Member-only story
[GPT4All Tutorial] How To Run Your Personal, ChatGPT-Like Model, Locally
Your own personal AI assistant
Context
It’s not breaking news to say that large language models — or LLMs — have been a hot topic in the past months, and sparked fierce competition between tech companies.
OpenAI has ChatGPT, Google has Bard, and Meta has Llama. Llama is a special one; its code has been published online and is open source, which means that everyone can access it and see how the model was built.
It’s better than nothing, but in machine learning, it’s far from enough: without the training data or the final weights (roughly speaking, the parameters that define a model’s decision-making), it’s virtually impossible to reproduce the model.
Open source, but not reproducible
To explain this, let’s say you have a company that produces a popular type of soda and for the sake of transparency, publicly shares the plans of its soda-making factory. Without the initial ingredients or the final composition, it would be impossible to reproduce their beverage.
