Understanding GPT-4: Text Input Requirements, NLP Task Support, Performance Considerations, Cost Analysis, and Integration with spaCy

Community
,
1

Hi,

Im working on a project and would like to know the following:

  1. What are the specific text input requirements for GPT-4? Does it expect tokenized text, sentence-level input, or paragraph-level input?
  2. Which natural language processing tasks are supported by GPT-4 out of the box? For example, does it provide built-in capabilities for tokenization, named entity recognition, part-of-speech tagging, or any other linguistic annotations?
  3. Are there any known performance considerations or hardware requirements for optimal usage of GPT-4? Does it require powerful hardware or specialized infrastructure to achieve optimal performance?
  4. What is the pricing structure, usage limits, and associated costs for utilizing GPT-4? Are there any constraints or considerations related to cost when using GPT-4?
  5. How flexible is the integration of GPT-4 with other tools like spaCy? Are there any known challenges or considerations when combining the preprocessing capabilities of spaCy with GPT-4?

Thankyou for any clarity you can provide

2

  1. Input is natural language ascii text, no pre tokenisation is required.

  2. All processing tasks are supported out of the box, if those tasks are purely text based. There is currently no audio or image input to the system as is. There are no limitations on tasks that can be requested by the prompt (with legal restrictions ofc)

  3. Inference is performed on enterprise level hardware behind the API and the results returned to you. The calling hardware only has to be able to create and maintain an internet connection for the duration of the transaction and handle the returning text, nothing else is required. I have implementations of ChatGPT on microcontrollers.

  4. GPT-4 8K starts out at $0.03 per 1K tokens as input (prompts) and $0.06 per 1K tokens output (completions) details can be found here Pricing

  5. The API can be used with any other language or system, so long as you are able to implement basic JSON string handling and related text processing tasks (serialisation etc)

2 Likes
3

Thanks Foxabilo for these details. You clarity on everything really helped me what understand what to do from this point. It would have taken me ages to find the answers to this.

kind regards

Pete

1 Like