Meta is set to bring generative AI to every one of its platforms, after reporting sales increases for the first time in three-quarters – thanks in part to the success of its AI-driven Instagram Reels feature.
After famously dragging its heels with AI investments, channeling the majority of its spending into its VR Metaverse project instead, the company now claims it's “no longer behind in building AI infrastructure”.
New AI chat features will soon be added to WhatsApp and Messager, as well as visual content tools for Facebook and Instagram.
But how will Meta's new AI model, LLaMA, compare to established tools like ChatGPT?
Meta Will Bring AI to Every Platform
In a development that can best be described as ‘better late than never' Meta has announced it will be adding generative AI to every one of its apps in a way that's “useful and meaningful”.
In Meta's recent earnings call, the company's CEO Mark Zuckerberg said it's exploring “chat experiences in WhatsApp and Messenger”. While the Silicon Valley company is keeping its cards close to its chest, it's likely these text-based AI tools will bare some resemblance to existing models like Google Bard and ChatGPT.
Meta also revealed they're working on visual AI creation tools for adverts and Facebook and Instagram posts, with video and multi-model experiences likely to be introduced over time.
“Over time, this will extend to our work on the metaverse, too, where people will much more easily be able to create avatars, objects, worlds, and code to tie all of them together.” – Meta CEO Mark Zuckerberg
While this shift to AI spending represents a change of tact for the social media giant, the company is keen to point out that they aren't abandoning the metaverse project, with Zuckerberg explaining that AI developments will “extend to our work in the metaverse too”.
Meta will be forking out $33 billion to fund this rollout, according to the company's Chief Financial Officer Susan Li. But unlike its Metaverse project, which cost its Reality Lab unit almost $4 billion in the first quarter of this year, these heavy investments into AI are likely to pay off.
Meta's Year So Far
Yet, in Meta's latest quarterly earnings report of the year, it revealed its sales have increased for the first time in three quarters, with the company hitting revenues of around $28.65 billion – beating estimates by almost a billion dollars.
This rebound can be largely attributed to Instagram's new Reel feature, which increased time spent on the app by 24% due to using TikTok-adjacent AI algorithms that give users more personalized recommendations.
“Reels also continue to become more social with people resharing Reels more than 2 billion times every day, doubling over the last six months.” – Meta CEO Mark Zuckerberg
Meta's ad revenue on Facebook is looking optimistic too, with the platform attracting a record two billion daily active users in February.
How Does Meta's AI Model Compare?
Last month, Meta officially joined the AI race by announcing LLaMA – a large conversational language model that works in a similar way to ChatGPT, Bing Chat, and Google Bard.
LLaMa and ChatGPT both use unsupervised learning to train their models, which means they use the internet and other sources, instead of human-labeled data to generate data.
Unlike ChatGPT, LLaMA is also trained on diverse sources such as scientific and news articles, making it better equipped at answering more technical prompts.
However, while LLaMA may have the edge when it comes to research, its 65 billion parameters pale in comparison to ChatGPT's 175 billion parameters, making its uses slightly less extensive than OpenAI's chatbot.
As it currently stands, LLaMA is still not accessible to the public, so it's impossible to see how it stands up against its competitors, directly. Yet, with Meta's gaze set on becoming a “leader in generative AI”, the race to become the dominant AI is anything but clear-cut.