My closing keynote of the Tweakers DevSummit – slides and resources

Yesterday I gave the closing keynote of the Tweakers Developer Summit in Utrecht, The Netherlands. The conference topic was “Webdevelopment – Coding the Universe” and the organisers asked me to give a talk about Machine Learning and what it means for developers in the nearer future. So I took out my crystal ball 🔮 and whipped up the following talk:

Here are the resources covered in the talk:

• GOAT raising $25M for marketplace for collectible sneakers – this was an example of how we as developers failing at influencing what our market is doing. We’re still at a stage where plain consumerism wins the big investments and technology is used to cater to the needs of a very small, very affluent group
• The Robot Takeover – an interactive site telling you how safe your job is from being taken over by a robot. Based on this research in PDF format
• Movie written by AI
• Daddy’s Car – a song written by AI after analysing lots of Beatles songs
• AI software learns to write AI software – Google Brain coding itself
• Beer brewed by AI
• Chinese factory replacing 90% of humans with robots – 250% more productivity, defects 90% down
• The Power of Big Data and Psychographics – an incredibly scary presentation of Cambridge Analytica at Concordia summit showing how applying the OCEAN psychology test to Facebook quizzes allowed them to influence voters in the Brexit and Trump votes
• Meitu’s App Permissions – an app to make sparkly selfies has full access to your phone now
• Google Photos and its impressive use of Machine Learning to tag and find your own photos
• Pinterest Lens – a Shazam for everyday objects
• AIPoly – an app to detect objects and tell blind users what they are using a Neural Network on the device instead of having to be online and suffering the latency that comes with that. Demo video here
• Google’s DeepMind’s lip reading success – AI can read lips four times better than humans. You can also read the scientific paper on it
• Microsoft Office using Cognitive Services – PowerPoint now creates text alternatives for images
• Learning tools in Microsoft OneNote and a video on how that helps educators
• Live translation of phone calls using Cognitive Services and Skype
• Google Allo using machine learning to create recommended answers to posted images
• Pixel recursive super resolution is a method to create facial images from 8×8 pixel input using machine learning. In essence it is making the zoom and enhance movie trope reality.
• Google open sourced their DeepMind Lab to train AI. The code is on GitHub and the “paper available on Arxiv
• Facebook explained and released their image segmentation algo showing you how to detect shapes in images and classify them
• ImageNet is a database of described images
• Open images dataset is Google’s alternative to ImageNet, a lot of metadata for image URLs described with humain aid
• CaptionBot is a demo by microsoft using Cognitive Services to describe images.
• Microsoft Cognitive Services is a suite of APIs using Machine Learning to understand images, sounds, text and enhance your projects to have a more human output and allow for human input
• Custom Recognition Service is a service by Microsoft to filter content by giving it context, thus improving the outcome of OCR or speech to text conversions
• Video Breakdown is a service to detect speakers, sentiments and create a transcript of videos
• My visit to the medical holodeck is my report about visiting Weill Cornell hospital in New York where they use VR for visualisation of diseases and HoloLens to design medication
• Sophia Roshal’s HoloLens demo shows the intern’s project at Weill Cornell using HoloLens to visualise cancer medication and tumor data
• Anamaria Cotîrlea is a Google employee whose research as an intern saves the company 1.5 petabytes of data each day by switching the Android store from gzip to brotli compression.
• Amazon’s Mechanical Turk trained AI for years showing that a lot of the computer magic we see today is the result of human endeavours
• The next blue collar job is coding explains how a lot of soon-out-of-job workers re-educate themselves working in IT as it is the only really growing market

Yes, this was a lot – maybe too much – for one talk, but the feedback I got was all very positive, so I am hoping for the video to come out soon.