Meta Platforms Inc., the parent company of Facebook and Instagram, is set to release a new generative AI model. This open-source model is anticipated to rival the prowess of GPT-4 and is expected to be “several times more powerful” than the earlier Llama 2 AI (Artificial Intelligence) model. Mark Zuckerberg is directly involved in this initiative, and the aim is to provide a platform for other firms to develop AI-driven services related to text generation, image creation, and analysis.
The driving force behind this move is Meta’s desire to lead in the creation of next-gen AI models that can produce human-like expressions. Despite significant progress by competitors like Google’s Bard chatbot and startups like Cohere Inc., Zuckerberg seeks to position Meta at the forefront of the AI sector.
OpenAI, backed by giants like Microsoft Corp., has been a dominant force this year. Services like ChatGPT have gained widespread popularity, leading to revenue streams through ChatGPT Plus and ChatGPT for Business. Google has also launched Bard and has plans for an advanced model named Gemini. These efforts come at a significant cost, and both OpenAI and Meta have invested heavily, with Meta specifically sourcing Nvidia Corp.’s H100 GPUs for training their AI.
Despite the enormous costs associated with AI development, Meta remains committed to an open-source model. Their approach ensures that other companies can freely utilize and build upon their tools. Meta’s commitment to open-source is evident in their consistent contributions to the AI community, releasing models like Code Llama, SeamlessM4T, AudioCraft, Voicebox, I-JEPA, and datasets like FACET this year alone.
Open-source models, as highlighted by industry experts, offer adaptability, which could give Meta an edge. However, they aren’t without challenges, including copyright issues and misuse for spreading disinformation. Additionally, Zuckerberg sees potential in generative AI for developing 3D visuals for the metaverse and enhancing features in Instagram, Messenger, and WhatsApp. However, the success of Meta’s AI pursuits, especially against OpenAI, is contingent on the volume of training data, with Llama 2 using 70 billion parameters and estimates for GPT-4 being around 1.5 trillion parameters.
Meta’s investment in an open-source model could redefine generative AI’s future, offering affordable and adaptable solutions. As AI continues to evolve, the industry will keenly watch how this strategic move reshapes the landscape, challenging established players while setting new standards for innovation and application.
Source: https://newslinker.co/
