The explosion of interest in AI due to the recent success of ChatGPT, the state-of-the-art natural language generation model that can write anything from essays to poems to code, is no surprise. However, now we are starting to see the excitement wane as ChatGPT usage numbers drop. This could be due to competition, concerns about privacy and security, or the overall excitement factor slowing down as users struggle to find uses of the tool. Further, to use the API available from OpenAI, you need a lot of technical skills and resources to train, fine-tune, and deploy it. You also need to be careful about the quality and safety of the generated text, as it might contain errors, biases, or harmful content.
The good news? This is just one tool in a sea of many other AI tools that are refined and purpose-built for many organizational needs. At the top of that list of tools is Microsoft’s Azure Cognitive Services tools, a collection of cloud-based APIs that provide ready-made AI solutions for various scenarios. Anyone who is familiar with Data Science and Machine Learning knows that we need troves of clean and trustworthy data to train an ML model to be able to predict results. The beauty of Cognitive Services is that Microsoft has already built these models around many categories, and even won many “human-parity” awards! Below are just a few examples of how Cognitive Services can help you with your AI needs:
• Speech Recognition: This service allows you to convert speech to text in real time or from audio files. You can use it for voice commands, transcription, dictation, captioning, and more. You can also customize it with your own vocabulary and acoustic machine learning models.
• Computer Vision: This service allows you to analyze and understand images and videos. You can use it for face detection, emotion recognition, object detection, optical character recognition, video indexing, and more. You can also create your own custom vision models using a simple interface. I recently created a video with an overview of the service here: https://youtu.be/ac8fvBWgUHg
• Text Analytics: This service allows you to extract insights from text data. You can use it for sentiment analysis, key phrase extraction, entity recognition, language detection, and more. Another example would be to use it to analyze healthcare documents and extract clinical information.
• Many more: Cognitive Services offer a wide range of services for different domains and scenarios, such as natural language understanding, conversational AI, anomaly detection, spatial analysis, personalization, and more.
You don’t need to worry about building or managing your own AI models, or if required, many of the services allow for custom models to be built as well. Once determined, you just need to connect to the API and start using it in your applications. Even better, many of the services can be containerized within a docker container where you can deploy the models locally or in other clouds for an even faster prediction for your application. Finally, you also get the benefits of Microsoft’s expertise and innovation in AI, such as high accuracy, reliability, security, and compliance.
To get started, many of the services have free service tiers for minimal transactions, and each service is billed based on consumption, so as long as you are controlling those transactions, you don’t have to worry about cost overruns, etc.
So, what are you waiting for? If you want to add AI capabilities to your applications without the hassle and complexity of ChatGPT and similar tools, Cognitive Services are the way to go! And if you really want to go deeper in understanding all the capabilities check out our recent book “Practical Guide to Azure Cognitive Services” from Packt or through other online book retailers: https://bit.ly/44NKm04
If you think this was valuable, or could improve, please leave a comment below and share it with your friends. And don’t forget to subscribe to my video blog https://youtube.com/bizdataviz for more Data and AI insights and tips.
Category Archives: Azure
3 Key Differences in ChatGPT and Azure OpenAI
In this vLog I discuss some misconceptions around ChatGPT and Azure OpenAI, to include:
- Who owns ChatGPT, OpenAI, and how Microsoft got involved
- Security and privacy concerns about Azure OpenAI and ChatGPT
- How each of the services is consumed and billed
Take a look to find out more!
| ChatGPT | Azure OpenAI | |
| Ownership | Owned by OpenAI LP a for profit arm of OpenAI the non-profit who’s mission is to development of society | Part of Azure AI offerings as APIs, and investor in OpenAI for exclusive rights to technology generated |
| Security | Insecure and open to the public. LLM is trained on dataset created in GPT 3 and currently only references data from 2021 and earlier. Questions and interactions are captured and can be used for further training | Secure to an Azure tenant using GPT-4, GPT 3.5, Codex and Dall-e Requires access for tenant to reduce the chances of the AI to be used for malicious purposes Data is not stored by the prompt, and model is not trained on data added |
| Costs | Free during preview stages, or paid for better hardware availability | Based on a consumption model like other Azure Cognitive Services. Biggest expense is re-training the model you’ve deployed for your data |
Azure Cognitive Services
AI solutions are exploding, and Azure has the most complete offering of any cloud provider! Watch this video to get started with our API based Cognitive Services in Azure and a sample architecture of how to employ them with the Azure Bot Service. Azure Cognitive Services are cloud-based services with REST APIs and client library SDKs available to help you build cognitive intelligence into your applications.
You can add cognitive features to your applications without having artificial intelligence (AI) or data science skills. Azure Cognitive Services comprise various AI services that enable you to build cognitive solutions that can see, hear, speak, understand, and even make decisions. Azure Bot Service enables you to build intelligent, enterprise-grade bots with ownership and control of your data. Begin with a simple Q&A bot or build a sophisticated virtual assistant.
https://docs.microsoft.com/en-us/azure/cognitive-services/what-are-cognitive-services
https://docs.microsoft.com/en-us/azure/cognitive-services/cognitive-services-apis-create-account?tabs=multiservice%2Cwindows
https://dev.botframework.com/
#Azure #AI #CognitiveServices #ArtificialIntelligence #Bots #ReferenceArchitecture #MachineLearning #API #Cloud #Data #DataScience
What is HTAP in Azure
Hybrid Transactional and Analytical Processing, or HTAP, is an advanced database capability that allows for both types of workloads to be performed without one impacting the performance of the other.
In this Video Blog, I cover some of the history of HTAP, some of the challenges and benefits of these systems, and where you can find them in Azure.
Overview of Azure Synapse Link featuring CosmosDB
Azure Synapse Link allows you to connect to your transactional system directly to run analytical and machine learning workloads while eliminating the need for ETL/ELT, batch processing and reload wait times.
In this vLog, I explain how to turn the capability to use Link on in CosmosDB, and what’s happening under the covers to give access to that analytical workload without impacting the performance of your transactional processing system.
Check it out here and let me know what you think!
Getting started with Spark Pools in Azure Synapse
In my latest video blog I discuss getting started on the newly Generally Available Spark Pools as a part of Azure Synapse, another great option for Data Engineering/Preparation, Data Exploration, and Machine learning workloads
Without going too deep into the history of Apache Spark, I’ll start with the basics. Essentially, in the early days of Big Data workloads, a basis for machine learning and deep learning for advanced analytics and AI, we would use a Hadoop cluster and move all these datasets across disks, but the disks were always the bottleneck in the process. So, the creators of Spark said hey, why don’t we do this in memory and remove that bottleneck. So they developed Apache Spark as an in memory data processing engine as a faster way to process these massive datasets.
When the Azure Synapse team wanted to make sure that they were offering the best possible data solution for all different kinds of workloads, Spark gave the ability to have an option for their customers that were already familiar with the Spark environment, and included this feature as part of the complete Azure Synapse Analytics offering.
Behind the scenes, the Synapse team is managing many of the components you’d find in Open-Sourced Spark such as:
- Apache Hadoop Yarn – for the management of the clusters where the data is being processed
- Apache Livy – for the job orchestration
- Anaconda – a package manager, environment manager, Python/R data science distribution and a collection of over 7500 open source packages for increasing the capabilities of the Spark clusters
I hope you enjoy the post. Let me know your thoughts or questions!
Connecting to External Data with Azure Synapse
In my latest video blog I discuss and demonstrate some of the ways to connect to external data in Azure Synapse if there isn’t a need to import the data to the database or you want to do some ad-hoc analysis. I also talk about using COPY and CTAS statements if the requirement is to import the data after all. Check it out here
The Modern Data Warehouse in Azure Part 4: The Serving Layer
In this video blog post I covered the serving layer step of building your Modern Data Warehouse in Azure. There are certainly some decisions to be made around how you want to structure your schema as you get it ready for presentation with whatever your business intelligence tool of choice, for this example I used Power BI, so I discuss some of the areas you should focus on:
- What is your schema type? Snowflake or Star, or something else?
- Where should you serve up the data? SQL Server, Synapse, ADLS, Databricks, or Something Else?
- What are your Service level agreements for the business? What are your data processing times?
- Can you save cost by using an option that’s less compute heavy?
Microsoft Reimagines Traditional SIEMs with Azure Sentinel
If you’re like most, security is at the forefront of your mind for your organization. You need the right tools and the right team to keep up with the balance of increasing number of sophisticated threats and with security teams being inundated with requests and alerts.
Today I’d like to tell you about Microsoft’s reimagined SIEM tool Azure Sentinel. Over the past 10 to 15 years, Security Information and Event Management (SIEM) has become extremely popular as an aggregation solution for security and events that happen in our network.
There are also software tools, hardware appliances and managed service providers that can help support your corporate needs to better understand the level of risks in real-time and over a span of time. They do things such as log aggregation, event correlation and forensic analysis and offer features for alerting, dashboarding and compliance checks.
These are great resources to help secure our environment, our users and devices. But unfortunately, the reality is security teams are being inundated with requests and alerts. Compile this with the noteworthy shortage of security professionals in the world – an estimated 3.5 million unfilled security jobs by 2021 – this is a major concern.
Microsoft decided to take a different approach with Azure Sentinel. Azure Sentinel provides intelligent security analytics at cloud scale for your entire enterprise. It makes it easy to collect data across your entire hybrid organization on any cloud, from devices to users to applications to servers. Azure Sentinel uses the power of AI to ensure you’re quickly identifying real threats.
With this tool:
- You’ll eliminate the burden of traditional SIEMs as you’re eliminating the need to spend time on setting up, maintaining and having to scale the infrastructure to support other SIEMs.
- Since it’s built on Azure, it offers virtually limitless cloud scale while addressing all your security needs.
Now let’s talk cost. Traditional SIEMs have proven to be expensive to own and operate, often requiring you to commit upfront and incur high cost for infrastructure maintenance and data ingestion. With Sentinel, you pay for what you use with no up-front costs. Even better, because of Microsoft’s relationships with so many enterprise vendors (and more partners being added) it easily connects to popular solutions, including Palo Alto networks, F5 networks, Symantec and Checkpoint offerings.
Azure Sentinel integrates with Microsoft Graph Security API, enabling you to import your own threat intelligence feeds and to customize threat detection and alert rules. There are custom dashboards that give you a view to allow you to optimize whatever your specific use case is.
Lastly, if you’d like to try this out for free, Microsoft is allowing you to connect to your Office 365 tenant to do some testing and check it out in greater detail. This product is currently in preview, so there may be some kinks but I’m looking forward to seeing how it develops in the future, as a true enterprise-class security solution for your environment, whether in the cloud, on premises, in data centers or remote users or devices.
Microsoft Announces Windows Virtual Desktop in Azure
Today I’m here with some exciting news out of Microsoft with the public preview of Microsoft Virtual Desktop. Virtual desktops are not a new invention and they are currently offered by multiple vendors.
Windows Virtual Desktop is comprised of the Windows desktops themselves and the application that you would pass out to users and the management solution for these are hosted in Azure. Inside public preview desktops and apps can be deployed on virtual machines in any Azure region in the US, with the management solution in data for these virtual machines residing in the US as well.
As the service moves closer to general availability, Microsoft will start to scale out the management solution and data localization to all Azure regions. Virtual desktops can be deployed with Windows 7 (with extended support) or Windows 10 for the workstation modes, and for the server versions you can run 2012 through 2019.
It will provide a full desktop virtualization environment inside your Azure subscription without having to run any additional gateway servers as you would if you were deploying this on other vendors’ platforms or in your on premises.
With Windows Virtual Desktop you can build custom images or pick some of the canned images provided in the Azure gallery. Images can be personalized and remain static with persistent desktops. There are also many configuration options, for instance; you want to run a single application for a connectivity for setting up a server/client type deployment of an application or you want to deploy pooled multi-session resources.
Another key point is you’re deploying this with significantly reduced overhead as you no longer need to manage the remote desktop roles like you would with remote desktop services on prem or with some of the other providers. You just have to manage the virtual machines inside of your Azure subscription.
There are many great use cases historically for virtual desktops in things like education and medical along with many others.