![How these EU countries may ‘stop’ Facebook-parent Meta from training its AI model](https://img.etimg.com/thumb/msid-111211182,width-650,height-488,imgsize-20454,resizemode-75/how-these-eu-countries-may-stop-facebook-parent-meta-from-training-its-ai-model.jpg)
During its pilot in India, users have experienced Meta AI for use cases such as information gathering, learning, questions about coding, social media content generation such as captions, Instagram threads, generating stylised brand logos etc.
One of the advantages with MetaAI is that it is integrated in all daily use platforms and does not need any app downloads or website to generate responses unlike Google's Gemini, OpenAI's ChatGPT or Microsoft Copilot, said Ryan Cairns, VP of Engineering, Meta.
“I think the benefit of the way that we've rolled out meta AI is that it's available in all of our apps, regardless of the platform. You can access it in WhatsApp, whether it's on Android or iOS, you get it on the web,” Cairn told the media in a virtual briefing.
“Our apps provide both the biggest distribution and the most availability to the end user. The Google Assistant works really well on Android. Siri works really well on Apple devices, but beyond those contexts, you can't really serve the end user,” he said.
When asked about Meta using users’ personal data to train its model, Cairns said that using public information to train AI models is an industry wide practice, “and we're going to continue to do that, it's very important to us to build AI in a responsible way. So, things like safety and red teaming are of the utmost priority and top of mind for us.”
Discover the stories of your interest
Regular updates
He added that Meta’s models are updated every two weeks to account for user feedback, prevent misinformation, control hallucinations and use this data to train and adapt the models.
“We know that we don't get every answer right. But the fact that we can turn the corner quickly and iterate means that we can learn and adapt and improve the quality bar over time,” the executive said.
When asked about probable misuse of Meta’s image generation capabilities to create deepfakes Cairns said that Meta’s images are stylized and not photorealistic. “We don't generate images of public figures or any explicit images and things like that.”
As per policy, Meta is already using “Made with AI” labels on AI-generated video, audio and images based on detection of industry-shared signals or self-disclosures by people when they are uploading such content.