Q&A

• What is the difference between the models?

  The difference between the three currently available models, are subtle capabilities, how much text they can consume and the data they are trained on. As a general rule of thumb, the GPT-3.5-turbo is the fastest, cheapest and most reliable model. GPT-4 is the largest, most advanced and most expensive model. GPT-4-turbo is the most recent model, with a cut-off date in April 2023 and you can give it the largest amounts of text - 128.000 tokens vs 8000 in GPT-4.

• What are these new 'assistants'?

  The assistants feature enables you to create a customized assistant, that has a specific pre-prompt written by you to more easily complete specific tasks. All assistants are private upon creation. The assistant can then be shared with others in your team (or across Egmont) who are doing the same tasks if necessary.

• How can my assistant be in the 'Community'?

  The assistants that are available across Egmont, are general assistants that are relevant to the majority of users. If you have an assistant that you think would be great as a featured assistant, please reach out to us at datatech@egmont.com.

• How does web search work?

  Web search works by taking your prompt and re-phrasing it to a web search query. It sends this to an API that searches Google and retrieves the first ten results. It summarizes these and gives you a response based on the extracted results.

• Why is web search not secure?

  In order to enable web search, your prompt must be sent to a search engine. In this case, we use Google search, and hence what you search will be logged as when you search Google normally. Treat it, like you would a normal search engine.

• Can EgmontGPT access internal Egmont data?

  No, currently EgmontGPT does not have access to any internal data, such as emails, documents, policies etc, though we are exploring ways to integrate EgmontGPT with internal data sources.

• Why is EgmontGPT's GPT 4 not as fast as OpenAI's ChatGPT?

  EgmontGPT is built to be secure and private, which means we don't use OpenAI's ChatGPT 4 API directly, but a private instance deployed in Egmont's Azure cloud. Azures OpenAI instances has build-in extra security and content moderation features that are not available in OpenAI's API, at the cost of some delay in streaming responses. This means that EgmontGPT is not as fast as OpenAI's chat UI, but it is still fast enough for most use cases.

• What data is being saved?

  EgmontGPT saves the messages you send and the messages the model responds with, to provide you with a chat history for your convenience only. No data is used for training or model improvement. You can delete your chat history at any time, the individual chats by clicking the trash can next to the chat name or all chats via Settings -> Delete all conversations.

• What features are coming in the future?

  We hope to add more features in the future, making EgmontGPT even more helpful. Besides making minor updates to the user interface, we investigate the possibilities in introducing other Large Language Models than ChatGPT 4 to interact with. Also, we're looking into developing an API integration to EgmontGPT, to cater for using it directly in business processes.
  If you have any ideas or request for future developments, please let us know: digital.egmont-it@mail.4me.com.

• Who pays for all this?

  Egmont IT pays for the hosting of the models, chat UI and token usage. This means that you can use EgmontGPT for free, as long as you are an Egmont employee.

• What can I use EgmontGPT for?

  EgmontGPT is designed to be used for internal Egmont use only. This means that you can use EgmontGPT for any Egmont related tasks, such as writing articles, emails, presentations, etc. Try asking EgmontGPT how it can help with our tasks or visit https://www.aiforwork.co/ for inspiration.

• Does EgmontGPT support plugins?

  No, EgmontGPT does not support plugins; Code Interpreter etc.

• How do I switch models?

  On the frontpage, you can click the small gear on the model card. Here you can select any available model.

• How do I change the initial prompt to the model?

  On the frontpage, click on the small gear on the model card, and you can see the selected model. Here you can click 'System Prompt' and change the prompt for that model for you alone.

• How do I delete my chats?

  If you hover over any chat on the sidebar, a small trashcan shows. Click this and then the resulting checkmark and the chat will be deleted. If you wish to delete all chats, click 'Settings' and then 'Delete all conversations'. After confirming deletion, they will be gone permanently.

• How can I improve my prompts?

  Use: www.aiforwork.co/