CIO Insider

CIOInsider India Magazine

Separator

Meta to Train Artificial Intelligence Models to Stay Ahead of its Competitors

CIO Insider Team | Tuesday, 11 June, 2024
Separator

Aiming to keep up with the pace against competitors like OpenAI and Google, Meta, the social media behemoth, revealed plans on training its artificial intelligence models using user data from European users.

The company stated that it needs to use public data from those users to train its Llama AI large language model to better reflect the languages, geography, and cultural references of its users in Europe.

Despite initial concern regarding EU’s privacy and transparency requirements, the move seems to put the company's approach in Europe broadly in line with how it approaches the data it feeds into its AI models from other parts of the world.

Large data sets are used to train AI language models, which enable them to anticipate the most likely word to appear in a phrase. Newer iterations of these models are usually more intelligent and powerful than their predecessors. For users in the US and thirteen other countries—but noticeably absent from Europe—Meta's AI assistant feature has been integrated into Facebook, Instagram, and WhatsApp.

The public content that Europeans share on our services and elsewhere, such as posts or comments, is what trains our models on. Without this, models and the AI features they drive will not able to effectively comprehend significant regional languages, cultures, or popular social media subjects. In a blog post, Stefano, the head of worldwide engagement for Meta's privacy policy, stated.

To provide users in Europe an opportunity to opt out before the global AI training regulations take effect on June 26, Meta claims to have issued billions of alerts to consumers since May 22. The Instagram behemoth guarantees that users can refuse at any time and that their photos will never be used to train artificial intelligence algorithms.

The company stated that it needs to use public data from those users to train its Llama AI large language model to better reflect the languages, geography, and cultural references of its users in Europe.

Opting out is simply not an option in the rest of the globe, which makes this quite distinct. Granted, it's too late to refuse training data used for Meta's LLaMa 3, but for Facebook and Instagram users outside of the EU, training is also required for next models. It's now a function that's only available to users inside the EU, but people outside of Europe could be able to choose to opt out in the future.

It's unlikely that there won't be any opposition at all, even if Meta probably believes it's in an excellent position to begin using user data from European users. The social media behemoth hinted at its plans last week when it updated its privacy policy, even before Facebook made its formal announcement. This led to complaints being filed around Europe by the consumer privacy advocacy group noyb.

According to Noyb, user data gathering should always be opt-in rather than opt-out. The European Union's Right to be Forgotten is another factor that is likely to present issues, since data cannot actually be removed from an LLM or other AI model.

Current Issue
Ace Micromatic : Pioneering Excellence in Comprehensive Manufacturing Solutions