Meta's AI assistant is a new product that can generate text, audio, and imagery based on user input.
Meta, formerly known as Facebook, is one of the leading tech companies in the world, with over 3 billion monthly active users across its platforms. Meta is also one of the pioneers in artificial intelligence (AI), with various products and services that use AI to enhance user experience, such as face recognition, content moderation, and recommendation systems. One of the latest AI products that Meta unveiled at its annual Connect conference in September 2023 is Meta AI, a virtual assistant that can generate text, audio, and imagery based on user input. Meta AI can help users with various tasks, such as writing, coding, researching, planning, and more. Meta AI can also access real-time information via a partnership with Microsoft's Bing search engine.
But how did Meta create such a powerful and versatile AI assistant? Meta used public posts from Facebook and Instagram to train parts of its AI assistant. These posts included both text and photos that users shared publicly on the platforms. Meta also used some publicly available and annotated datasets to train its AI assistant. Meta's president of global affairs Nick Clegg told Reuters that the company tried to exclude datasets that have a heavy preponderance of personal information, such as LinkedIn. He also said that Meta did not use private posts or chats as training data for the model. He added that Meta imposed safety restrictions on what content the AI assistant could generate, such as a ban on the creation of photo-realistic images of public figures.
Meta's approach of using public posts from Facebook and Instagram to train its AI assistant has some implications and challenges. On one hand, it can be seen as a way of leveraging the vast amount of data that Meta has access to and providing value to its users by creating a useful and personalized AI product. On the other hand, it can also raise some ethical and legal concerns about privacy, consent, and ownership of the data that Meta uses without permission.
These questions are not easy to answer, as they involve complex issues of data governance, transparency, accountability, and responsibility. Meta may have to face some legal challenges or regulatory scrutiny from authorities or stakeholders who may question its data practices and policies. Meta may also have to deal with some backlash or criticism from users or activists who may feel violated or exploited by its data use.