OpenAI’s got exciting news! They’re letting folks customize their AI buddy, GPT-3.5 Turbo, with their own data. This means you can make it smarter and bend it to your needs.
They say that if you fine-tune this GPT-3.5 Turbo, it could even beat GPT-4 at certain tasks. Fancy that!
Businesses and tech whizzes can now tweak the model to make it follow rules better, speak in a certain language, format text neatly, or match a specific style.
Plus, you can save money and speed things up by using shorter instructions. Some folks have trimmed their prompts by 90%!
Just know, to fine-tune, you’ll have to prepare data, upload files, and use OpenAI’s tools. They’ve got safety checks to keep things clean.
And here’s the price tag:
- Training costs $0.008 per 1,000 tokens
- Input costs $0.012 per 1,000 tokens
- Output costs $0.016 per 1,000 tokens.
Tokens are just the chunks of text, like “fantastic.”
Also, they’ve got some new GPT-3 models (babbage-002 and davinci-002) with cool features, but they’re retiring the old ones on January 4, 2024.
Oh, and GPT-4 is coming soon, and it’s going to understand images too! But they didn’t spill the beans on when exactly.