Meta, formerly known as Facebook, has announced a series of updates to its generative AI capabilities across its platforms. The company introduced Llama 3, a new version of its open-source large language model (LLM), and additional integrations and updates for Meta AI. These updates will be noticeable on Facebook, Instagram, Messenger, and WhatsApp.
The generative AI chatbot, previously powered by Llama 2, will now utilize Llama 3. This upgrade aims to improve user experiences by enabling Meta AI to better assist with tasks such as learning, content creation, and connecting users with their interests or areas of concern.
Meta AI has also expanded its reach to 12 new countries, including Australia, Canada, Ghana, Jamaica, Malawi, New Zealand, Nigeria, Pakistan, Singapore, South Africa, Uganda, Zambia, and Zimbabwe. However, the service is currently only available in English.
One of the significant enhancements is the ability for users to invoke Meta AI through the search bar on Facebook, Instagram, Messenger, and WhatsApp. This integration allows users to interact directly with Meta AI on posts in their Facebook feed. Additionally, a standalone site for Meta.ai has been launched, enabling users to converse with the model in a manner similar to using OpenAI ChatGPT or Google Gemini.
The most noticeable updates include the Meta AI search bar and its integration into Facebook feeds. With these additions, users can ask Meta AI specific questions and receive relevant information in the form of Reels or Facebook pages. Furthermore, through partnerships with Google and Bing, Meta AI can provide real-time information and summaries of search engine results.
However, concerns have been raised regarding the potential misuse of Meta AI. Given the history of social media platforms spreading inflammatory political and medical misinformation, it remains to be seen how Meta AI will handle controversial topics. In response to these concerns, Meta has assured users that Llama 3 has been stress-tested and refined to restrict discussions on risk areas, such as weapons, cyber attacks, and child exploitation. The company aims to prevent prompts that could incite harm while allowing prompts that seek harmless information.
It is important to note that generative AI systems are not infallible and may produce inaccurate or inappropriate outputs. Meta acknowledges this limitation and emphasizes the need for users to exercise caution when relying on the information provided by AI systems. Additionally, the training data used by Meta AI raises privacy concerns. Meta has stated that the models are trained on a combination of publicly available online information and annotated data. It is unclear which specific data sources are used, but Meta claims to exclude or remove sources containing a high volume of personal information about private individuals.
Regarding user data, a Meta spokesperson explained that interactions with Meta AI may be used to train the model. While users have the ability to opt out of sharing their data with a paid version of ChatGPT, it is unclear if this option exists for Meta AI. As a free service, it is likely that users’ data is utilized as part of the agreement for using Meta AI.
In addition to the generative AI updates, Meta has also introduced a new image generator feature. Unlike traditional AI image generators that create an image based on a final prompt, this feature generates images in real time as the user continues typing. Meta aims to make the process of generating images more approachable and user-friendly.
Overall, these updates from Meta demonstrate the company’s ongoing commitment to enhancing its AI capabilities across its platforms. While the improvements offer convenience and assistance to users, concerns surrounding misinformation, privacy, and data usage persist. As users continue to engage with Meta AI, it is essential for both the company and users to remain vigilant in ensuring the responsible and ethical use of AI technologies.
