Public Service Announcement For CustomGTPs and ChatGPT…

OpenAI has several products it offers, a few of them are ChatGPT and an Application Programming Interface (API).

ChatGPT is a user interface only, NOT a Large Language Model (LLM)!

ChatGPT accesses Large Language Models created by OpenAI, specifically Generative Pre-Trained Transformers (GPTs).

GPTs were developed by Open AI.

Several other LLM providers have their own LLM technology.

For example, Google has Bert, PaLM and LaMDA.

Meta has LLaMA and, OPT.

Anthropic has Cluade.

IBM has Watson.

Etc.

One cannot ebbed ChatGPT (Or a CustomGPT) on a website, because OpenAI provides you with free tokens or a maximum amount of tokens (if you are a paid user) to keep it and you in the ChatGPT interface.

They want the data.

If they forgo the data, then one needs to pay for the tokens via their API.

Using the API, one can replicate a ChatGPT-style (the word style used intentionally) interface (although there is likely some extra secret sauce to prevent anyone from creating the exact output) and embed it outside the ChatGPT interface.

They would undermine the ChatGPT business model.

Likewise, CustomGPTs are just applications built with the ChatGPT interface using a GPT as the LLM model.

One CANNOT embed a CustomGPT outside of the ChatGPT interface for all the above reasons.

There are, however, dozens and dozens of vendors who offer to embed “CustomGPT-style” (the word style used intentionally) chatbots on websites etc.

Although they may use OpenAI GPT models, these are NOT CustomGPTs.

They may also use other LLM providers at the base model to power them via the relevant API products.

GPT is an OpenAI technology, regardless of what others say or believe.

Hopefully, this clears up some confusion about what you can and cannot do and with ChatGPT and CustomGPT and what they are and what they are not.