body,html{margin:0px;padding:0px;background:#000}.loader{width:100%;height:100%;position:fixed;z-index:99999;opacity:1;transition:opacity 1s ease-in-out;background:#000}.loader .item{width:230px;margin:0px auto;position:absolute;top:50%;left:50%;transform:translateY(-50%);transform:translateX(-50%);opacity:1;transition:opacity 1s ease-in-out;display:none;}
Much ado about data – what brands could learn from Shakespeare and machine learning:
Why design matters more than ever in the charity sector
Microsoft told us how to create killer AI robots… Just kidding…or am I...
Blog / Web Development

Microsoft told us how to create killer AI robots… Just kidding…or am I...

When I say killer, I don’t mean murderous, but I do mean awesome.

When I say robots, I don’t mean big mechanical machines that shoot lasers out their eyes, but I do mean Chatbots.

I’m Arthur, a Developer here at Pull. Recently, a couple of other developers and I were lucky enough to attend Microsoft’s Azure AI technical brief, which included discussions about Chatbots, AI and Machine Learning.

Pull started working with Microsoft earlier this year and have been in discussions about AI and what it can do for us and our clients ever since. We all went to the technical briefing thinking it might be a steep learning curve, having limited knowledge and experience of AI previously. We were wrong.

Microsoft ran the event to show what products they produce, or are in the process of creating, which can assist us when developing our own Artificially Intelligent programmes. The talk began with discussion of the “Digital Revolution”. If you haven’t heard (don’t worry I hadn't), the ‘Digital Revolution’ is apparently happening right now! This revolution began with a data explosion and we have so much data now we barely know what to do with it.

According to Robin Lester, Cloud/Data Solutions Architect at Microsoft, we generate 2500,000,000,000,000,000 bytes per day. That’s 2500,000,000 GB or 2500 PB (Petabytes). If you're wondering how big a Petabyte is, think of it like this – if stars were bytes, 1 Petabyte is about 5,000 Milky Way Galaxies…  which is pretty insane! We’re now moving into a time where we want to do things with this shared data and have it more easily accessible to all individuals.

Another focus of the event was that Artificial Intelligence is artificial. It’s easy for us (as humans, not machines), to make assumptions, but AI will only learn from what we give it and can only make decisions based on information it has. For example, if I tell my AI program that a bird is a dog, until it learns otherwise, every bird to it will be a dog.

Microsoft referenced ‘Tay’ when talking about this. Tay was an AI chatbot that would “learn” from the users… this lead to a variety of issues, including certain trolls getting her to tweet some inflammatory things, turning ‘Tay’ into a sexist, geocidal, racist.

What we need to take away from ‘Tay’ is that AI needs ethics. Microsoft CEO Satya Nadella states “The most critical next step in our pursuit of AI is to agree on an ethical and empathic framework for its design. This begs the question, “Should we be using philosophers?”, which is a valid point when you have something making decisions which could potentially affect humanity. Lots of industry heads are sending letters asking for laws to be put in place around AI, no law has been set yet. Microsoft have their own set of rules for AI, as stated by Satya:


Cognitive services are where the sell is here. Microsoft is developing a multitude of different services that can help process this (now huge) amount of data, whilst also creating programmes which can relieve previously boring and repetitive jobs. These services can do things from figuring out what an image is, to analysing the sentiment of Shakespeare.

One of the most brilliant things about these services is that Microsoft have done the hard part, teaching services to recognise certain things, for instance one of the technologies can understand what a picture of a dude on a skateboard is, because Microsoft has already programmed it with this information. Basically, they’ve done the grunt work, all you need to develop are custom tweaks for your particular purpose.

We came out of the event with some brilliant tools we hope to use in future:
 
The Bot framework

This program and what it can do when integrating with Cognitive services, even just out of the box, is incredible. It gives a great template to work with when creating your own Chatbot, allows you to create a bot for most platforms and makes it easier to create previously complicated Chatbot builds.

 
The Video Indexer

This tool allows you to get insights from video - who’s talking, when they’re talking and what their sentiment is - without watching the video yourself.

For instance, when I put a video from the music retailer Andertons Music Co. in Microsoft’s Video Indexer tool I discovered keywords in the video included Led Zeppelin and White Snake, that they were indoors, there was a man playing a guitar and that guitarist appeared for 88% of the video's duration.

Give it a go yourself and find out all the other things it has to offer.
 
The Machine Learning Studio

This is wonderful if you want to try machine learning, but the cognitive services aren’t specific to your needs and you don’t have a data scientist to hand.

The program has a simple drag and drop interface which allows you to add complex algorithms for the AI to make decisions and predictions about your data.

For example, if you opened a box of eggs to find one cracked, you likely wouldn’t be able to immediately pinpoint how this happened if you previously thought they were all whole. Now imagine someone gave you all the data of how those eggs got there, the journey taken by ones that got cracked and the ones that didn’t, then suddenly asked you to predict how many eggs would be cracked in the next delivery – this would be almost impossible as there would be too much information and potential scenario’s for you to go through.

However, if you gave the Machine Learning Studio all that data along with lots of other ‘egg journeys’ within a few hours of processing the data you could train the service, with a certain amount of confidence, to predict the likelihood of any eggs being cracked in the next delivery and how many of them.

Obviously, this analogy is a little out there, but this allows you to see how the Microsoft Machine Learning tool can calculate and process complicated, large amounts of data and come out of it with a result and action you can use to inform potential business decisions.
 
Microsoft ran the event to “democratise AI”, something Satya Nadella has been quoted with, and make it more accessible, letting “Every developer be an AI developer”, which in my opinion is a great thing.

We’re now in a new age of development, where there are so many shared technologies available enabling us to make incredible things your childhood self would have never even dreamed of… so I say let’s use these amazing tools and start building!
 

 

Console