Synapse Espresso: Using OpenAI GPT-3 in Synapse!

Welcome to the 22nd video in our Synapse Espresso series!
In this video, Thomas joins Stijn to show how we can make use of Azure OpenAI GPT-3 from within Spark in Synapse Analytics. GPT-3 is a natural language AI model and it's able to understand text that you feed into it and then generate now text based on your input.
For the integration between Synapse and Azure OpenAI, SynapseML is used. This open-source library simplifies the creation of scalable Machine Learning pipelines. One of the features it provides is the ability to send requests to Azure Cognitive Services using APIs. It's this feature that allows to easily access Azure OpenAI from Synapse Analytics.
The spark notebook showcases an example of sentiment analysis on restaurant reviews using GPT-3.
The reviews are loaded into a dataframe and the corresponding prompts are written, the SynapseML OpenAICompletion object is defined and finally the prompts are sent to OpenAI for analysis.
Related blog post: techcommunity....
SynapseML: microsoft.gith...
What is Azure OpenAI? learn.microsof...
Thomas Costers - Cloud Solution Architect Data & AI / thomas-costers-3442b059
Stijn Wynants - Fasttrack Engineer / sqlstijn
#azuresynapse #openai #chatgpt

[@PicaPauDiablo1]

I don't know how I missed this. Thank you!

[@avicool08]

Thank you for sharing this basic and great video how and what we can do with GPT3 using Synpase. Please share more ideas how we can use different kinds of datasets to analyze and give meaningful insights. Looking for more videos.

[@sdhilip]

Nice one. Can you share the code?

[@AndresPadillaAndrade]

Hello Synapse Espresso - great video - can you describe any other scenarios besides sentiment analysis or generating sql code or creating data sets that you could use with your synapse workspace?

[@stijnwynants7307]

Hi Andres, yes, we shared the data quality one and generating test data. more to come

[@myhst2712]

code doesn't work with gpt-35-turbo and gpt-4