Large Language Model
Event Recap: First-Ever Meta Llama 3 Hackathon
May 20, 2024

This Mother’s Day weekend, we teamed up with Cerebral Valley to host the first-ever Meta Llama 3 hackathon along with 10 other sponsors. At the event, which took place at SHACK15 in San Francisco’s iconic Ferry Building, attendees were encouraged to leverage the full collection of Llama models including Meta Llama 3 and Meta Llama Guard 2 to build open source tooling projects. Meta VP of AI Manohar Paluri kicked off the event with a warm welcome to the developers, sponsors, and influencers who came together to build innovative projects using Llama 3.

By the numbers

  • 1,200+ applications
  • 354 attendees
  • 51 projects built in 24 hours
  • 50 boxes of coffee + 500 cans of energy drinks consumed

Meet the winning projects

OpenGlass: The first-place team built smart glasses on a budget of $20 USD that can provide information about what the wearer is seeing. For instance, when looking at a fruit bowl, the glasses were able to identify the fruits and provide their calorie count. Check out the GitHub.

Deb8: The second-place team built an AI agent debate arena where different models compete in a Lincoln-Douglas debate. A third agent rates the quality of the debaters and scores them.

Team AAA (Activation Ablation Augmentation): The third-place team discovered jailbreaking methods to neutralize the activation layers in Llama 3, which can prevent the model from answering prompts deemed problematic.

Other notable projects included:

  • Feycher: An app that lets people create their own lifelike AI chatbots with customizable appearances with playful animation, personalities, and skills
  • Hound:A project that uses the power of generative AI and retrieval-augmented generation (RAG) to help law enforcement crack down on established human trafficking routes globally
  • Joe CRM: A Llama 3-powered CRM that gathers customer-related signals, conducts research, and autonomously creates outreach campaigns
  • Mongoose Miner: A search system to push LLM-based code generation beyond average human-level performance
  • SixthSense: An app for Ray-Ban Meta smart glasses that augments physical knowledge consumption by letting you use physical highlighters to digitally bookmark memorable pieces of content, get explanations on the fly, and more

Click here for a full gallery of all 51 projects.

Llama in the wild

Meta has long been committed to the development of an open AI ecosystem, and this first Llama 3 Hackathon is a testament to what the community can do (and quickly) when they have access to high-quality open models. We heard from multiple developer attendees that the introduction of Llama with a commercial license was a game-changer for the AI community. We can’t wait to see what else the community comes up with.

Thank you to all of the participants who helped make this hackathon a success. Onward to the next one!

Visit the Llama 3 website to download our latest models and check out the Getting Started Guide for the latest list of all available platforms.


Share:

Our latest updates delivered to your inbox

Subscribe to our newsletter to keep up with Meta AI news, events, research breakthroughs, and more.

Join us in the pursuit of what’s possible with AI.

Related Posts
Computer Vision
Introducing Segment Anything: Working toward the first foundation model for image segmentation
April 5, 2023
FEATURED
Research
MultiRay: Optimizing efficiency for large-scale AI models
November 18, 2022
FEATURED
ML Applications
MuAViC: The first audio-video speech translation benchmark
March 8, 2023