Meta Releases Beefed-Up AI Models
Meta, the tech titan formerly known as Facebook, has introduced an improved AI assistant built on new versions of its open-source Llama large language model. In a blog post, Meta announced that its AI assistant, Meta AI, is now smarter and faster thanks to advances in the publicly available Llama 3.
According to Meta co-founder and CEO Mark Zuckerberg, “We believe Meta AI is now the most intelligent AI assistant that you can freely use.” The company is excited about the potential of generative AI technology and aims to develop and release it in a way that reduces risk and incorporates user feedback.
Customizable and Open-Source
One of the key features of Meta AI is that it is open-source, allowing developers outside of Meta to customize the Llama 3 model as they wish. Meta can then incorporate these improvements and insights into future versions of its AI assistant. This open approach fosters collaboration and innovation within the broader ecosystem.
Meta is also committed to incorporating protections and being cautious when adding generative AI features to its platforms, including Facebook, Instagram, WhatsApp, and Messenger. The company wants to ensure that the technology is used responsibly and that potential risks are minimized.
Integration and Accessibility
Meta has made efforts to make Meta AI more accessible and easier to use across its apps. The AI assistant has been built into the search boxes of WhatsApp, Facebook, and Instagram Messenger, allowing users to ask questions directly from these platforms. This integration aims to provide a seamless user experience and make AI assistance readily available.
However, it’s important to note that AI models, including Meta AI, have been known to occasionally provide inaccurate or bizarre responses, referred to as “hallucinations.” Meta acknowledges this issue and is continuously updating and improving its AI assistant to address such concerns.
Meta’s Approach and Future Plans
Meta’s slower approach to building its AI has put the company behind in terms of consumer awareness and usage. However, it still has significant potential due to its massive user base across its social media apps. Meta can leverage this user base to test AI experiences and gather valuable insights.
By integrating AI into its family of apps, Meta aims to quickly deliver features powered by the technology to billions of people. This approach allows Meta to observe how users interact with AI and make improvements accordingly.
Meta also emphasizes the importance of refining AI models to provide balanced and unbiased responses. For example, the company has tuned Llama 3 to better discern innocuous prompts from out-of-bounds ones. Meta wants its AI assistant to provide relevant points about political or social issues without offering a single point of view.
Transparency and Multilingual Capabilities
Meta is committed to transparency when it comes to AI-generated content. The company notifies users when they are interacting with AI on its platform and puts visible markers on photorealistic images that were generated by AI. Starting in May, Meta will also label video, audio, and images as “made with AI” when it detects or is informed that the content is generated by AI.
While Llama 3 is currently based in English, Meta plans to release more capable models in the coming months that can converse in multiple languages. This expansion will further enhance the accessibility and usefulness of Meta AI across different regions and cultures.
Overall, Meta’s release of beefed-up AI models, particularly the improved Meta AI assistant built on Llama 3, showcases the company’s commitment to advancing AI technology while prioritizing responsible development and user experience.
