Howspace & Azure Open AI
Updated over a week ago

Howspace now allows selected organisations to choose between "classic AI" and the new "Microsoft Azure Open AI" engine.

By enabling OpenAI engine you'll get superior results in summarization and can start exploring conversations by asking questions from the AI via prompting.

What is the "Azure OpenAI engine"?

Howspace platform is hosted on Amazon cloud service and can leverage Microsoft's Azure OpenAI API services. This service is located outside Amazon cloud.

You can find more details about Azure OpenAI service here:

What are the benefits of enabling "Azure OpenAI engine" inside Howspace?

You will experience significantly improved summarization and gain access to exploring dialogue through "prompting".

Prompting means you have e.g. chance to ask questions about superchat contents, such as "What could be the top 5 initiatives in the dialogue" and AI will give you an incredibly good list of answers.

How to enable the OpenAI engine in Howspace?

Under workspace settings (features) you need to tick the "Allow alternative engine selection". After saving (and possibly refreshing the page) you can click the "AI engine" selection and choose "OpenAI" from there.

What are the limitations of the OpenAI engine?

We use generative models (similar to "chatgpt") and they have their limitations. For example the gpt3.5 model struggles to understand certain questions such as "who mentioned 'transparency'" or prompts about "counting something". AI can generate unrelated answers. GPT4.0 model has better quality. Different model versions are being tested inside Howspace.

Generative models can make mistakes.

What happens to "Howspace classic AI engine" in the future?

Our "classic AI engine" will be retired and completely replaced by OpenAI engine during 2024-2025.

How is Howspace AI different from OpenAI's ChatGPT?

In terms of "AI models" we use (almost) the same models so there isn't much practical difference in model quality.

Howspace provides context differently (e.g. for example "summarization of a superchat discussion" uses "superchat comments" as context while for chatgpt you'd have to copy & paste the data and write your own prompts).

Privacy is a factor too: Public OpenAI aka "chatgpt" may use some data for training their models. Howspace AI uses private service (Azure OpenAI API) and therefore no data is used to train models.

What does it means that data is transferred to Azure OpenAI service?

Howspace data (such as "superchat comments") are sent to our Azure OpenAI service from our Amazon cloud environment.

Data is not stored in Azure.

Where is Azure OpenAI data transferred or stored?

Data is transferred to Europe (Sweden) or Canada, but not stored.

  • Howspace Canada services (ca.howspace) will transfer data to Azure OpenAI endpoint that is located in Canada.

  • Howspace European services (in.howspace) will transfer data to Azure OpenAI endpoint that is located in Europe (Sweden).

According to Microsoft the data is not stored in the model: "The models are stateless: no prompts or generations are stored in the model."

What data is being transferred to Azure OpenAI?

At the time of writing (November 2023) we'll process "superchat comments", "superchat authors", "prompted questions" in Azure OpenAI. The data that's being sent can (and will) change as the development progresses.

Is Howspace data used to train Azure OpenAI models?

No.

According to Microsoft Azure's service description, they do not train models with sent data and therefore Howspace data is not used for further training their models.

How Azure OpenAI uses data that's being sent there?

From Microsoft Azure security/privacy page:

Your prompts (inputs) and completions (outputs), your embeddings, and your training data:

  • are NOT available to other customers.

  • are NOT available to OpenAI.

  • are NOT used to improve OpenAI models.

  • are NOT used to improve any Microsoft or 3rd party products or services.

  • are NOT used for automatically improving Azure OpenAI models for your use in your resource (The models are stateless, unless you explicitly fine-tune models with your training data).

  • Your fine-tuned Azure OpenAI models are available exclusively for your use.

The Azure OpenAI Service is fully controlled by Microsoft; Microsoft hosts the OpenAI models in Microsoft’s Azure environment and the Service does NOT interact with any services operated by OpenAI (e.g. ChatGPT, or the OpenAI API).

How long will data remain at Azure OpenAI service?

Data will not remain at Azure.

It says on Microsoft's data privacy document that:

"The models are stateless: no prompts or generations are stored in the model."

Based on this we assume no data is stored long-term in their services (with the exception of possible abuse monitoring if content is flagged harmful).

What GDPR considerations there are?

Howspace is sending first name, last name and person's comment (when discussion is not anonymous) to Microsoft Azure OpenAI for processing.

It says on Microsoft's data privacy document that:

"The models are stateless: no prompts or generations are stored in the model."

Therefore no data remains at the Azure service after processing has been completed.

Did this answer your question?