Event banner
Event details
Question to Azure OpenAI Platform team: Question to Travis or Allan: 5. How does Microsoft ensure data security and privacy when using GPT-4 in business applications, especially for industries like healthcare or finance? Are there compliance certifications built into the Azure deployment? 6. For someone just starting with this technology, what are the cost implications of using the GPT-4 API on Azure? Are there recommendations for managing costs while scaling up usage? 7. Are there any upcoming features for GPT-4 on Azure that could further benefit no-code developers? What should we be looking out for over the next few months? 8. What are some common mistakes non-programmers might make when setting up or using GPT-4 with Azure? Any tips on how to avoid them?
- nbrady
Microsoft
(5) Yes! Azure OpenAI Service is an enterprise-grade platform hosting the latest models from OpenAI. Check out our data, privacy, and security documentation for more information.
(6) Cost implications are use case dependent. Items that are likely to impact your costs include: Input/output audio ratio, average length of audio session, number of concurrent connections, deployment type, and throughput requirements. We recommend that developers start with small scale tests and development and evaluate costs before scaling to production.
(7) The "On your data" feature of Azure OpenAI is always a great way to get started quickly, then you can create a web app or a chatbot to share with your team or organization. Check it out!
(8) Common mistakes I've witnessed firsthand typically include:- not adding enough of your default quota to your deployment
- not spending enough time on refining the system message to specify instructions
- when building retrieval augmented generation (RAG) applications, folks don't spend enough time on their chunking strategies and investigating relevancy of the documents in the vector store.
I'm sure there are plenty of others you'll uncover as you begin your journey. Best of luck! - Travis_Wilson_MSFTFor getting started with no code, Azure AI Studio (https://ai.azure.com/) includes the new /realtime capabilities in its Playground experience (much like OpenAI's) and allows you to interact with the new gpt-4o-realtime-preview model in the browser without needing to write any lines of code -- it's also the best way to create and manage the model deployments on your Azure OpenAI Service resource. Low- and no-code development of /realtime-based experiences is an area we're actively looking into. The new API endpoint and its WebSocket-based capabilities are considerably more complex than the prior REST-based operations like /chat/completions, and making it as approachable and easy to integrate as possible (including with no code at all, where it makes sense) is a major priority for us when it comes to Developer Experience.
- I see you've posted this question twice - it looks like a complete duplicate. Please let us know which of these threads you want us to answer these questions in - I'll delete the other one.
- Broken it down in two parts.
