The phrase ‘Artificial intelligence’ or ‘AI’ can conjure a vision of an apocalyptic future ruled by robot overlords in the form of a young Arnold Schwarzenegger. The truth is that human annihilation at the hands of AI is way, way, way off (or at least we hope so…..Stephen Hawking, Bill Gates and Elon Musk might think differently, but hey, what do they know!). In the meantime, there’s a lot more mundane AI stuff we need to get to grips with – including what it means for customers of financial services!
These days ‘AI’ – like its cousin (and important resource!) ‘Big Data’ a few of years ago – is a term bandied about, but often only vaguely understood, so we thought we’d write a couple of (humble) blogs laying out the turf. Firstly, here follows a bit of a primer, for those of you that aren’t entirely sure what the term means and where it comes from..
“AI is the new electricity” – Andrew Ng
One of the leaders in the field, Stanford Professor, turned VC Andrew Ng stated: “AI is the new electricity”. Meaning in the same way electricity revolutionised the way people lived in the 19th – 20th century so will AI completely shift how we live and work today.
Already AI is changing how we engage and interact with technology – for example, Amazon’s Alexa or Apple’s Siri and FaceID enabled by powerful backend architectures facilitating state-of-the-art voice and facial recognition. More exciting, perhaps, are self-driving cars – already being tested extensively – which have proven to outperform human drivers in some circumstances. This brand of AI really gets the juices going.
Enter your email below and you’ll be the first to know when Xinja is ready to launch.
AI is everywhere, and has been for a while…
However, more subtle but equally disruptive forms of ‘AI ’include recommendation systems. Sorry to disappoint, but these have been around for about 2 decades. Amazon developed ‘collaborative filtering’ very early on – using standard statistical models to identify significant correlations between products (in their case books) which created the ‘people who enjoyed this, also liked this’. (Incidentally there were some epic fails in the early days, possibly before their significance testing was robust, causing people who chose Tolstoy to be referred to obscure punk rock bands etc.) This technique was picked up and deployed across a huge range of ecommerce platforms, from groceries to fashion to music, and has just merged into everyday life. But the algorithms have become more nuanced and ‘intelligent’ – taking in multiple types of data. Netflix, apparently, uses a highly complex one to proffer the next show (although quite how it gets ‘Legally Blonde’ from my staple diet of ‘House of Cards’ and documentaries escapes me – the problems of different members of the family using the same login of course…).
Natural Language Processing
The manifestations are becoming more exciting – particularly in the form of NLP. We get excited at the ‘newness’ of chatbots communicating apparently like humans, but again, NLP has been around for a while in areas of search for about 20 years. All the talk then was of the ‘semantic web’ and computers understanding context, and this extended to them understanding meaning (not just individual words). This gave way to, amongst other things, dynamic FAQs – mundane yes but useful. A computer can understand that question a is actually very similar to question b and therefore pull the answer for question b, which it already has.. Chatbots are an understandable evolution for NLP, but like any form of AI, are only as good as the quantity and richness of the data off which they feed and learn.
New uses of AI
Now of course the innovation is touching many sectors including design, where people like Maurice Conti are using AI to design lighter, faster and more durable drones. And similarly everything from health to cybersecurity to protection of endangered species is finding ingenious uses for AI. Although it’s not without its fair share of hype, AI is having impacts that far outreach what was previously thought possible.
So…what’s the definition of Artificial Intelligence?
AI has notoriously always been difficult to define. Even now there is no agreed consensus, with multiple interpretations all evolving overtime. To break it down into the simplest possible, however, ‘artificial’ can be referred to as ‘not made by humans’, and intelligence as the ability to acquire knowledge and/or skill.
Still not sure?
Alan Turing – of ‘Enigma’ fame and one of the pioneers in early computer science – also had the same difficulty pinning down the exact scope of AI. It was for this reason he devised the Turing test, which acted as a framework in determining machine intelligence.
In this test, a computer and human are to required to complete a task and if the evaluator cannot reliably differentiate the machine from human output, then the program is accepted to be AI.
Our interpretation although simplistic captures the essence of AI as something that drives a cognitive-like task. Simple business rules or conditional programming (like the programming described above in ‘collaborative filtering’ or simpled ‘if/then’ statements) if executed well enough could be called ‘AI’ – particularly if they remember and automatically optimise based on previous results or outcomes..
Artificial Intelligence, Machine Learning, Deep Learning…What’s the difference?
In the AI, big data and analytics space, you’ve probably heard of other terms like machine learning (ML) and deep learning come up (DL) and perhaps used interchangeably.
What is Machine Learning?
Where we can think about AI as computer performing human like tasks you can think of machine learning as a subdiscipline of AI. Machine Learning can be defined as “the field of study that gives computers the ability to learn without being explicitly programmed.” For the most part Machine Learning is comprised of supervised learning, which is where the computer learns from a labelled training set.
What is deep learning
The term Deep learning is a buzzword that is finding its way up with big data, machine learning and AI. To make it simpler This is where most of the time and attention has gone in developing AI in recent years.
Brushing up on your AI history
Fun AI fact! While AI has exploded in the last decade and we’ve seen innovation after innovation, what most people don’t realise is that many of foundations in the field were accomplished in the early 60s and (even more shocking) the neural networks architecture applied today is not so dissimilar to the work Walter Pitts and Warren McColloch did in this area in 1943!
AI went through periods of peaks and toughs, with two tough winters (reduced attention and funding) that lasted from 1974-1980 and 1984-1993. There was in fact an AI winter that lasted all through the 1980s to the late 1990s. But even setting aside the hype, there HAS been a massive acceleration in AI activity. Whilst building on traditional mathematics, as Fortune Magazine’s Roger Parloff put it in September 2016 (in his seminal article).
“AI is suddenly changing everything” – Roger Parloff
The resurgence in AI can largely be attributed to Moore’s Law (or at least what it charts – the doubling in computer processing power every 2 years) which paved the way for smarter, faster and more powerful devices that connect the physical to the digital world. This enabled not only capturing a wealth of ‘big data’ that simply wasn’t accessible before, but also the ability to store and analyse this information in a meaningful way.
Another major contributor was a man named Geoffrey Hinton, known by some as the Godfather of modern AI – (interestingly both a computer scientist and a cognitive psychologist) – as it was Hinton that made significant breakthroughs in the development of these neural networks – accelerating the sophistication of cognitive tasks computers were able to perform.
What’s to come?
AI is still in its infancy, and you can expect the hype to accelerate as fast as the progress, however the rate of change is such that it can develop, by its nature, very very rapidly. For this reason, we should probably be listening to the destruction advocates a little more attentively. However, the reason Xinja is focused on this is because AI is at the heart of what we do. To begin with we will be gathering data on our customers, this will be the core of our AI in improving customer experiences. In fact, AI in financial services, and what we will be doing at Xinja, is a whole other topic so watch out for our next AI blog.
Join the waitlist below, if you haven’t already, to get early access to the Xinja pre-paid card & other products, & hear about our developments & opportunities to own a piece of the action in future crowdfunding.