Llama 3, OpenAI, AI News 4/22/24 with Ray & Ronald

Llama 3, OpenAI, AI News 4/22/24 with Ray & Ronald

Published
April 15, 2024
🎙️Show link:
Video preview

Who are we?

Ray

Ray intro worked at Apple for 12+ years and went from working at the Genius Bar to being in the studio with Jony Ives and being in the office with Tim Cook before presenting the Apple Watch.

Ronald Mannak

Rapid Fire

AI headlines that caught our eye. Early trends. Prototypes. Breaking News.

Ronald

Ray

  • AI Pin Mood: Throw it out the window. https://twitter.com/RayFernando1337/status/1781736989085327742
    • Show-stopping bugs
      • AI Pin gets into a state where it stops taking requests despite being in an area with known WiFi. This is the second day it has happened and the steps to get into state aren’t clear.
      • Frequent overheating despite not making many requests
    • Series of live streams from unboxing to using apps to getting stuck in a state of not being able to use the device.
  • Meta Horizon OS https://x.com/Andyparackal/status/1782448650326868240
    • Meta has identified three partners for this initiative: Asus's Republic of Gamers, Lenovo, and Xbox.
    • Meta Horizon Store and App Lab: The Meta Quest Store will be rebranded as the Meta Horizon Store, and App Lab titles will be featured in a dedicated section of the store. 
    • Social Layer and Mobile App: All headsets running Meta Horizon OS will share the same social layer, allowing users to maintain their accounts, avatars, and friends list across different devices. The Meta Quest mobile companion app will be renamed to the Meta Horizon app.
    • New Spatial App Framework: Meta is developing a new spatial app framework to help mobile developers create mixed reality experiences. 
    • Strategic Implications: This move by Meta is seen as an emulation of the Android model, where Google's operating system is used by various smartphone manufacturers.
  • Llama-3 Surpasses GPT-4 as King of Open Models https://chat.lmsys.org/?leaderboard https://twitter.com/lmsysorg/status/1782483699449332144
    • English Rankings: 70B model reaching the top-5 on the Arena leaderboard
      • Arena Leaderboard is a benchmark platform that evaluates large language models (LLMs) through a crowdsourced, competitive format using the Elo rating system.
      • Elo: calculates the relative skill levels of these models in a competitive setting, often by comparing their performance in generating language-based responses.

Trend for the week -

     

    Show Summary

    Introduction

      Closing Remarks