Winterwell Newletter #1 December 2008

The Winterwell Newletter: The First Issue

newsEarlier this year Joe Halliwell and Daniel Winterstein founded Winterwell. Working with clients of all shapes and sizes, we use maths and artificial intelligence to solve problems and make businesses more efficient.

Our work is varied, taking in IT, new media, emerging technologies, robotics, risk management and data mining. The best way to find out about what we do (and how we could help you) is to see some of our work. Hence this, our first newsletter. In this issue:

Smart email

Some companies receive dozens of emails a day. Others receive hundreds, and many companies receive thousands. And someone has to deal with these emails as promptly, politely and efficiently as possible. But there's only so much a person can do in one day. The problem is time versus resource. That was enough challenge for us, so with support from the Technology Strategy Board, we set about finding an answer - and we're well on our way.

chatbot conversation logo We're developing a piece of software that will be able to automatically answer emails for you. We don't mean those automatic responses saying 'I'll be back on Monday' you can set up in your email software. We're talking about software that will learn from examples, and carefully choose the correct response based on its understanding of the message content. Obviously there will always be some messages you'll want to reply to personally, but an application like this could really save businesses a lot of time and effort - in call centres, for example.

It's still early days, and there's obviously a lot of testing and polishing to be done before this software can be let loose on valuable customers or suppliers. But the prototype is due early in 2009, so if you'd like to learn more before then, email or call us. Back to top

The world is your playground

The Puppet Master

With an increasing emphasis on realism and interactivity, game players are constantly seeking new and more involving game experiences. And the next phase of gaming is emerging with Alterative Reality Games (ARGs - also known as interactive drama or transmedia fiction). These games get players more involved by using a combination of media - anything from newspapers to TV, phones and of course the Internet with Twitter, Facebook and also custom-built sites. In short, the player's world becomes the gaming platform - the ultimate in realistic graphics.

But there's one big problem with ARGs, perhaps best summed up by ARG creator Sean Stewart:

"It just, you know, costs you unbearably in blood, sweat and tears over an extended period of time."

Yes, while potentially lucrative, ARGs are very labour intensive. That's why we're looking to develop a new system called Puppet Strings to provide the tools for making ARG development quicker and more cost-effective.

We're also involved with a certain media company on an exciting project based around ARGs... but we're not allowed to talk about it yet. Watch this space!

A limited-functionality version of Puppet Strings should be available in 2009. But if you want to hear more about the possibilities of ARGs and Puppet Strings before then, you can email us at or call us. Back to top

I'll be Bach.

The idea of machines being able to complete human tasks is nothing new - just look at washing machines, spell checkers or the films of Arnold Schwarzenegger.

HMM Bach drawn by Joe Ruddy However, the notion of a computer actually being creative is something different. A washing machine just follows a given set of instructions every time you use it, but to be creative a computer would actually need to learn. So, working with the Institute for Adaptive and Neural Computation (ANC) at Edinburgh University, we set out to test whether this was possible by creating a computer program that could learn how to play music.

Specifically, we wanted our program to harmonise in the style of classical composer J. S. Bach. In the 1500s the Lutheran church began to use hymns called 'chorales'. At first these hymns were simple melodies for people to sing, but then composers started to elaborate them with complex harmonies. Possibly the greatest of these composers was J. S. Bach, so we gave our program about a hundred of his pieces to see if it could teach itself how to harmonise.

The program successfully identified patterns in Bach's music. In other words, it made up the rules of how to compose like Bach on its own, purely by analysing his music. The result is pretty impressive - although Bach needn't worry about the competition, the program can produce a plausible harmony based on any melody it's given.

Aside from making music, machine learning like this can be used to solve problems in a wide range of fields; from computer gaming to medicine. As artificial intelligence and statistical consultants, we put this technology to work for our clients in all areas of business. So if there are parts of your business you'd like to improve, we can help. Just email us at or if you'd like a chat, call us. Back to top

Do you know what the future holds?

Might. Possibly. Potentially. Pick up any startup report on financial projections or predicted costs and you'll find words like these. Such reports are based on guesswork, because none of us can predict the future. They're a 'best guess'.

TriStart Out latest client is a company called TriStart. Their online risk assessment tool gives startups and very young companies a more accurate idea of how a business will perform, taking into account detailed factors such as market, technology, the entrepreneurial qualities of individuals and teams, training and skills. An assessment can be easily completed online, with a report generated instantly. The success of this approach means TriStart have public and private sector clients all over the world. And now they've signed a partnership deal with Winterwell, so we'll be developing the next generation of their system. Again, this will be a convenient, online application, and as the system learns more about you and your company, statistical learning software will deliver continuous improvements to the service.

If you're considering setting up in business, or you have a young business and you want a more accurate idea of your future, TriStart can help. Their new service will launch in January 2009.

And if you'd like to speak to us on the subject, feel free to email or call us. Back to top

How artificial intelligence is creating petrol-powered dogs

Anyone with an interest in robotics (one of our own fields of interest), will tell you that enabling a robot to walk is hair-tearingly difficult. So we're in awe of Boston Robotics and their creation 'Big Dog', which not only walks, but can also correct itself when it slips. The breakthroughs that made this feat possible are all down to the maths behind the robot's brain. In fact you'd be surprised at the amount of maths involved in walking, (the more curious amongst you can read more about this in an old Scotsman article written by Daniel).

Of course, you may just be amused at the bizarre, four-legged creature slipping on the ice. But that works too. Enjoy. Back to top


That's all for this issue, we hope you found it useful, (or at the very least, interesting). We're always here to discuss any problems in your business, or ways of making it more efficient - you'd be surprised at the things we can help you with. Feel free to email us at or phone us. Thank you for reading!