Azure OpenAI helps customers accelerate innovation with large AI models; Microsoft is expanding availability

by Vanessa Ho

Customers who buy a used car can sometimes feel overwhelmed by searching many car specifications and reviews, but CarmaxAnd The largest used car retailer in the United States, makes it easy for customers to access even more useful information.

Thanks to powerful AI language models, potential buyers can now see summaries of customer feedback for every make, model and year of vehicle the company sells, nearly 5,000 combinations in a massive inventory of nearly 45,000 vehicles. Summaries provide easy-to-read key information from real customer reviews: whether it’s a great family car, how comfortable the ride is, or whether there’s enough room in the luggage compartment for weekend adventures.

CarMax has also used templates to create new site content that allows customers to easily see what’s new in each vehicle version, helping them decide whether the new features are worth it.

company Produce tons of original content In just a few months – a rate that was previously impossible – using powerful natural language models GPT-3 built by the company open ai. To extend the work, CarMax now uses a file Azure OpenAI ServiceAnd Which combines access to OpenAI templates and Azure enterprise-grade features, such as security, compliance, and regional availability.

“Through the Azure OpenAI service, we are able to create content that enables our customers to be well informed before making a decision,” commented Shamim Mohamed, Executive Vice President and Head of Information and Technology at CarMax. “The ability to innovate at scale, the ability to make the customer experience easier, the ability to experiment — all of these things are possible because of this service and the partnership we have with Microsoft.”

In the build going on today, Microsoft announced that the Azure OpenAI service is available from a limited-access preview. Customers who wish to use the service can Entry request. Introduced at Ignite 2021 as a A new product in the Azure Cognitive Services familyAnd Part of Azure AI, the Azure OpenAI service was previously only available to guests.


“Through the Azure OpenAI service, we are able to create content that enables our customers to be well informed before making a decision.”


The service has a new responsible artificial intelligence system that filters out malicious content and helps detect abuse. In addition, Azure service open ai Now allows access to more models, including GPT-3 models, manuscript and implication. Codex can generate code and translate simple language into code, while embeddings make semantic searches and other tasks easier. The service also provides new capabilities for customers to modify forms for more personalized results.

Providing customers with Azure guarantees and promises

Azure OpenAI Service enables customers in all industries, from healthcare to financial services and manufacturing, to quickly accomplish a range of tasks. Innovations include creating unique content for customers, summarizing and categorizing customer feedback, and extracting text from medical records to simplify billing. The most common uses were to aid in writing, translate natural language into code, and gain insights from data through research, entity extraction, sentiment, and classification.

Azure OpenAI Service helped Farmlands, a rural supply cooperative in New Zealand, to summarize and categorize customer interactions as ‘neutral’, ‘negative’ or ‘positive’, and extract keywords to gain insights into customer trends faster. Image courtesy of Farmlands.

“One of the interesting things is the variety of use cases that can be supported outside of a single model,” said Eric Boyd, Vice President of Azure AI at Microsoft. “The Azure OpenAI service is really leading the way with these big new language models and giving customers the assurances and promises of Azure that this will be reliable and secure, and their privacy will be protected, as they explore these amazing frontiers of what is possible with these new technologies.”

Through the OpenAI API and Azure OpenAI Service, CarMax used GPT-3 to abstractly summarize 100,000 customer reviews and tune them into 5,000 well-written summaries. Kevin Hopwood, the company’s software engineer, said the work would have taken 11 years for the editorial team at CarMax.

He commented, “Summaries and other content generated by templates improved customer engagement and search engine optimization, while the time saved allowed CarMax content creators to focus on deeper research, longer articles, and more creative tasks.”

“The best thing we can do is free up their time so they can explore new content ideas and new ways to engage customers,” Hopwood said.

Azure’s security, compliance, and reliability features and other enterprise-grade features will allow CarMax to extend its use of GPT-3 to cases where millions of keywords are required. The model’s ability to learn from just a few examples of the intended output, a process called low-snap learning, will help the CarMax team of 60 products to use the models without the need for additional teams of data scientists.

“As an integrated service for Azure, this tool puts a great deal of power into our traditional engineers with Microsoft C# they didn’t have before,” said Sean Goetz, Director of Application Systems. Give Carmax. “We can extend it to other teams, just like any other Microsoft toolkit.”

Building systems to support responsible AI

The power of GPT-3, pre-trained with a huge amount of text on the Internet, comes with the risk of generating malicious or unintended results. Microsoft has made significant investments to help protect against abuse and unintended damage, which includes requiring candidates to demonstrate and integrate well-defined use cases. Microsoft Principles for the Responsible Use of Artificial Intelligence. An important way that CarMax and other customers meet standards is by having humans in the flow to ensure that model outputs are accurate and meet content standards before they are published.

The new responsible AI system built into the Azure OpenAI service can help filter out sexually relevant, violent, hateful or self-harming content. The team plans to add additional filters and personalization features as they work with customers during the preview period and see what is practically needed..


We built the system not to say we filter everything and have the right answer, but to get that feedback and adapt quickly.. “


The filtering system also identifies potential patterns of abuse or unintended harm, in which case Microsoft will work with customers to investigate, respond to, and block, if necessary, any abusive user. The Incident Response Team is available to quickly update content filters as the language evolves. For example, if a new racial insult appears, the team will respond quickly and ban the term as hate speech.

“We built the system not to say we filter everything and have the right answer, but to get that feedback and adapt quickly,” said Sarah Bird, Microsoft’s lead AI for Azure AI. “We also do not judge whether or not it is appropriate for users to post certain content online. We say that we do not believe that our AI system should generate this type of content.”

To help customers build robust systems, Microsoft also provides User Experience Design Guidelines and Standards, and a Transparency Note that defines the Service’s boundaries, uses, and intended characteristics.

“We’ve thought very well about the tools we’re making and how they’re going to be used, and we’re really trying to make sure they’re used responsibly so that the benefits of this technology can benefit everyone,” Microsoft said. Boyd.

The Principles of Responsible Artificial Intelligence has been a helpful guide FarmlandAnd New Zealand’s largest rural sourcing cooperative, where human auditors also work with Azure OpenAI to improve customer service and create new content for their website.

Farmlands, New Zealand’s largest rural supply cooperative, used Azure OpenAI Service to summarize customer interactions so employees don’t have to put their customers on hold and spend precious minutes scrolling through feedback to catch up. Image courtesy of Farmlands.

Farmlands used the OpenAI API and Azure OpenAI Service to condense 350,000 customer interactions into short summaries that help call center partners quickly understand a customer’s situation and respond effectively. In the past, colleagues had to scroll through emails and status notes while keeping the customer waiting. Summaries can capture minute details like sarcasm and complexities like a case involving 11 interactions with a customer who received the wrong fuel.

“This saves time for our call center staff and improves our customer experience,” said Gareth Poulard, Director of Farmlands Insights and Analytics.

The retailer also uses the service to extract keywords from interactions and categorize interactions as ‘positive’, ‘neutral’ or ‘negative’. She explores an in-house tech chatbot and creates product descriptions for her e-commerce site. He said that using the Azure OpenAI service means that engineers using the full Azure package will be able to more easily gain data insights into customer trends, while expanding their use of model-generated content across the enterprise.

“In terms of content generation, it’s a game changer,” said Jan van Schalkwyk, Business Intelligence Lead at Värmlands. “It’s a very cool technology.”

Related links

Tags: Azure, Azure OpenAI, Artificial Intelligence

Leave a Comment