We already did a detailed review of machine learning services and tools provided by each company from the grand IT trio. To sum it all up in one article, we decided to do a quick comparison of the main ML tools offered by Amazon, Microsoft, and Google. We believe it is necessary for everyone who is preparing to make their final decision and choose one provider to start a fascinating journey into the world of machine learning.
All these companies are market leaders not only in the machine learning segment but in all other IT departments as well. Needless to say that they compete with each other on all fronts trying to invent better, faster, and more affordable products. Their approach to machine learning may differ and can be truly unique – remember Amazon DeepRacer League, Azure Machine Learning Studio, and Google AutoML? But the majority of available options and APIs do exactly the same – translations, text analysis, image recognition, and so on. To save you from spending too much time on discovering all the details about the machine learning services of each provider, we wrote this summarizing article that highlights the main strengths of each company’s solutions and compares them with each other.
All three have a full set of machine learning features available for use as a service (MLaaS). Amazon is known for launching services like AWS Rekognition that does decent image recognition, AWS Polly for transforming the text into speech with the help of deep learning algorithms. Still, the most popular tool in their arsenal is definitely Amazon SageMaker, which is designed to simplify the process of creating, training, and deploying machine learning models.
As for Microsoft, with the machine learning services packed in Azure, they are aiming to make ML more accessible for everyone. And if we’re to judge by the latest versions of Visual Studio, we can say that they are doing it pretty well. We absolutely love the graphic UI that allows building ML models without extensive knowledge or experience with the subject. Also, some companies have reported that this tool significantly reduces the time required to train and deploy accurate models based on the pre-loaded data.
Google Cloud advertises its APIs and machine learning services as one of the main advantages. Their range of various ML and AI-based products, including AutoML and Machine Learning Engine, looks truly impressive, allowing users to build and run superior ML models with ease. Regarding AI tools, Google offers access to its AI Hub, which is a hosted repository that contains various plug-and-play AI components. For example, end-to-end AI pipelines, algorithms that can be used right-of-the-box, AI building blocks, and more.
Most Notable Tools and Services
So, all these corporations are confident in Machine-Learning-as-a-Service. Despite the fact that some started earlier and had a major advantage in accumulating the customer base (AWS), while others had to break seven sweats to catch up (Google Cloud), we can say that all companies are in equal conditions when it comes to machine learning. This parity happened because machine learning is still a new field and the companies added ML services to their range of products not so long ago, in 2015.
Let’s check what they have achieved in the past four years and what techs did they bet on. Bear in mind that it is not a comprehensive review of all available services but merely a glimpse at the most popular product(s) on each side. For in-depth analysis of machine learning services by Amazon, Google, and Microsoft, check individual reviews we posted in our blog earlier.
Predictive Analytics Framework. The solution that does not need data preprocessing and does not require its users to choose learning methods. Although those things can be done by the program automatically, it comes with a payoff in the form of limited capabilities. Available options are regression, binary classification, and multiclass classification.
SageMaker. This machine-learning platform has been a breakthrough for the company since its release in late 2017. It grants almost unlimited freedom to ML practitioners and makes their job much easier as it comes with a multitude of integrated ML algorithms and pre-trained ML models. SageMaker supports TensorFlow managed instances, so developers are free to experiment with their own ML algorithms created from scratch. SageMaker integrations are not limited only to TensorFlow – Keras, Apache MXNet, Caffer2, and many others are on the list as well.
DeepLens. This service lets data scientists experiment with ML algorithms by providing not only source code, tutorials, and pre-trained ML models for deep learning, but also a programmable video camera. With these instruments, anyone can get practical experience with deep learning. Unfortunately, DeepLens is available only in eight countries so far.
Azure ML Visual Studio. Arguably the most comprehensive ML tool available on the market today. A clean and simple interface with drag-and-drop controls is ideal for building your own ML models. It may take a while to adapt and learn all available options (especially for ML beginners), but after familiarizing with the basics, model building becomes a breeze. All ML models created in Visual Studio can be exported in used in applications or sold as prediction services.
Azure ML Services. This is a great solution for those who are aiming at custom model engineering. Experienced developers prefer to use Azure ML Services with frameworks and deploying in containers using Docker. Microsoft also provides Cognitive Services, which is a collection of SDKs/APIs/complementing tools that make ML applications more intelligent.
Cloud AutoML. The service for building custom ML models with a simple graphical interface that lowers the entry threshold level for data scientists and developers. It offers products in three categories: Sight (AutoML Vision/Video Intelligence), Language (Natural Language/Translation), and Structured Data (AutoML Tables).
Google Machine Learning Engine. This is Google’s answer to AWS SageMaker. Most commonly paired with TensorFlow, but it does not mean that you are limited to using deep neural networks with this tool. Requires good expertise and experience with machine learning, target audience – ML specialists and seasoned practitioners. Flexible and versatile instrument.
In the end, it comes down to the project requirements, budget, and maybe even personal preferences because all these providers have the same ML services for managing particular tasks. To prove these words, here is a table with the most popular tools and their analogs:
Comprehensive Summary of Services
All three providers are committed to inventing new ML tools, services, and APIs to attract more clients with more advanced solutions. Currently, all of them are going nose-to-nose without any significant gaps in all market segments. We decided to gather all available services and split them into three categories: AI/ML, IoT, and Serverless. We thought that we would not limit ourselves only to machine learning tools because the Internet of Things, for instance, is also a strategically important area that can use the fruits of machine learning. Today every self-respecting IT vendor is willing to explore the IoT niche, so you might also be interested in complementary technologies. Below you will find a table with services provided by Google, Amazon, and Microsoft in terms of ML/AI, IoT, and serverless solutions.
It is not surprising that Amazon, being the first of the trio to invest heavily in the cloud, now is trying to lead the way in the ML sphere offering a lot of interesting tools as well as IoT services. But the cherry on top is still their SageMaker solution for training and deploying ML models. The only thing AWS lacks compared to GCP/Azure is the automatic generation of ML models. When it comes to the serverless aspect, AWS has one option more than its competitors. Their Lambda serverless environment neutralizes the need for on-premises hardware. Plus, you can always to deploy any app from their serverless repository.
Although Microsoft has fewer ML/AI tools in comparison with Amazon or Google, the company is aiming at high-grade and somewhat specific solutions. For example, their Cognitive Services package includes Anomaly Detection, which is useful in banking for finding and analyzing potentially fraudulent transactions. Just like Google, Microsoft has only one serverless platform. But it does an excellent job managing complex workloads. Azure’s IoT options appeal mostly to business people rather than data scientists and developers.
Google reigns supreme being the algorithm trendsetter and SEO champion. As for their AI and ML solutions, those are vast and comprehensive. Google APIs excel in the speech and text department, offering more languages that can be automatically recognized, transcribed, and analyzed. With the number of services and APIs Google has, it can play on par with others in the AI/ML segment, but in the IoT category, Google falls behind – there are just Cloud IoT Core and Cloud IoT Edge. Serverless technologies are represented by Cloud Functions, but the company promises to extend the range of services soon. Right now, we can say that the future of ML implementation with GCP looks very promising.
Instead of a conclusion, we thought it would be wiser to give you a quick summary of each company’s tools instead of writing straight recommendations. Why? Because we feel like each ML project is unique with its requirements, features, and design. If one service works best for a particular type of apps for one organization, it is not guaranteed that it will do miracles for you. Another major factor is pricing – companies have different budgets and recommending top solutions for small businesses that need simple ML tools is not the brightest idea.
AWS ML verdict: the company has invested more than anyone into the diversification of its services and tools. This includes machine learning services as well. Amazon is dedicated to enhancing current AI and ML technologies and expanding the current limits of image, voice, and face recognition capabilities. Pricing of MLaaS is difficult to estimate, but if you need massive cloud storage for data on your project, then you should note that Amazon offers very cost-effective options. We recommend using the cost calculator tool on the official website due to the complex structure of calculations or maybe even consulting with a third-party organization to pick the most effective options within your budget.
Azure ML verdict: compatibility with Microsoft products certainly opens many possibilities for organizations that are already using MS products. Add open source availability to the mix, and you will get a flexible and adjustable platform, which is Azure. Scale your project up and down whenever you want without worrying about the legacy data as it will be saved by Microsoft. The company may have fewer ML tools than Google and Amazon at the moment, but it is not going to stop investing in machine learning. Right now, Azure is probably the most flexible platform in terms of price on machine learning services. However, it requires thorough analysis, planning, and precise calculations for each project to determine the best prices.
GCP ML verdict: Google Cloud is growing fast, so does its ML segment. The corporation is working on new services while enhancing the already existing ones. Even in beta, some APIs are already showing fantastic results allowing Google to compete on par with time-tested Amazon services. There is not much to see in the IoT department, but if it is not your main concern, and you would rather sacrifice it to have cost-effective ML services, then Google ML is might be what you need. Google’s pricing is more transparent and client-friendly. In addition to that, the company has started a trend of giving out major discounts and other bonuses to win the audience.
If you're still not sure which one to pick and don't really feel like teting it out manually on your projects - contact us today and receive a professional recommendation from our data scientists!