Meta, the parent company of Facebook, is embarking on an ambitious journey to train a new artificial intelligence (AI) model that rivals OpenAI's latest chatbot, GPT-4. This endeavour includes a significant investment in Nvidia H100 AI-training chips and an expansion of the company's AI infrastructure, as reported by The Wall Street Journal. Meta aims to initiate the training of this new large language model (LLM) in early 2024.
Mark Zuckerberg, CEO of Meta, is reportedly considering an initial phase where access to the new AI model will be provided free of charge to businesses and researchers interested in crafting AI tools. This strategic move aligns with Meta's vision of accelerating the development of AI tools capable of replicating human expressions.
In a recent development, Meta introduced Code Llama, an AI tool tailored for generating code and debugging human-written programming work. This LLM utilizes text prompts to create and engage in discussions about code, a promising step that could enhance developer productivity and lower entry barriers for coding enthusiasts.
Code Llama represents a code-specialized variant of Llama 2, achieved through further training on code-specific datasets and extensive data sampling from the same dataset. The tool has garnered recognition for its prowess in coding tasks, signifying its potential as a productivity and educational aid for programmers striving to produce robust, well-documented software.
ALSO READ: Apple's upcoming event expected to launch the new iPhone 15 with USB-C and iOS 17
Meta's foray into the realm of AI intensifies its focus on generative AI technologies, as previously revealed by Mark Zuckerberg in February of this year. With a dedicated team assembled to construct this groundbreaking AI model, Meta aims to rival the capabilities of OpenAI's GPT-4, propelling the AI industry into a new era of innovation and applications.
ALSO READ: Google introduces AI-driven 'Proofread' feature in Gboard Keyboard
Inputs from IANS
