Azure AI Search: Generating the right answers, every time

Chapters
00:00 Welcome to the AI Show
00:18 On today's show
00:46 Azure Cognitive Search is now Azure AI Search
01:07 Semantic search is now Semantic ranker
01:56 Azure AI Search Announcements from Ignite
03:08 Semantic ranker pricing
04:22 Integrated vectorization
06:37 Demo
18:20 Learn More
On this episode, we'll dive into the future of search with Azure AI Search:
🌐 Revolutionary vector database technology
🔍 Cutting-edge search ranking for superior user experiences
🔄 Advanced retrieval capabilities: vector search, hybrid search, semantic re-ranking
🤝 Deep platform integrations for seamless AI app experiences
Learn more:
Aka.ms/AISearchIgnite2023
Connect:
Seth | / sethjuarez
Watch on-demand at aka.ms/AIShow/OnDemand
AI Show Live Playlist: aka.ms/AIShowPlaylist
Join us every Monday, for an AI Show livestream on TH-cam aka.ms/AIShowLive

[@KarlJuhl]

Seth you're so good at pointing out all the useful things and summarizing the slightly complicated parts 👏🏼👏🏼

[@machevellian79]

Fascinating thanks folks!

[@yusufkemaldemir9393]

is it just for pdf files in blob? how about sharepoint files?
how do you set up the credentials to start with.

[@crow-gf8hr]

Great video thanks. I saw a similar question on another video. Can you integrate rating for the generates answers of your model with Azure AI search? Thumbs up and down and feed this back to the model? I am not talking about the playground but after I deployed the model.

[@rahulghosh4031]

Indexer: Could not execute skill because Web Api skill response is invalid (exercise 5, task 3.22). I am getting this error while trying to run the integrated vectorization python program. Seems some issue with the AzureOpenAIEmbedding skillset. The error is coming due : Error while sending the Request
Please help me out with this

[@chinny4953]

Great stuff guys

[@pranavbilurkar2805]

Any github repo, to refer the code?

[@seatconnect224]

how does chunking performed? By fixed amount of tokens or by semantic meaning of each chunk?

[@MultiMohsinkhan]

Just quick question on Azure AI Search, I have around 1 TB of document data, each document is approx 8MB. During enrichment it tooks 2 minutes for a document seems it will take year to complete the indexing any suggestion on this ?

[@virupakshaguptha.h.a8453]

one question - When you delete few paragraphs present in documentation, then those exact chunks present in azure AI search should also be deleted (is this taking care).
second question - When you update few paragraphs in documentation, those those exact chunks should be updated, or old chunks should be deleted and new have to be created
How well indexer is handling this? I have been asking this question to various Microsoft developers in AI search no one is answering

[@MiSt3300]

Implementing this is .NET is an absolute nightmare, you guys left absolutely NOTHING in the documentation, no examples, nothing. I am getting error 400 bad request when trying to insert data into the vector DB all the time, and if I serialize my chunk into json like it tells me, it throws deserialization errors. What kind of software is that?

[@procexsolutions]

How do you get access to the prerelease of search documents 11.4.0b12?

[@rahulghosh4031]

Copy the file from whl folder under python-demo

[@johnwhite7218]

Sounds a bit like “Semantic Rancor” which is what I do when my code doesn’t work

[@daviduhlmann8119]

I would really admire if MS stops changing the names of their products for lets say the next 3-5 years. This would make everyone's life so much easier.

[@pn1302]

But then they'd have to get rid of the Product Renaming department and lose the 300 people who work there.

[@Momentum_Option_Buyer]

haha..Agreed..and their product names are so similar that they are hard to differentiate ex: Azure AI Studio, Azure Open AI, Azure AI Service.