Meta AI rolls out several enhancements across apps and websites with its newest Llama 3

By Marty Swant  •  April 19, 2024  •  4 min read  •

Ivy Liu

The ways Meta AI is integrated varies for each app. On apps like Messenger and WhatsApp, users can ask the chatbot questions and receive answers within a conversation or chat with it directly. On Facebook, it’ll show up within a user’s feed and provide info about videos, photos and posts. For starters, an English version of Meta AI is rolling out in the U.S. and at least a dozen other countries including Australia, Canada, Jamaica, Nigeria, Singapore and South Africa. 

In a brief video posted on Facebook, Meta CEO Mark Zuckerberg said “our goal is to build the world’s leading AI and make it available to everyone.”

“We’re investing massively to develop the leading AI,” Zuckerberg said. “And open sourcing our models responsibly is an important part of our approach. The tech industry has shown over and over that open source leads to better, safer, and more secure products, faster innovation and a healthier market.”

Beyond updates for Meta AI, the company also introduced Llama 3, its next-generation large language model. Just like its predecessor, Llama 2, the new AI model will be available open-source on Hugging Face and more than a dozen platforms. Some of those include cloud giants — including Amazon AWS, Google Cloud and Microsoft Azure — and other major AI players like IBM WatsonX, Databricks and Snowflake. Llama 3 will also be offered through AI hardware providers like Nvidia, Intel, AMD and Qualcomm. Meta’s new LLM also marks the latest effort by companies to collaborate with open-source models while also partnering to address ways to responsibly develop ethical AI. 

“The addition of Llama 3 builds on IBM’s collaboration with Meta to advance open innovation for AI,” IBM said in a press release about Llama 3. “The two companies launched the AI Alliance — a group of leading organizations across industry, startup, academia, research and government —  late last year, and it has since grown to more than 80 members and collaborators.”

While Meta released two size models of Llama 3 today, it also plans to release more versions in the coming months that will also have capabilities like multiple languages and larger context windows. The company also plans to add in more multimodal capabilities to power features on Ray-Ban glasses. Although the company didn’t provide details, one possibility is enabling Meta AI to explain what someone is seeing through their smart glasses as they look around.

In Meta’s blog post published yesterday, the company also explained that its training data used for Llama 3 was seven times larger than the previous model. It also claimed Llama 3 beat competing AI models when compared across several benchmarks. However, rivals-turned-partners like Google also celebrated Llama 3 and its addition to the slate of open-source models powering AI platforms like Vertex AI.

Llama 3 arrives at a time when some marketers say there is a growing sense of “AI fatigue” as AI providers race to compete with LLMs and countless other companies compete to market various AI applications. There also is uncertainty when it comes to future regulations. In the U.S., Congress is considering multiple AI-related bills as state lawmakers also look at how to regulate various uses of AI.

It’s too soon to tell how Llama 3 could provide new opportunities and challenges for companies and users alike. However, it’s likely to still raise similar concerns as other AI models when it comes to copyright infringement, data privacy and misinformation. Officials in Europe and the U.S. are also looking at ways to regulate Meta’s practices around user data and use for both advertising and AI.

It’s also unclear how usage of Meta AI might impact the ways users consume information about news and other topics, especially as Meta moves away from partnerships with publishers and removes tools like the news tab in the U.S. and Australia.

There’s also growing concern from consumers about how AI-generated misinformation will be both created and distributed by various social platforms. A report released yesterday by Adobe said adults in the U.S. and Europe are increasingly worried about how AI will impact 2024 elections around the world. According to the survey, 39% of U.S. respondents said they had either stopped using social platforms or cut back on their habits because of misinformation or amount of information they were seeing. Meanwhile, 29% in the UK, 37% in France and 24% in Germany also said they’re cutting back.

“As the results of this study clearly show, it is critical that we educate consumers about the dangers of deepfakes and provide them with tools to understand what is true,” Dana Rao, Adobe’s evp, General Counsel and Chief Trust Officer, said in a statement about the findings. “With elections coming, now is the time to adopt protective technologies like Content Credentials to help restore trust in the digital content we are consuming.”

https://digiday.com/?p=541725

More in Media