Photocredit: Mashable

Machine Learning, Neural Networks and Artificial Intelligence (AI): How are intelligent technologies predicting decisions?

Part of the ‘Artificial Intelligence for Beginners’ series

What is Artificial Intelligence and how is it influencing the decision making process? What must be understood about artificial intelligence in order to examine how this form of technology is providing businesses with a tool to understand human decisions? These are the sorts of things that I will explore more deeply in this piece.

To someone that has not explored Artificial Intelligence, the term itself may lead him or her to imagine a post-apocalyptic world as a result of ‘Skynet’, or perhaps the so-called ‘corrupt’ computer system V.I.K.I in I-Robot. What these dramatized versions of the ‘future’ portrayed by Hollywood forget to tell you however, is that Artificial Intelligence is an area where some of the most intelligent minds spend their time exploring, and with examples such as IBM’s ‘Watson’ or the even earlier ‘Eliza’, that aimed to put the Turing test to work, it’s no surprise that we’re seeing continued advances in the AI industry. Before I examine exactly how these advances in technology are leading to ‘intelligent’ decisions by complex algorithms, I think it’s important that we understand exactly what the terms above are, and how they play their part with a number of other systems in the AI space.

Photocredit: Geekwire

Starting with Machine Learning, this area of AI aims to explore algorithms that allow machines to act without being programmed to do so. These so-called machines and algorithms learn from patterns and adjust their behavior accordingly. Thus to some extent, it can be considered as a form of semi-autonomous decision making with the intention of removing human interaction. Artificial Neural Networks (ANN), are models based on the biological neural networks present in the brain, that aim to solve tasks that would be too difficult for traditional methods of programming. Thus a combination of the two, can lead to a set of algorithms that not only learn from patterns, but also use forms of programming that traditional methods would not be able to solve. It is this form of analysis that I will explore in more detail, although I will in the future look into Deep Learning, Data Mining and Natural Language Processing (NLP) and how all these forms of AI can also benefit a business and its wider consumers.

Companies such as Apple, Google and Microsoft are all using forms of Artificial Intelligence, or more specifically, forms of Machine Learning, and this is presented in the form of Siri, Google Now and Cortana. Although the focus in this case is perhaps looking into creating a deeper understanding of speech through NLP, these systems are constantly working on recognizing patterns in human behavior. A better example in the case of Apple is their pattern recognition and analysis in predictive keyboards, which are constantly learning about how a user interacts with a mobile device and the keyboard, with the end game of allowing the user to quickly write messages in their own style and for common mistakes to be eliminated in the process. In this respect, this is an algorithm that is constantly evolving, one that demonstrates Apple’s understanding of self-developing algorithms. As Facebook begins to explore the idea of creating their very own Siri-like NLP software in messenger, they too are exploring the different forms of AI that can be used in order to develop a user’s overall experience, but in the background begins to analyze a users interests and behavioral patterns. It’s Google however, that’s leading the race to create something capable of passing a Turing test, and with their recent acquisition of DeepMind it looks like they don’t have any plans to slow down in the near future. AI is something that Google are implementing across a range of their products, from Maps to Advertising, and everything in-between.

Photocredit: Techcrunch

So how does this technology really impact the everyday user’s lifestyle, and how can it be used to predict decisions made by those that are using it. One way has already been clearly demonstrated, in the form of predictive text analysis, but other forms are already beginning to present themselves to a larger market, such as ‘NEIL’ or the Never Ending Image Learner. NEIL was released at Carnegie Melon University, and uses complex algorithms to constantly compare relationships between different images. Although the technical capabilities may be present, the question is how can this tool be used to benefit the everyday user and the connections between businesses and individuals? Certainly one way whereby this technology could be formulated is through the combination of pattern analysis in fashion. As people become accustomed to their own ‘style’, technology such as NEIL could be implemented to present the user with further ‘similar’ items of clothing that match this, hence removing the barrier of searching for something new that suits their style. This does however, rely on a user’s style remaining constant, and naturally forms of automated learning must be implemented in order to help the algorithms adapt according to changes in style and fashion. It’s also very powerful for those involved in the business side of fashion, as those that operate in the industry have an ability to clearly analyze the popular items of clothing, and one may even predict that forecasting in the future points to trends based on algorithms. Manipulated algorithms look to influencing a consumer’s decision to the point of purchase, and hence create the possibility of businesses being able to benefit from this sort of technology. Although this may raise ethical questions, is it unreasonable to assume that this is the way that technology is headed? As Google and other leading companies pour money into this new intelligence, only time will tell.

So where can AI be implemented in order to generate an understanding of a user, or to make predictive assumptions about a user. One could argue that just about any industry can benefit from this sort of technology, as mass data can be repositioned and presented in a personalized way. Recent research papers are effectively able to show that a combination of Data Mining and data analysis with the use of self-learning algorithms, that computers are now able to understand individuals from their Facebook accounts, given that enough information is supplied. This doesn’t just mean understanding a user more thoroughly, but it leads to computers being able to make emotionally intelligent decisions, and we could even argue that computers are becoming more socially skilled. Thus data sets become more specific to each user, and businesses are able to present tailored information to every individual. What I’m arguing here of course, does lead to the idea that computers understand the individual better and better, and as we continue to build a digital footprint, these algorithms allow businesses to make more informed decisions about a person. This can however lead to the wider public feeling uncomfortable, but that’s an ethical issue, and something that I will leave to be understood by each person differently. One thing that is becoming clear however is that online privacy is something that’s going to become harder to achieve for generations after us today, unless a counter-product is created.

Photocredit: Louis Dorard

Now that I’ve explained the increasing level of intelligence that computers possess today, it’s easy to see that the future points towards a world where a digital presence of someone can be easily interpreted and predicted by intelligent algorithms. As computers carry out processes quicker than the human mind, the ability to process data becomes more possible, and more accurate. The important thing to understand here however, is that this should not be seen in a negative light, or a dangerous tool possessed by businesses, but instead I would argue that these tools should be used to drive innovation further, allowing humans to build products that not only work more effectively, but work to benefit the end user on a personalized level, as software begins to understand the end game of the one using it, and thus is able to present this data more quickly. Over the next few months I will be writing pieces that touch on Artificial Intelligence, and an ever-changing technological realm, with the aim to make it easier for the everyday person to understand the rules of AI, and how it works. I aim to demystify the industry, making it more easily understandable so that a wider audience is able to understand both the negatives and positives of the industry and it’s potential.