Intel AI PC Pilot Hackathon ‘24: Intel Student Ambassadors Created High-Performance AI Solutions
Last December, the Intel Core Ultra Processors (Series 1), also known as Meteor Lake, were made available to the public. The Software Ecosystem Enablement (SEE) division of Intel, in partnership with the Client Computing division (CCG), held a developer-focused AI PC pilot hackathon with the goal of gathering early feedback and comprehending the extent of innovations on the Meteor Lake AI PC platform. During the “Rockstar Student Ambassador hackathon,”12 of the most active Intel Student Ambassador Program participants in the United States, gave them Intel AI PC development kits, and trained them for nine weeks on how to create creative solutions that capitalize on AI PCs.
Intel AI PC Pilot Hackathon Winning Projects
The top five initiatives created by the student ambassadors during the event are listed below:
Virtual Reality (VR) AI Playground
Migara Amarasinghe of Fiorida State University created the Virtual Reality (VR) AI Playground, an interactive learning environment for teachers, students, and AI enthusiasts. Using virtual reality and real-time AI applications, it produces an immersive learning environment that covers fundamental to advanced AI ideas. With interactive courses and live demos, the goal is to make learning AI more interesting. The primary workstation was the Intel NUC (Next Unit of Computing).
Three programs make up the project, which is based on the C# and Python programming languages: (i) YOLOv10, a VR program for object recognition that uses the OpenVINO Toolkit to locally perform model training and inference on GPU; (ii) Hugging Face APIs for VR integration for interactive text-to-image conversion; and (iii) the main VR AI playground for speech-to-image interaction.
TikTalk:
Bill Zhang of the University of Southern California created this project, which uses user input to create brief instructive movies. The researchers used the OpenVINO Toolkit to optimize the Meta Llama 3.2 model for Intel CPU, which produces music descriptions and video scripts. The music description is used to create the audio, which is transcribed to obtain subtitles. Subtitles are used to generate pertinent images, which are then put together to create a finished video output in the form of an MP4 file.
Maestra
Keerthi Nalabotu of the University of California, San Diego created the Python-based voice-to-voice software Maestra. The user provides the tool with a topic or a document, notes, or other type of file as input. Based on the contents of the file, the tool answers with a preliminary query. After then, the user can record his own answer to the question, starting a dialogue that will be documented in a transcript for later use. In order to achieve faster inference, the researchers trained the models on a GPU rather than a CPU using the OpenVINO version of the Microsoft Phi-3 model.
NUC AI PC Drone
Yuri Winche Achermann of RWTH Aachen University created this AI-powered drone project. Using the Intel Core Ultra 7 processor-powered NUC AI PC as the main hub for controlling hardware resources and software programs, it mimics the brain of an autonomous drone.
Voice-to-text and text-to-speech conversions are handled by the Audio Processing module; question-answering, order placement, and customization are handled by the Natural Language Processing (NLP) module; and object detection, depth estimation, and facial recognition are carried out by the Computer Vision module of the project. You may find the drone’s architecture details on GitHub.
YouTube Copilot
Developed by University of Michigan student Kieran Llarena, the project serves as a virtual assistant that can respond to your inquiries simply using the URL of a YouTube video. Hugging Face LLM on AI PC is operated by the Python-based project, which also makes use of OpenVINO for quicker inference. A Google Chrome plugin uses JavaScript to insert a graphical user interface (GUI) into a YouTube video URL. After that, the URL is processed to save subtitles and retrieve the video ID. The local LLM on the AI PC provides a response to the user’s search question and sends it back to the Chrome extension, which parses the data and shows the result on the injected GUI.
What Comes Next?
Use AI PCs with Intel Core Ultra CPUs to combine the capabilities of CPU, GPU, and NPU for faster AI development. Use AI PC development tools, such as the AI PC development kit, to get started.
Intel also invite you to investigate AI tools and framework optimizations for multiarchitecture, cross-vendor, parallel computing, which are driven by the oneAPI programming model.
