Following a post by a 4chan user, the Meta AI leaked online and the incident has taken over the news. LLaMA is now accessible via the internet and the company has not yet responded.
Almost a week ago, Meta began responding to access requests for its new language model, which is now widely available online via a downloaded torrent link. After one of the users submitted the link and started the spread, it initially surfaced on 4chan. The AI community is already debating it, and some are even experimenting with the new cutting-edge technology, which is now available via the internet.
Finally finished the LLaMA torrent and managed to load 7B in about 22GB VRAM. Running some fun prompts to test it out. Really not sure about that first answer here 🧐 pic.twitter.com/MOI93bNCBS
— David Rose 🫧 (@drose101) March 4, 2023
How was Meta AI leaked?
Instead of releasing LLaMA as a public chatbot, Meta chose to create a separate product for widespread use. On the other hand, LLaMA was created in response to demands from the AI community to support researchers. The business claimed it wanted to “better democratize access” to AI while promoting study into its drawbacks.
Yet, it has turned into a significant problem for everyone who uses the internet because this potent AI language model could be employed for harmful purposes like customized spam and phishing attempts, and as the Meta AI leaked, this potential became more threatening for online users. Benchmark results indicate that LLaMA performs significantly better than competing open-source models. According to an Ars Technica study, LLaMA-13B fared better than ChatGPT-like technology while being 10 times smaller.
Industry professionals condemned Meta for making the language paradigm open-source, including Jeffrey Ladish, who tweeted: “Well, Meta’s 65 billion parameter language model just got leaked to the public internet, that was fast. Get ready for loads of personalized spam and phishing attempts. Open sourcing these models was a terrible idea.”
“To maintain integrity and prevent misuse, we are releasing our model under a noncommercial license focused on research use cases. Access to the model will be granted on a case-by-case basis to academic researchers; those affiliated with organizations in government, civil society, and academia; and industry research laboratories around the world,” Meta said last week. However, the gates of Meta didn’t hold for long.
How will Meta AI join the artificial intelligence race?
LaMDA from Google Bard is said to compete with LLaMA. A wave of malevolent use is forthcoming, said Arvind Narayanan and Sayash Kapoor of Princeton University in a blog post. But, there haven’t been any reported instances of such usage. Many are blaming Meta for allowing that to happen and for not taking sufficient measures, as the widespread distribution of it may result in a significant surge in cybercrime.
Compared to other publicly accessible chatbots like OpenAI’s ChatGPT and Microsoft’s Bing Chat, the robust language model is very different. It is more complex than free chatbots and calls for knowledge, the appropriate hardware, and time. At that time, the issue arises because a large number of persons are prepared to commit crimes and meet all three criteria, making it more dangerous since the Meta AI leaked online. The Verge also noted that, unlike the other chatbots in the preceding example, LLaMA has not been fine-tuned. The chatbot may concentrate all of its resources on a single task through fine-tuning.
Because the corporation had a different objective in mind when developing the language model, Meta didn’t focus on user experience or fine-tuning. The language paradigm might be used for a variety of nefarious activities, although the corporation hasn’t yet made any announcements about it. Although it is a 219GB file and a big model cannot be operated on a typical PC, the download file also contains the smallest model with 7B parameters. We advise Meta users to try to exercise a little more caution, especially when it comes to phishing scams.
Meta AI leaked only a few days ago but the company continues to work on Meta AI despite the leakage. However, the company is expected to make an announcement about the issue as many users are concerned and frustrated about the leak. If you want to keep up with Meta and its latest developments, make sure to check out the links below to stay on track.