Comparing OpenAI ChatGPT - Microsoft OpenAI and Microsoft Cognitive Services
In recent years, Artificial Intelligence (AI) has become increasingly important in the business world, with companies looking for ways to automate processes and improve their customer experience. There are a number of AI tools available in the market, and some of the most popular are OpenAI ChatGPT and Microsoft OpenAI, as well as Azure Cognitive Services.
In this blog post, we will rely on the analysis of an article by Molly White to compare the cost of these three offerings and determine which one may be the best fit for your organization's needs and budget. We will provide insights into the different pricing models, features, and capabilities of each tool, giving you a clear understanding of the costs involved and helping you make an informed decision.
For our test we want to have each AI solution return the following for the article:
- Article Author
- Article Categories
- Article 100 word Summary
- Article detected language
Using ChatGPT is simple, simply navigate to https://chat.openai.com/ in your web browser, and create an account and log in.
We use the following prompt:
Prompt: Determine what language the following content is written in, who the author of the following is, the top three categorizations, and a 100 word summary Content:
This prompt is followed by the content of the article (covering the explanation of the bankruptcy hearing regarding the companies Genesis and Gemini).
When we run the prompt we receive this result. Basically, this response is perfect and demonstrates the power of the OpenAI GPT model.
At this time this article was written OpenAI does not have an API to call the ChatGPT model, nor do they indicate the number of tokens, used for determining the cost.
However, their most advanced model, at the time of this writing, is $0.02 per 1000 tokens.
Using the OpenAI documentation, we can see that the prompt should not be more than 1000 tokens so the cost, using the old model, would be about $0.02
Using Azure OpenAI
Next, we will run the same example using Azure OpenAI. Azure OpenAI is a Microsoft cloud platform using OpenAI models (See: Comparing Azure OpenAI and OpenAI).
To use it, you first apply for access to Azure OpenAI Service at https://aka.ms/oai/access, then you can then go to: portal.azure.com and Create a resource.
Search for the Azure OpenAI service and click Create.
You can then get to the Azure OpenAI playground at this link: https://oai.azure.com/portal/playground
If this is your first time using the playground you will be prompted to deploy a model. Select and deploy the text-davinci-002 model.
After it is deployed, navigate to the GPT-3 section.
This content is written in English and the author is Molly White. The top three categorizations are business, finance, and cryptocurrency. In brief, the article discusses the recent bankruptcy hearing of Genesis, a cryptocurrency company, as well as the legal troubles of Gemini, another cryptocurrency company
When we run the exact same prompt we receive this result. We see that this query consumes 585 tokens.
We can then consult the pricing table and we see that the cost of this query would be approximately $0.01
Using Azure Cognitive Services
Azure Cognitive Services is a collection of pre-built APIs that allow developers to add intelligent features to their applications. It includes a range of AI services for tasks such as speech recognition, language understanding, computer vision, and more. With Cognitive Services, developers can integrate AI into their apps without the need for extensive technical expertise or resources.
We can get to the playground at this link: https://language.cognitive.azure.com/home
Microsoft Cognitive Services did not have the ability to determine the article author and the article categories, but we can use a call to the Summarize information API to detect the language and get a summary.
Rather than producing an original summarization of the article, this service just extracts key sentences.
To get a unique summarization you will have to apply for access to call that API (at the time of this article I have applied, but not received that access).
We can then consult the pricing table and we see that the cost of this query would be approximately $0.004
The FAQ pricing page offers a further explanation of the method to determine what a Text record is when determining pricing.
Note: while Microsoft Cognitive Services is more limited than the other OpenAI GPT models, it does offer unique services such as: document and conversation summarization.
- The GPT models from OpenAI produce unique results at a cost
- Azure Cognitive Services cannot match the features, but what they do offer is a fraction of the price
Calling OpenAI GPT-3 From Microsoft Blazor
Analyze text with the Language service
Announcing the new beta release of Azure Text Analytics libraries
Review the quick start documentation
Review the responsible AI documentation and requirements