outpainting for complex image edits. Answer mode functions as a local chatbot that can process documents and answer queries, all while keeping data private by operating entirely on the user’s device.
The new update introduces features tailored for Ultra 200V processors, such as an exclusive theme and an improved LLM Picker preloaded with models like Phi3, Qwen2, and Mistral.
Users can install their own native PyTorch LLMs too, so long as they are compatible with Transformer version 4.39. The new version also includes a Conversation Manager for saving chat sessions, a font size adjustment tool, and a preset aspect ratio picker to streamline image creation.
To find out more about the AI Playground project you can go to its page on GitHub, or join the Intel Insiders Discord.