AI Unleashed: Llama Takes the Crown as the Biggest AI Story of 2023
VentureBeat is excited to present AI Unleashed, an exclusive executive event for enterprise data leaders. This event provides a unique opportunity for networking and learning with industry peers. Discover more about this event and join us to explore the latest innovations in AI.
Could a furry camelid steal the spotlight as the biggest AI story of 2023? If we consider Llama, Meta’s large language model, which made waves in the AI research world in February, followed by Llama 2 in July and Code Llama in August, the answer is a resounding yes.
Now, I can already hear the objections. “What about ChatGPT? It launched in November 2022, reached 100 million users by February, and brought generative AI into popular culture!” But please hear me out.
ChatGPT, undoubtedly, was a game-changer in the world of generative AI. As Forrester analyst Rowan Curran noted, it ignited the fire around generative AI. However, when Meta released Llama in February, it sparked a red-hot debate and solidified open source AI as a significant force in the industry.
According to Meta, since Llama’s release, the open source AI community has produced over 7,000 Llama derivatives on the Hugging Face platform. Popular variations like Koala, Vicuna, Alpaca, Dolly, and RedPajama have emerged, making it akin to the Marvel Universe with its numerous spin-offs and offshoots.
Kjell Carlsson, head of data science strategy and evangelism at Domino Data Lab, expressed his belief that Llama and its open source AI cohorts will foster real-world, impactful GenAI applications and strengthen the foundations of open-source GenAI going forward.
The era of closed, proprietary models began with ChatGPT, which launched in 2022. However, in 2023, OpenAI co-founder Ilya Sutskever admitted that sharing their research was a mistake due to competitive and safety concerns. In contrast, Meta’s chief AI scientist, Yann LeCun, pushed for the release of Llama 2 with a commercial license and model weights, emphasizing the need for open access to large language models.
Carlsson contends that comparing ChatGPT and Llama is like comparing apples and oranges. He believes Llama 2 is the real game-changer due to its open-source nature and the ability to fine-tune and operationalize it at scale. However, ChatGPT’s significance lies in bringing LLMs into the public consciousness, especially among business leaders.
Matt Shumer, CEO of Otherside AI, acknowledged that Llama’s impact wouldn’t have been possible without ChatGPT laying the groundwork. Nevertheless, he firmly believes that Llama has paved the way for countless companies to thrive.
Sridhar Ramaswamy, SVP of data cloud company Snowflake, stated that Llama 2 is the first truly capable open source AI model and has prevented the formation of a monopoly in the AI industry.
The leak of the Llama weights on 4chan allowed developers worldwide to access a GPT-level LLM, leading to a surge of new derivatives. Meta’s decision to release Llama 2 for commercial use, along with Microsoft making it available on Azure, further propelled its influence.
Despite concerns about misuse and potential harm, Meta remains committed to open-source AI. Mark Zuckerberg, Meta’s CEO, reaffirmed the company’s dedication to an “open science-based approach” during an all-hands meeting. Meta’s long-standing advocacy for open research includes creating an open-source ecosystem for the PyTorch framework.
As 2023 draws to a close, Meta will celebrate FAIR’s 10th anniversary (Fundamental AI Research). This milestone highlights the company’s enduring commitment to advancing AI through open research. Joelle Pineau, VP of AI research at Meta, joined the organization in 2017 due to FAIR’s dedication to transparency and open science.
Open research is now not just about raising the bar on research quality, but also driving the productivity of the entire AI ecosystem. Numerous startups are grateful for the alternative model provided by Meta.
In conclusion, while ChatGPT made a tremendous impact on the AI landscape, Llama and its open source AI counterparts have solidified their position as the long-term game-changers. With Meta at the forefront of open research and the continuous evolution of Llama derivatives, we can expect more real-world applications and a strong foundation for GenAI going forward.
