Meta rolls out an updated AI assistant, built with Llama 3


Meta has just unveiled a significant update to its AI assistant platform, Meta AI, leveraging the highly anticipated open-source Llama 3 large language model (LLM). According to the company, it’s now positioned as “the most intelligent AI assistant available for free.” In terms of functionality, Meta AI boasts capabilities such as assisting users in studying for exams, planning meals, and organizing social outings – the usual AI chatbot repertoire.

Expanding beyond its initial trial run on Instagram DMs, Meta AI has infiltrated nearly every aspect of Meta’s extensive portfolio. While it remains accessible on Instagram, users can now interact with it across Messenger, Facebook feeds, and WhatsApp. Additionally, a dedicated web portal,, has been established, enabling users to engage with the chatbot without requiring a company login, albeit without image generation capabilities. Integration with the recently launched Ray-Ban smart glasses is already in place, with integration for Quest headsets on the horizon.

On the front of image generation, Meta claims significant improvements in speed and the ability to generate images in real-time as users type. Moreover, the chatbot now supports custom animated GIFs, a noteworthy enhancement. However, concerns arise regarding its proficiency in generating images depicting diverse races, as previous tests revealed a tendency to favor images of individuals from the same racial background.

Meta is also expanding the global reach of Meta AI with this update, extending its availability to over a dozen countries outside the US, including Australia, Canada, Ghana, Jamaica, Pakistan, and Uganda. Notably, this expansion is limited to English language support, potentially limiting its utility for non-English-speaking users.

Addressing safety and reliability, Meta highlights enhancements in the Llama 3 model, emphasizing an expanded dataset and the exclusion of personal data sources. The company conducted thorough evaluations to mitigate risks associated with sensitive topics such as weapons, cyberattacks, and child exploitation.

While AI development has become a focal point for Meta CEO Mark Zuckerberg, alongside his ventures in cattle farming, the company still trails behind industry leaders like OpenAI and Google. Meta’s previous Llama 2 iteration fell short of expectations due to its limited feature set, but the revamped AI assistant holds promise. Whether it will truly captivate users remains to be seen, but it’s poised to offer innovative functionalities and potentially raise the bar in AI assistance.