– OpenAI revealed GPT-4o before Google announced Gemini improvements, followed by Microsoft announcing Copilot+ PCs and AI improvements
– Meta released a research paper on its multi-modal AI model Chameleon, claiming it competes with models like Gemini, GPT-4, and Llama-2
– Chameleon is built on an early-fusion token-based mixed-modal architecture, similar to Google’s Gemini, and Meta has been pushing open source initiatives for its AI models
In the ongoing arms race between technology companies in the AI space, OpenAI, Google, Microsoft, and Meta have all made significant announcements in the last two weeks. OpenAI unveiled GPT-4o, Google announced improvements to its Gemini model, Microsoft introduced Copilot+ PCs with AI enhancements, and Meta quietly released a research paper on its multi-modal AI efforts.
Meta’s research paper showcases its development of a state-of-the-art multi-modal large language model called Chameleon. Built on an early-fusion token-based mixed-modal architecture, Chameleon combines images, code, text, and other inputs to create sequences, similar to Google’s Gemini model.
Unlike late fusion multi-modal architectures that fuse together separate models trained on single modalities, Chameleon and Gemini are built on early-fusion architectures that learn from a combination of inputs from the start. Chameleon is claimed to be an end-to-end model and competes evenly with other AI models like Gemini, GPT-4, and Meta’s own Llama-2.
The paper suggests that if the claims about Chameleon are true and reproducible, it could potentially surpass many existing AI models. Meta’s focus on open-source initiatives, such as the existing Llama 3 and opening up the Quest headset operating system to hardware manufacturers, indicates a commitment to advancing AI technology through collaboration and innovation.
While the release date of the Chameleon model remains uncertain, Meta continues to make strides in AI development, with the recent launch of the Llama 3 assistant on Facebook, Instagram, and WhatsApp. The company’s dedication to cutting-edge technology and open-source initiatives suggests a promising future for AI advancements.