Meta's Llama 3.1 Assertively Stands to Take on Established, Closed-Source Competition
Meta, the tech giant behind Facebook, has unveiled its latest open-source AI model, Llama 3.1. Claimed to be the leading open-source model of its size, Llama 3.1 promises to outperform leading private models from OpenAI, Google, and Anthropic. However, the use of the term 'open source' for Llama models has been met with criticism and concerns.
Jeffrey Ladish, executive director at Palisade Research, has warned that the safety guard rails of the 70 billion parameter version of Llama 3 could be stripped off in minutes for a negligible cost. Theo Priestley, a technology futurist and author, has stated that Meta's use of the term 'open source' for Llama models is misleading.
Llama 3.1 enables researchers and developers worldwide to explore, innovate, and build upon state-of-the-art language AI without proprietary APIs or expensive licensing fees. However, approval from Meta is required before large enterprises can utilize Llama 3.1, and the company places restrictions on its use for such enterprises with hundreds of millions of users.
The Open Source Initiative (OSI) and the Free Software Foundation (FSF) have criticised Meta's licensing of Llama, arguing that it does not meet the OSI's Open Source Definition. Specifically, Llama's licenses have restrictions that prevent commercial use in certain fields, such as controlled substances and critical infrastructure, and initially restricted use by individuals in the European Union. The FSF classified Llama 3.1's license as nonfree due to these restrictions.
Critics argue that Llama's license discriminates against certain persons or groups and fields of endeavor, which is contrary to the principles of open-source software. This includes restrictions on commercial use and specific fields, which some see as "openwashing" rather than truly open-source.
The OSI also criticised Meta for not disclosing the training data used for Llama, which is now expected under the newly published Open Source AI Definition. Meta has disagreed with this definition, leading to further controversy.
There are concerns that Meta might move away from open sourcing its AI models, potentially ceding ground to other companies or countries that embrace open-source AI more fully. This shift could impact the global AI landscape by favoring closed ecosystems over open-source models.
Despite these criticisms, Meta's founder and CEO, Mark Zuckerberg, has reiterated his claim that open source is the future of AI, projecting it will follow a similar path to open source operating systems.
[1] Source: The Verge, TechCrunch, Ars Technica [4] Source: The Register, VentureBeat, ZDNet
- Amid ongoing criticisms about Meta's licensing of Llama and concerns about its commitment to open source, Theo Priestley, a technology futurist, suggested that the company's use of 'cloud computing' for Llama could pose an additional challenge to 'cybersecurity', as the lack of transparency in training data could potentially lead to unintended consequences.
- As Meta continues to face criticism for the limitations it imposes on the use of Llama, researchers and developers worldwide have raised concerns about the impact on 'technology' innovation and collaboration. Some fear that Meta's restrictive practices could favor closed ecosystems, potentially tilting the future of AI development towards more proprietary models.
