Meta uses Public Facebook and Instagram Posts to train its Virtual Assistant

Meta Platforms uses Public Facebook and Instagram Posts to train its Virtual Assistant | CIO Women Magazine

Meta Platforms, the parent company of Facebook and Instagram, has revealed its approach to training its new Meta AI virtual assistant. In an exclusive interview with Reuters, Nick Clegg, the President of Global Affairs at Meta, emphasized the company’s commitment to respecting user privacy in the development of AI technologies.

According to Clegg, Meta used publicly available Facebook and Instagram posts as part of the training data for Meta AI. However, they deliberately excluded private posts shared only with close friends and family to safeguard users’ privacy. Additionally, private chats from Meta’s messaging services were not utilized in the AI model’s training. Meta also took measures to filter out private details from the public datasets used for training.

Clegg stated, “We’ve tried to exclude datasets that have a heavy preponderance of personal information,” emphasizing that the “vast majority” of the data used for training was publicly accessible. He cited LinkedIn as an example of a platform that Meta Platforms chose not to include in its training data due to privacy concerns.

These actions by Meta come at a time when tech companies like Meta, OpenAI, and Alphabet’s Google face criticism for using internet-scraped information without explicit permission to train their AI models, which rely on vast amounts of data to generate content and summaries.

Meta Platforms AI Unveiled at Connect Conference

Meta Platforms AI took center stage at Meta’s annual Connect conference, marking a significant step in the company’s journey into consumer-facing AI tools. Meta CEO Mark Zuckerberg introduced the new product during the event. Unlike previous conferences, this year’s focus shifted from augmented and virtual reality to artificial intelligence.

Meta AI was developed using a custom model, primarily based on the powerful Llama 2 large language model, which Meta made available for public commercial use in July. Additionally, a new model called Emu was employed to generate images in response to text prompts. The virtual assistant is designed to generate text, audio, and imagery, and it will have real-time access to information through a partnership with Microsoft’s Bing search engine.

During the development of Meta AI, both text and photos from public Facebook and Instagram posts were used. Emu was trained using this data for image generation, while chat functions were built using Llama 2, supplemented by publicly available and annotated datasets. Meta Platforms also indicated that interactions with Meta Platforms AI would contribute to ongoing improvements in its features.

Clegg emphasized that Meta had implemented safety restrictions on what content the Meta AI tool could generate, including a ban on creating photo-realistic images of public figures.

Copyright Concerns and Litigation

One of the significant challenges faced by companies like Meta Platforms in the development of AI models is the potential infringement of copyrights. Clegg expressed his expectation of a “fair amount of litigation” surrounding the issue of whether creative content is covered by existing fair use doctrine, which permits limited use of protected works for purposes like commentary, research, and parody.

Regarding copyrighted materials, Clegg acknowledged that the matter was likely to be decided through legal proceedings. Some AI companies have taken steps to avoid reproducing copyrighted imagery. For instance, OpenAI signed a six-year agreement with content provider Shutterstock to use its libraries for training.

When asked if Meta had taken similar steps, a Meta spokesperson pointed to their new terms of service, which prohibit users from generating content that violates privacy and intellectual property rights.

Meta’s commitment to user privacy and its cautious approach to copyright issues reflect the company’s efforts to navigate the complex landscape of AI development while addressing ethical and legal concerns. As AI technology continues to evolve, these principles will remain at the forefront of Meta’s strategy.

Share:

LinkedIn
Twitter
Facebook
Reddit
Pinterest

Related Posts