Today was the first Global AI Bootcamp. The Global AI Bootcamp took place in 67 locations. An initiative created by Henk Boelman. London was organised by Laura de Silva (https://twitter.com/lauraDataSci), Pablo Alverez Doval (https://twitter.com/PabloDoval) , Terry McCann (https://twitter.com/SQLShark) and Amy Boyd (https://twitter.com/AmyKateNichol). Let me start by saying, what a fantastic event! We go to a load of events and this has to be one of my highlights of 2018. The curation of the sessions was great. We had sessions on a huge variety of topics. I normally dislike keynotes, but the GlobalAIBootcamp Keynote was fantastic and incredibly inspiring.
Applied AI
Amy took us through an example of how to stitch together a series of services to create an application to auto tag items of clothing. Amy used the Custom Vision AI to create a model which can identify items of clothing. Amy used 30 pre tagged (labelled) images of items of clothing to create her model. Amy asserted that normally that would never be enough, however the Custom Vision API uses transfer learning, in order to reduce the need for thousands of labelled images. Amy then combined this with a PowerApp and LogicApps to take a photo of an item of clothing and have it tagged and email to her. Such a good demo. Individually these tools are great, working together they are better. Great stuff Amy!
(Photo provided by Laura de Silva)
Machine Learning concepts
Following Amy and the keynote video was a great session from Harry from studentsfor.ai. A lot of subjects were thrown around very quickly. What does it all mean? Harry took us through a lightening tour of all the key points mentioned. What do we mean when we say learning, parameters, features, labels etc. These concepts are not difficult, however there is a level of understand that we often forget about. This short session was a fantastic bridge that set the tone and understanding for the rest of the day. Make sure you take a look at www.studentsfor.ai it is a great idea.
The past and future of Deep Learning - Pablo Doval. There is quite a lot to deep learning, and in under an hour Pablo covered most of it. His session was amazing. Looking at the rise of deep learning and the reason the first AI winter came about, then AlexNet Alex Krizhevsky thawed the Ai winter. Pablo then looked at how to build a Neural Net and some examples of how they can be used. He showed an inspiring list of projects he has worked on before. Finally Pablo focused on an important aspect of Deep Learning. Lots of data scientists have advanced degrees, however a team of this kind will not work, you need a mixture of Data Scientists and Engineers. That could be a Machine Learning Engineer. Not sure if you need an ML engineer? Take a look at our blog: 7 Reasons you need an ML Engineer
Next up with Terry McCann our Principal Consultant and Owner of Advancing Analytics. Terry looked at how deploy a Batch Machine learning model in to SQL server using SQL Server Machine Learning Services. Deploying machine learning models is hard, using a tool such as SQL Server can make this a lot easier. Terry showed the initial interactive phase of Machine Learning in Jupyter notebooks, iterating until you're happy with a model. He then took the Python code from our Jupyter notebook and prepared it for SQL Server. He then built a table for storing the models, trained a model and deployed the serialised version of the model as a Varbinary in a table. Then using another stored procedure, we got a prediction. Terry will post a more in-depth blog on this shorty.
Auto ML
Jen Stirrup (https://twitter.com/jenstirrup) took us up to lunch time. Jen demoed an awesome new feature which she helped Microsoft to deliver around Auto ML. Auto ML is a is basically magic. Auto ML takes a dataset and iterates over a series of algorithms trying different approaches until it hits on an accuracy you're happy with. Auto ML processes such as this from Microsoft or tpot are great for quickly understanding which models work best. Then take that model and tune it. It can reduce the cycle of machine learning from days to minutes. Great talk Jen.
Lunch arrived and we stuffed our faces on a ranged of lovely sandwiches. Thanks to Plain Concepts for providing the lunch. It is lush!
Azure ML Studio
Following lunch we split in to two. Laura took half the audience through Azure ML Studio and half followed Sherin Mathew (https://twitter.com/SMdisrupt) for an introduction to the Bot Framework. This was a great lab looking at how to get a bot set up. Sherin demoed his Christmas bot and also touched on the importance of DevOps for deploying and managing a Bot in production. We will have a blog on this in the future. So much food for thought in this session after the excessive amount of sandwiches we ate!
We have had a good day, but how would I know? Look at the faces of the attendees? Maybe ask them. Or we could use sentiment analysis to see. Matt How kicked off another session looking at orchestrating a series of tools in Azure to create a PowerApp which does real-time sentiment analysis using cognitive services. Great talk Matt. Really slick slide design too!
Finally we had a session from Sherin again looking at how to integrate an AI strategy in to your business. strategy is often overlooked as it is not quite as fun as creating a Deep Learning model, but it is much more important. Having a direction and knowing what it takes to run Enterprise Applied AI is hard. This session condensed years of knowledge in to a shirt burst!
We have been to a lot of events in the last 12 months. The Global AI Bootcamp has to be one of the best. Great speakers, great topics and great attendees. We had a blast talking to you about the work you are doing around AI. Advancing Analytics are proud to sponsor the Global Ai Bootcamp in London.
If you have any questions or you're looking to understand more about any of the topics discussed then get in touch with us.
