I'm out of the loop but is GPT-3 a pre-trained model like RedNet is for images? If such how can it be used easily? The article mentions getting access to an AP

I'm out of the loop but is GPT-3 a pre-trained model like RedNet is for images? If such how can it be used easily? The article mentions getting access to an API, so is it an online service? Or, there's the pre-trained model (please, a tutorial!) and a service for those that don't want to configure it locally?