By Maham Javaid
The Meta AI assistant is coming to WhatsApp, Messenger, and Instagram, along with dozens of AI characters based on celebritiesIn order to offer real-time online results, Meta is announcing a partnership with Microsoft’s Bing. This differentiates Meta AI from many other free AIs out there that don’t have recent information.
Another significant feature of the Meta AI is its capacity to produce visuals using the “/imagine” prompt that resemble Midjourney or OpenAI’s DALL-E. It quickly generated eye-catching high-res photographs. This image generation is entirely free to use, just like all of the other AI tools that Meta has unveiled this week.
The assistant’s development has been overseen by Ahmad Al-Dahle, VP of generative AI at Meta, although he wouldn’t say specifically what it has been trained on. He defined it as a “custom-made” huge language model that is “based on a lot of the core principles behind Llama 2,” Meta’s most recent model that is essentially open source and is currently gaining wide acceptance.
According to him, Llama 2’s quick uptake has enabled Meta to improve the functionality of its own assistant. “We just saw a huge demand for the models, and then we saw an incredible amount of innovation happening on the models that really helped us understand their performance, understand their weaknesses, and help us iterate and leverage some of those components directly into product.”
Al-Dahle claims that his team spent time “refining additional data sets for conversations so that we can create a tone that is conversational and friendly in the way that the assistant responds” when comparing Meta AI to Llama 2. Many of the current AIs can be robotic or boring.
The context window of the model was increased by Meta “so that we can build a deeper, more capable back and forth” with users. The context window allows the model to draw on earlier interactions to generate what it outputs next. He claims that Meta AI has been modified to provide “very concise” responses.
The company is starting to roll out an initial roster of 28 AI characters across its messaging apps in addition to Meta’s helper. Charli D’Amelio, Dwyane Wade, Kendall Jenner, Mr. Beast, Snoop Dogg, and Paris Hilton are just a few of the famous people that have inspired many of them. Others are designed for certain use cases, such as a travel agency.
These characters have an intriguing twist that Al-Dahle refers to as “embodiments.” One of them will have a subtle animation on their profile picture that changes while you interact with them. In comparison to the 2D chatbots I’ve used so far, the impact is more immersive.
Though it appears that this will change in the future, Meta AI is not currently trained on publicly available user data from Facebook and Instagram. Asking it to “show me reels from the south of Italy” is a simple example of a use case that other chatbots may find difficult to mimic. “We see a long roadmap for us to tie in some of our own social integrations as part of the assistant to make it even more useful,” said Al-Dahle.
Al-Dahle and other Meta executives that I spoke with made it very apparent that the firm views its unmatched distribution—billions of daily users across all of its messaging apps—as a crucial competitive advantage versus ChatGPT and rivals.
The assistant can be found “right there inside of your chat context, and our chat applications are quite popular,” according to Al-Dahle. You don’t need to remove yourself from the situation in order to communicate, engage, or ask the assistant for assistance.
Although OpenAI may have launched the chatbot craze, given the enormous reach of Meta’s social networks, its assistant may end up being the AI that most people interact with for the first time.
Produced in assosciation with theverge.