Facebook parent Meta recently announced Llama 3.1 open source AI model, enabling developers to access it and accelerate AI development. CEO Mark Zuckerberg said that it is the most advanced open-source AI model to date, and outlined the company’s commitment to open-source AI, arguing that it benefits both Meta and the broader tech community.
Open-source software, like Android, allows for widespread collaboration and innovation by providing others access to the source of the code. In contrast, proprietary software, such as iOS, is owned and controlled by a single company. The CEO said he believes that an open-source approach will accelerate AI development and benefit users worldwide.
“When I talk to developers, CEOs, and government officials across the world, I usually hear several themes,” Zuckerberg said in a blog post
Zuckerberg on why open source AI models are way to go
Zuckerberg said that since organisations need the ability to tailor AI models to their specific needs, they need to train and fine-tune models with proprietary data and then optimise model size for different applications.
“Now you’ll be able to take the most advanced Llama models, continue training them with your own data and then distil them down to a model of your optimal size – without us or anyone else seeing your data,” he said.
Secondly, relying on closed-source AI models from a single provider creates dependency and potential risks, such as changes in terms of service or model availability. Open-source models offer greater control and flexibility.
“Open source enables a broad ecosystem of companies with compatible toolchains that you can move between easily,” the CEO noted.
The third important thing is data privacy as handling sensitive data often necessitates on-premises AI solutions. Zuckerberg said open-source models allow organisations to keep data secure within their own environments.
“Open source addresses privacy issues by enabling organisations to run the models wherever they want. It is well-accepted that open source software tends to be more secure because it is developed more transparently,” he noted.
Finally, open-source models can be more cost-effective to run compared to proprietary alternatives, both for user-facing and internal applications, Zuckerberg noted.
“Developers can run inference on Llama 3.1 405B on their own infra at roughly 50% the cost of using closed models like GPT-4o, for both user-facing and offline inference tasks,” he said.
Open-source software, like Android, allows for widespread collaboration and innovation by providing others access to the source of the code. In contrast, proprietary software, such as iOS, is owned and controlled by a single company. The CEO said he believes that an open-source approach will accelerate AI development and benefit users worldwide.
“When I talk to developers, CEOs, and government officials across the world, I usually hear several themes,” Zuckerberg said in a blog post
Zuckerberg on why open source AI models are way to go
Zuckerberg said that since organisations need the ability to tailor AI models to their specific needs, they need to train and fine-tune models with proprietary data and then optimise model size for different applications.
“Now you’ll be able to take the most advanced Llama models, continue training them with your own data and then distil them down to a model of your optimal size – without us or anyone else seeing your data,” he said.
Secondly, relying on closed-source AI models from a single provider creates dependency and potential risks, such as changes in terms of service or model availability. Open-source models offer greater control and flexibility.
“Open source enables a broad ecosystem of companies with compatible toolchains that you can move between easily,” the CEO noted.
The third important thing is data privacy as handling sensitive data often necessitates on-premises AI solutions. Zuckerberg said open-source models allow organisations to keep data secure within their own environments.
“Open source addresses privacy issues by enabling organisations to run the models wherever they want. It is well-accepted that open source software tends to be more secure because it is developed more transparently,” he noted.
Finally, open-source models can be more cost-effective to run compared to proprietary alternatives, both for user-facing and internal applications, Zuckerberg noted.
“Developers can run inference on Llama 3.1 405B on their own infra at roughly 50% the cost of using closed models like GPT-4o, for both user-facing and offline inference tasks,” he said.
You may also like
"Submit NRC application number to get Aadhaar card...": Assam CM Himanta Sarma makes big announcement
Drone visuals show intensified search operation in Bahraich to catch remaining two wolves
Chinese man dies after working for 104 days with 1 day off, company held liable
CRED Users Left High And Dry As Fintech Company Cancels Jackpot Promising Macbook, iPad, AirPods Due To 'Technical Glitch'
