Jerome Pesenti is celebrating Meta’s recent decision to release Llama 3, a powerful open-source large language model that is available for anyone to download, run, and build upon. Pesenti, the former vice president of artificial intelligence at Meta, had been advocating for the company to make its technology more accessible to others. Now, with the release of Llama 3, Pesenti’s new startup, Sizzle, can benefit from a high-quality AI model that rivals OpenAI’s GPT-4 but is more cost-effective and open to modification.
Pesenti believes that the release of Llama 3 could signal a shift in the AI landscape, potentially leading developers and entrepreneurs to choose open-source models over proprietary ones like GPT-4 and Google’s Gemini. Sizzle’s engineers are currently exploring the possibility of using Llama 3 in place of OpenAI’s model for crafting problem sets and curricula for students.
In addition to Llama 3, other open-source AI models, such as Mixtral 8x22B from Mistral, are gaining attention for their capabilities and efficiency. Meta’s latest model, Llama 3, has been enhanced with larger amounts of high-quality training data, making it more powerful and cost-effective to run compared to other models like GPT-4.
The emergence of open-source AI models like Llama 3 could level the playing field between open and closed models, leading to increased competition and innovation in the AI industry. With ongoing developments in AI technology, including the potential for even more powerful open-source models in the future, the landscape of AI development is set to evolve significantly.
