AWS is Amazon’s secure cloud services platform which offers features from content delivery to database storage. The software is used by the UK Government, The BBC and other public sector organisations to help them improve innovation.
The event was composed of multiple interactive workshops on topics such as serverless architecture and AI/cognitive services. The keynote speaker of the summit was Amazon’s CTO and Vice President, Werner Vogels, who spoke about the democratisation of AI and how this is Amazon’s focus over the next year.
Machine Learning and AI
Hosted by Julian Simon, the first workshop we attended discovered some of Amazon’s cognitive and AI services.
Some of the features included AWS Rekognition video analysis (streamed via AWS Kinesis video stream pipeline). This feature is used to recognise objects or people and can track them between cut scenes, and can recognise faces from very little pixel data at high speeds.
Rekognition was used in Sky’s coverage of The Royal Wedding on May 19th, which allowed Sky to automatically subtitle celeb and nobility faces from the live coverage. Individuals names also appeared on Sky’s app for viewers to interact and find more details. This is the first use of the feature in a royal wedding broadcast, enhancing the delivery of the event for viewers.
Another feature and driver behind Amazon Alexa, is Polly. Polly creates lifelike text-to-speech conversion, providing the response to any conversational element of an app. AWS Polly also allows users to use SSML (speech synthesis markup language) which lets users modify the way that speech is returned.
For example, when using an abbreviation in normal discourse, this is provided as a markup. Polly is then given the marked up text and returns it in the form of voice.
AWS also provide a service for speech-to-text translation called AWS Transcribe, which is a complimentary service to Polly and is also used to drive Alexa. This is used if you provide an audio stream, it will respond with the corresponding text.
Finally, AWS Deeplens is a local hardware solution which enables the user to play with video learning. This feature is currently online available in the US but we’re watching this space for its deployment in the UK.
Deep Learning for Developers: Apache MXNet & Tensorflow
Again, hosted by Julian Simon, this workshop focused on the move away from the main AWS cognitive services and was aimed at using Amazon Sagemaker, which allows you to write your own deep learning algorithms in Python.
Simon explained how AWS provides an AMI to setup an EC2 with all of the prerequisite libraries and environments to start using the Jupyter Notebook web application. From this, users can gain access to dive into Machine Learning from an easy-to-use interface. Using Jupyter Notebook, he then explained how to use Training Accuracy and Validation Accuracy to train and apply dataset.
The second day was another day full of great insight and the event was packed to full capacity with eagle-eyed developers and Amazon fans.
Head of UK/EMEA at AWS, Gavin Jackson, kicked off the talks explaining how AWS Marketplace attempts to decrease the time required to negotiate a large service contract with a participating vendor, down to nearly zero. Amazon has almost created a plug-and-play contract negotiation function that can be performed between parties to try and replicate commonly used licensing constructs.
Next to take to the stage was Peter Vik from Jaguar Land Rover who outlined how they have implemented AWS to power their connected cards program- embedding Alexa in their cars to drive innovation. He also explained how implementing AWS has helped reduce costs and increase business agility.
Werner Vogels then presented some of the new AWS announcements and gave examples of how SMEs and blue chip organisations are using it. One great example was from Babylon Health who are using AWS in Africa in their medical design technology. The system is used to diagnose a user by aligning terabytes of data to model conditions and helps them tackle lack of medical professionals in rural areas.
Vogels emphasised how data tools can provide businesses the capability to stand out from the crowd - “it’s all about data- the data that you have as a company and the way you use it to engage with customers.”
The “Democritisation of AI” was a fantastic soundbite from Werner Vogels during the Keynote. Services like Polly, Transcribe and Rekognition allow regular developers (rather than Data Scientists and Machine Learning experts), and as a consequence - our clients - to easily integrate broad machine learning into products and projects with little more than calls to Amazon Services. The next year is going to be interesting to see what direction developers take this tech.
Serverless is an amazing leap forward, both in time to build, but also in how we’re able to quickly iterate on projects without the monolithic application sitting in front of us. Prototyping can be done so easily, even with server-side code and database constructs, that our time to Minimum Viable Product is decreasing all the time.
Overall a really interesting Summit, with some technologies that should enable us to do more, in less time, with less resource.