The Future. Meta’s new AI system, which was trained on 40% more data than its last one, is free for all to use, tinker with, and create businesses with. Meta hopes the move will spark usage and bring it to the level of OpenAI’s ChatGPT and Google’s Bard. But by making the tech so public, Meta may make its system vulnerable to bad actors beyond the company’s control.
You get a bot, you get a bot
Meta’s LLaMA 2 is going open source with the help of Microsoft’s Azure cloud service, Amazon Web Service, and Hugging Face.
- By making the system accessible to all (and free of charge), Meta hopes it can catch up with competitors… or at least be used in their ecosystems.
- It also believes going open source will let third-party developers build on and find flaws with the system, which Meta can then leverage or patch over.
- And by putting it under a commercial license, developers can build apps that use Meta’s AI as the engine, becoming the de facto OS for new companies.
Meta’s open-source gambit has been met with mixed reviews. Companies like Nvidia (riding a sky-high stock because of AI), Zoom, and Dropbox have issued statements of support. But many AI researchers have said giving out the tech so freely has dangerous repercussions, like scams and misinformation.
In defense, Meta says it’s done copious “Red Team” testing — that is, “testing software for potential misuse and figuring out ways to protect against such abuse.”
Time will tell if that was enough.