Artificial intelligence software is all the rage these days, ranging from applications relating to art, writing, video, and much more.
This has drawn both interest and criticism to the concept as it has hit the mainstream. One type of popular AI is chat related software. Notable examples include ChatGPT and Microsoft’s Bing Chat which allow users to talk with a bot that learns from its conversations.
Meta has also thrown their hat into the ring with their own implementation called LLaMA, which they publicly released in the name of their commitment of “open science”. This meant they released it under a noncommercial license which would be given to academic researchers “on a case-by-case basis”.
This semipublic release didn’t last long, as just a couple weeks after their announcement it’s being reported (via Yahoo News) that it leaked publicly via popular image board site, 4chan.
The leak allows for people to see what was under the hood of Meta’s technology, where some have pointed out it’s smaller than GPT-3, while maintaining better performance. Others have also noted that it hasn’t been trained for chat, meaning it also is inferior to other options for now.
As it’s still the early stages of this leak, not much has been done with LLaMA as of this writing. However, as shown with the numerous other AI technologies that have popped up in the past year, it’ll likely see many breakthroughs.