Amazon ramps up its AI offerings
Machine Learning and Artificial Intelligence are core parts of a lot of applications nowadays. Thus, it is no surprise that there exists a growing number of machine learning frameworks and services. These include pre-designed models that can be trained on your own data and then applied in your applications, or frameworks that can be used to freely design your own algorithms to train and deploy.
At its re:Invent conference Amazon introduced a number of tools for developers. Amazon Machine Learning is not a new service, however, as we have stated about a year ago, the initial offerings by Amazon Web Services (AWS) have been not as extensive as those of some competitors. With the introduction of these tools, Amazon is now putting more focus on its machine learning and AI offerings on the AWS platform.
Sagemaker is tool intended to facilitate the easy realization of an end-to-end machine learning pipeline. Analyzing the data, then building and training a model, and finally deploying it. You can explore the data lying on your S3, clean it and preprocess it using a built in Jupyter Notebook IDE. The models can then be built using pre-installed machine learning algorithms, pre-configured open-source frameworks such as TensorFlow, or a framework of your own using Docker containers. Models can then be deployed to EC2 instances, e.g. with HTTPs endpoints for invoking the models. From data cleaning to deployment can all be covered with Sagemaker. More details on Sagemaker can be found in this blog-post.
Rekognition, Transcribe, Comprehend, and Translate are API based services for your image and NLP tasks. With Rekognition you can for example do object, scene, and activity detection, facial recognition and analysis, or person tracking in videos. Transcribe and Translate both do what their name suggests, namely convert speech to text and language translation, respectively. Comprehend is an NLP tool offering keyphrase extraction, entity recognition, sentiment analysis and topic modeling, among others.
These are all accessed via an API and can be combined to build your application of choice with a tight integration into the AWS ecosystem, eliminating the heavy lifting of developing and deploying your own algorithms.
How will your App employ AI
Amazon is giving us the choice. Do we use the built in algorithms and employ the power of machine learning exposed by the provided APIs or do we build our own, from end-to-end?
Thanks for reading! If you enjoyed this article, please leave a 👏