Dude, that intro is epic. I like it. At first thought you were twins, but then I focused on finding a shadow between both sides and where it cuts off, and I found it. The dialogue alone sold me at first.
1:15 into this video and I have no idea how good it is going to be at the end, but man can you act. The synchro is un believable. The best I've seen of people doing this kind of videos.
Thanks MG! The are very informative and helpful. What is often left out of videos is the fact that many developers want to create SaaS platforms using the techniques you present. This generally involves a multi-tenant / user scenario. Yes this adds complexity, but it would be great if you could elaborate on siloing customers data in a saas model approach using cognitive search and openAI.
What a great video! I have one question regarding embeddings, I found current embeddings model is not working well in handling document with tables, do you have any ideas how to handle tables?
Hi, thanks for the nice video.. It would be great if you prepare a end to end tutorial video explaining the codes to develop such application from scratch.
I'm totally new to microsoft Azure, so which Azure's products (storage accounts, database,... ) should I choose in order to process this kind of model, and could you share the approximately price to run the model on monthly basis
I love this video. However, please be careful - the way this is implemented and the way the methods are used results in some major problems. 1) There are some (surprisingly high) costs - I have spent 60$ on the search alg in the background of the site application that gets deployed on Azure. This has happened with less than 20 documents with less than 2000 tokens uploaded. There is no justification for such a cost other than the implementation not being that great or the algs/chains in the background being needlessly complex. 2) The control you have over the little, but important parts of your implementations is not optimal. Examples and history have to be manually given to the prompt, you have no power over the way in which the files are sliced when the embedding is happening, you have no control over the size of the docs being retrieved from your index (dataset of text slices). 3) The flexibility is not that great and you need to go the extra mile to change the models used (for translation, embedding and chat), you do not know how many instances of them are running in the background, nothing. I know this is extremely convenient, but I think it is cheaper and overall better to spend three days on learning LangChain than it is to depend on this solution - automatic example generation and selection, easy to use prompt templates, different kinds of history generation that can be combined into one that culminate into one great solution for building ChatBots.
you need to set up an azure account, it is a hell and should take 2-3 working days. After half a day, I still cannot figure out where to find/define the key access to the blob storage service...
Open your storage account in the portal, click on the key on the left side and you should be able to see the key. if you already have azure evnoremnt with Open AI, it should be just several minutes to deploy this solution
Thanks a lot for the awesome video. I have successfully deployed it. I also added one document. It has not computed the embeddings yet. Any idea on how long does that take? I used Bank of America report, that was used in the video..
How to deal with general questions related to my own documents ? Example: suppose i upload documents of a meeting transcription, and I my questions are: "how many people asked something during the meeting?" or "who many people participated in the meeting?" I understand that the embedding solution is good for specific questions, but my doubt is, can it deal with those type of general questions?
Is it possible to integrate ChatGPT with my existing database that contains around 40K products? I'm planning to create a chatbot using ChatGPT which can intelligently respond to user queries based on our current product availability. I'd appreciate any guidance or tips on how to achieve this.
OpenAI has private trained model for enterprises. Does it have Role-based access control for the Privately trained data in OpenAI? (like employees can access only their data; another employee cannot access his/her data)
Sent application to get the Azure Chatgpt permission. They expect a lot of company information, including company website, company phone number etc... Seems like they don't want anyone to just be playing around with it.
The questions look less open than what is possible via chatgpt3 or chatgpt4. It's as if you had to look for a direct quote from the text. Should we deduce that these data uploaded on the clood do not allow to ask questions as freely as on chatgpt itself ?
The questions were just for demo purpose but you can certainly question much more complex scenarios as you ask from ChatGPT. also check out Part 2 : th-cam.com/video/RcdqdWEYw2A/w-d-xo.html
Thanks a lot MG! One issue I am facing after setup environment and upload documents; for Index I am getting error: Traceback (most recent call last): File "/usr/local/lib/python3.9/site-packages/openai/api_requestor.py", line 672, in _interpret_response_line data = json.loads(rbody) File "/usr/local/lib/python3.9/json/init.py".... it is long list of error. What I am doing wrong?
Great video. I had a question. Does exposing your data to chatgpt make it public? Is it possible for me to not allow access to my data and yet train it using chatgpt algorithm? Much appreciate a response..
Azure open ai fine-tuning is still unavailable, did anyone facing this issue, plz help me how to resolve this? My resource group region is South Central US not able to do custom fine-tuning..
@@MG_cafe Thanks for this link. I'm building enterprise grade solutions(where scalability matters) to the customers. Should i go with (Cognitive search+chatgpt) combination or (redis + chatgpt) combination or (langchain+chatgpt) combination ?
Great video! Is it possible to identify the source document the answer refers to? How?
yes you can and here I have explained how in part 2 : th-cam.com/video/RcdqdWEYw2A/w-d-xo.html
⁷😊😊😊😊😊😊😊😊😊😊😊😊😊😊😊😊😊😊
Dude, that intro is epic. I like it. At first thought you were twins, but then I focused on finding a shadow between both sides and where it cuts off, and I found it. The dialogue alone sold me at first.
haha, you made my day, Thanks so much buddy!
1:15 into this video and I have no idea how good it is going to be at the end, but man can you act. The synchro is un believable. The best I've seen of people doing this kind of videos.
MG, I'm so proud of you as an Iranian!
What are the operating costs of Azure? I've seen some implement embeddings with langchain & pinecone, even though your method seems simpler.
Thanks MG! The are very informative and helpful. What is often left out of videos is the fact that many developers want to create SaaS platforms using the techniques you present. This generally involves a multi-tenant / user scenario. Yes this adds complexity, but it would be great if you could elaborate on siloing customers data in a saas model approach using cognitive search and openAI.
What a great video! I have one question regarding embeddings, I found current embeddings model is not working well in handling document with tables, do you have any ideas how to handle tables?
Thanks so much , please check out the video I posted : chat with your tabular /SQL data
Hi, thanks for the nice video.. It would be great if you prepare a end to end tutorial video explaining the codes to develop such application from scratch.
Amazing intro, that summarizes the problems we are currently having about chatgpt
Please make a video on how to do this locally in Python with Conda or venv. Thanks for the great video btw.
I'm totally new to microsoft Azure, so which Azure's products (storage accounts, database,... ) should I choose in order to process this kind of model, and could you share the approximately price to run the model on monthly basis
you dont host the model. Microsoft will host and handle everything and you just call these models through API and get charged for it
killing ending.. salute your motivation
Thank you. Is there a similar way to solve the exact problem on AWS?
thanks for the information. Is it only open for Company accounts? I was stuck in "Request Access to Azure OpenAI Service"
Really good. How can we display the source of the document which contain the answer ?
Please check out part 2
@@MG_cafe I did. But I meant, without azure search 😉
Very well done, congratulation ! Very high quality presentation and information ! 🙏❤🙏
I love this video. However, please be careful - the way this is implemented and the way the methods are used results in some major problems.
1) There are some (surprisingly high) costs - I have spent 60$ on the search alg in the background of the site application that gets deployed on Azure. This has happened with less than 20 documents with less than 2000 tokens uploaded. There is no justification for such a cost other than the implementation not being that great or the algs/chains in the background being needlessly complex.
2) The control you have over the little, but important parts of your implementations is not optimal. Examples and history have to be manually given to the prompt, you have no power over the way in which the files are sliced when the embedding is happening, you have no control over the size of the docs being retrieved from your index (dataset of text slices).
3) The flexibility is not that great and you need to go the extra mile to change the models used (for translation, embedding and chat), you do not know how many instances of them are running in the background, nothing.
I know this is extremely convenient, but I think it is cheaper and overall better to spend three days on learning LangChain than it is to depend on this solution - automatic example generation and selection, easy to use prompt templates, different kinds of history generation that can be combined into one that culminate into one great solution for building ChatBots.
Is langchang able to run private? I believe azure openai is HIPPA compliant
Is this process suitable for sensitive data that should not be shared?
I tried to deploy, all went well until I tried to use it. It just kept throwing errors at me.
double-check if you have the models deployed in your Azure OPen AI service with the same name chosen oni form deployment
you need to set up an azure account, it is a hell and should take 2-3 working days. After half a day, I still cannot figure out where to find/define the key access to the blob storage service...
Open your storage account in the portal, click on the key on the left side and you should be able to see the key. if you already have azure evnoremnt with Open AI, it should be just several minutes to deploy this solution
what type of subscription do you have? what is the minimum subscription needed to do this?
Thanks a lot for the awesome video. I have successfully deployed it. I also added one document. It has not computed the embeddings yet. Any idea on how long does that take? I used Bank of America report, that was used in the video..
Really informative, I would like to use azure sql instead of redis. is it possble?
Is it more advantageous to utilize GPT-3.5 instead of Davinci for improved performance?
From my own experience, yes.
@@xvmx.mp3 i tried to deploy GPT3.5 on OpenAI engine but did not work
I subbed because of the message at the end
you made my day, thanks so much!
How different is this from the Add your data option in Azure Open AI that is available?
How to deal with general questions related to my own documents ? Example: suppose i upload documents of a meeting transcription, and I my questions are: "how many people asked something during the meeting?" or "who many people participated in the meeting?"
I understand that the embedding solution is good for specific questions, but my doubt is, can it deal with those type of general questions?
Is it possible to integrate ChatGPT with my existing database that contains around 40K products? I'm planning to create a chatbot using ChatGPT which can intelligently respond to user queries based on our current product availability. I'd appreciate any guidance or tips on how to achieve this.
Yes it is possible . Please check chat with SQL data video I uploaded
if i want to convert one programming language to another will this method useful?
Great Stuff!!
OpenAI has private trained model for enterprises. Does it have Role-based access control for the Privately trained data in OpenAI? (like employees can access only their data; another employee cannot access his/her data)
Great Video Thanks ! Can I use the question part as an API for my application ?
Great stuff, whats the possibility of using this as an api using power app so instead of using stand alone application
I’m what ways is this different from connecting Cognitive Services Search to Azure OpenAI?
Great video! Thank you so much.
Sent application to get the Azure Chatgpt permission. They expect a lot of company information, including company website, company phone number etc... Seems like they don't want anyone to just be playing around with it.
Will this work with CSV files as well?
Excellent video.. Thank you!!
The questions look less open than what is possible via chatgpt3 or chatgpt4. It's as if you had to look for a direct quote from the text. Should we deduce that these data uploaded on the clood do not allow to ask questions as freely as on chatgpt itself ?
The questions were just for demo purpose but you can certainly question much more complex scenarios as you ask from ChatGPT. also check out Part 2 : th-cam.com/video/RcdqdWEYw2A/w-d-xo.html
Thanks a lot MG! One issue I am facing after setup environment and upload documents; for Index I am getting error: Traceback (most recent call last): File "/usr/local/lib/python3.9/site-packages/openai/api_requestor.py", line 672, in _interpret_response_line data = json.loads(rbody) File "/usr/local/lib/python3.9/json/init.py".... it is long list of error.
What I am doing wrong?
Can WE use sharepoint as data source?
Yes , check this out : th-cam.com/video/IcwbjD_nXG4/w-d-xo.html
Great video. I had a question. Does exposing your data to chatgpt make it public? Is it possible for me to not allow access to my data and yet train it using chatgpt algorithm? Much appreciate a response..
If you’re using the API they don’t store data or train their base model with your data… unless you opt in to allow them.
Azure open ai fine-tuning is still unavailable, did anyone facing this issue, plz help me how to resolve this? My resource group region is South Central US not able to do custom fine-tuning..
謝謝！
Thanks so much !
Great video! Is it possible to feed data to chatGPT via API?
This is what this demo does
How can I set this up locally? I don’t wish to use any cloud service. Pl suggest
in the same github, try the local deployment option
I can do this with a CSV file that has problem & solution columns?
definately you can
Could this be used for large data sets from a spreadsheet?
Can people outside of enterprise customers still get access to Azure OpenAI?
Is it possible to feed openai with search console api data of my site and get seo insights from it ?
Hello, is it free to use ? i have azure but i do not know how much that will cost me. can i deploy it in local ? if yes is it free ?
You need to have azure open AI service deployed which is not free and it will be charging based as you go usage
बाते अच्छी हैं लेकिन विज्ञापन भययुक्त है।
Amazing 😍 any updates?
How secure is this?sharing data with OpenAI?
Is there a way to do website links?
Which website?
hii their it was great exploitation , I have a doubt where dose the embeddings store and will we able to access them
Can I use chatgpt with cognitive search ?
Absolutely you can ! :)
@@MG_cafe is it like we use cognitive search to get all the relevant docs and then pass it on to chatgpt for completion of the query ?
@@ashishjohnsonburself Exactly! Check this put :
github.com/Azure-Samples/azure-search-openai-demo
@@MG_cafe Thanks for this link. I'm building enterprise grade solutions(where scalability matters) to the customers. Should i go with (Cognitive search+chatgpt) combination or (redis + chatgpt) combination or (langchain+chatgpt) combination ?
@MG if my question doesn't make sense, please do let me know. Thanks
⚠️ I'd like ChatGPT read all my book's library, so that I could research easily.
certainly possible :)
Ive been trying to convert data into a table, less than one page of data, Chatgpt only does less than half, not worth using it at this stage
GPT4 has higher token limit which can help further
I thought what you’re doing here is not limited to the token size?
You guys are hilarious.
its document embedding not word embedding
.......durgaparamesvari devi.......
😊
Aka give your data to the AI machine
iranians?
Hilarious
Hi, your video is awesome! I need this for my company. Do you freelance? Pls contact me if you do. Thank you.
Please send me an email , happy to help you : mo.Ghodrati95@gmail.com
@@MG_cafe ok I’ll email you. Thank you!