π» Boost with Fountain.FM β Fountain 1.0 has a new UI, upgrades, and super simple Strike integration for easy Boosts.
Building for our AI future β But ultimately we are a workplace and our policies and expectations are clear: this is a business, and not a place to act in a way that disrupts coworkers or makes them feel unsafe, to attempt to use the company as a personal platform, or to fight over disruptive issues or debate politics.
Meta Debuts Llama 3, Latest AI Model for Powering Chatbots β Llama 3, unveiled Thursday, is an upgrade from an AI model that Meta released last summer. Chief Product Officer Chris Cox said that model, Llama 2, has been downloaded 170 billion times. Llama 2 was billed as open source, though its license included some restrictions such as requiring companies with more than 700 million users to ask for permission to use it.
Dylan Patel on X: β Synthetic data pipelines are massive improvements over internet data
Flywheel only continues with big models too when these techniques are applied
Aran Komatsuzaki on X: "A few caveats about Phi-3:" β Phi-3-medium performs well on TriviaQA but noticeably underperforms rel. to GPT-3.5. We can guess that Phi-3 recipe doesn't magically make it understand more random factoids. It's more focused on valuable knowledges (e.g. STEM), which makes sense.
π» Boost with Fountain.FM β Fountain 1.0 has a new UI, upgrades, and super simple Strike integration for easy Boosts.
Building for our AI future β But ultimately we are a workplace and our policies and expectations are clear: this is a business, and not a place to act in a way that disrupts coworkers or makes them feel unsafe, to attempt to use the company as a personal platform, or to fight over disruptive issues or debate politics.
Meta Debuts Llama 3, Latest AI Model for Powering Chatbots β Llama 3, unveiled Thursday, is an upgrade from an AI model that Meta released last summer. Chief Product Officer Chris Cox said that model, Llama 2, has been downloaded 170 billion times. Llama 2 was billed as open source, though its license included some restrictions such as requiring companies with more than 700 million users to ask for permission to use it.
Dylan Patel on X: β Synthetic data pipelines are massive improvements over internet data
Flywheel only continues with big models too when these techniques are applied
Aran Komatsuzaki on X: "A few caveats about Phi-3:" β Phi-3-medium performs well on TriviaQA but noticeably underperforms rel. to GPT-3.5. We can guess that Phi-3 recipe doesn't magically make it understand more random factoids. It's more focused on valuable knowledges (e.g. STEM), which makes sense.