As the AI competition between internet rivals Google, Microsoft and OpenAI intensifies, Meta is creating a top-level product focused on generative AI. Led by Ahmad Al-Dahle, this team will report directly to Chief Product Officer Chris Cox and aim to expedite AI work throughout the company, according to Bloomberg reporting.
Generative AI, or the capacity for creating content, is one of the key technologies reshaping the future of technology. While it has become a hotspot for investments in this sector, growth has slowed substantially resulting in widespread layoffs and reduced experimental bets.
Meta has unveiled its own large language model, LLaMA, to disrupt the artificial intelligence space which has long been dominated by internet rivals like Google and Microsoft.
According to Meta, their system has "shown a lot of promise" when it comes to generating text, having conversations, summarizing written material as well as more complex tasks such as solving math theorems or predicting protein structures.
LLaMA was developed with researchers in mind, rather than the general public. It has been trained on large amounts of unlabeled data from CCNet, C4, GitHub, Wikipedia, books, ArXiv and Stack Exchange according to a spokesperson for CNNMoney.
This new model can generate content on its own based on queries, and it's optimized to run on a single GPU. It comes in various sizes (from 7 billion parameters up to 65 billion) that will be made available to the AI research community.
Meta is working to democratize access to large language models (LLMs), which have long been recognized as one of the most powerful and beneficial forms of AI.
Unfortunately, most researchers have been unable to take advantage of LLMs due to their need for highly advanced computational infrastructure in order to train and run them effectively. With LLaMA, Meta hopes to change this by making LLMs accessible again to everyone with an interest in research into language models.
Facebook CEO Mark Zuckerberg recently declared LLaMA "the next step in exploring generative AI's potential," signaling that his company is refocusing its efforts on this technology.
Meta's AI-focused chat in WhatsApp and Messenger has been a hit, but it's not as efficient as other chatbots. To make the experience more conversational and less robotic, the company released an AI chatbot last year that generated scientific articles and math problems; however, this bot later proved inaccurate.
Creative Instagram Filters and Advertising Formats, Too
In recent months, the company has been pushing creative Instagram filters and ad formats in an effort to boost engagement. Along with traditional square and circular images, it's added various filters that alter color or shape when users scroll through them.Kosoy told CNNMoney, he plans to expand the functionality into other platforms and applications, as well as add the capacity to hide irrelevant data so users don't get distracted by extra fields they see. This way, they can stay focused on their tasks instead of getting distracted by irrelevant details.
Managing Work with More Flexibility
By streamlining their processes and eliminating confusion, the team has been able to streamline work and prevent confusion.For instance, they created a job-specific workspace so team members stay on track and see upcoming project milestones. They can also expand the view when necessary to view additional fields - an enormous timesaver for development relationships teams which often need to respond to user requests for changes to data formats.