The role of natural language processing in AI

What is natural language processing?

Natural language processing (NLP) is a branch of artificial intelligence within computer science that focuses on helping computers to understand the way that humans write and speak. This is a difficult task because it involves a lot of unstructured data. The style in which people talk and write (sometimes referred to as ‘tone of voice’) is unique to individuals, and constantly evolving to reflect popular usage.

Understanding context is also an issue – something that requires semantic analysis for machine learning to get a handle on it. Natural language understanding (NLU) is a sub-branch of NLP and deals with these nuances via machine reading comprehension rather than simply understanding literal meanings. The aim of NLP and NLU is to help computers understand human language well enough that they can converse in a natural way.

Real-world applications and use cases of NLP include:

  • Voice-controlled assistants like Siri and Alexa.
  • Natural language generation for question answering by customer service chatbots.
  • Streamlining the recruiting process on sites like LinkedIn by scanning through people’s listed skills and experience.
  • Tools like Grammarly which use NLP to help correct errors and make suggestions for simplifying complex writing.
  • Language models like autocomplete which are trained to predict the next words in a text, based on what has already been typed.

All these functions improve the more that we write, speak, and converse with computers: they are learning all the time. A good example of this iterative learning is a function like Google Translate which uses a system called Google Neural Machine Translation (GNMT). GNMT is a system that operates using a large artificial neural network to increase fluency and accuracy across languages. Rather than translating one piece of text at a time, GNMT attempts to translate whole sentences. Because it scours millions of examples, GNMT uses broader context to deduce the most relevant translation. It also finds commonality between many languages rather than creating its own universal interlingua. Unlike the original Google Translate which used the lengthy process of translating from the source language into English before translating into the target language, GNMT uses “zero-shot translate” – translating directly from source to target.

Google Translate may not be good enough yet for medical instructions, but NLP is widely used in healthcare. It is particularly useful in aggregating information from electronic health record systems, which is full of unstructured data. Not only is it unstructured, but because of the challenges of using sometimes clunky platforms, doctors’ case notes may be inconsistent and will naturally use lots of different keywords. NLP can help discover previously missed or improperly coded conditions.  

How does natural language processing work?

Natural language processing can be structured in many different ways using different machine learning methods according to what is being analysed. It could be something simple like frequency of use or sentiment attached, or something more complex. Whatever the use case, an algorithm will need to be formulated. The Natural Language Toolkit (NLTK) is a suite of libraries and programs that can be used for symbolic and statistical natural language processing in English, written in Python. It can help with all kinds of NLP tasks like tokenising (also known as word segmentation), part-of-speech tagging, creating text classification datasets, and much more.

These initial tasks in word level analysis are used for sorting, helping refine the problem and the coding that’s needed to solve it. Syntax analysis or parsing is the process that follows to draw out exact meaning based on the structure of the sentence using the rules of formal grammar. Semantic analysis would help the computer learn about less literal meanings that go beyond the standard lexicon. This is often linked to sentiment analysis.

Sentiment analysis is a way of measuring tone and intent in social media comments or reviews. It is often used on text data by businesses so that they can monitor their customers’ feelings towards them and better understand customer needs. In 2005 when blogging was really becoming part of the fabric of everyday life, a computer scientist called Jonathan Harris started tracking how people were saying they felt. The result was We Feel Fine, part infographic, part work of art, part data science. This kind of experiment was a precursor to how valuable deep learning and big data would become when used by search engines and large organisations to gauge public opinion.

Simple emotion detection systems use lexicons – lists of words and the emotions they convey from positive to negative. More advanced systems use complex machine learning algorithms for accuracy. This is because lexicons may class a word like “killing” as negative and so wouldn’t recognise the positive connotations from a phrase like, “you guys are killing it”. Word sense disambiguation (WSD) is used in computational linguistics to ascertain which sense of a word is being used in a sentence.

Other algorithms that help with understanding of words are lemmatisation and stemming. These are text normalisation techniques often used by search engines and chatbots. Stemming algorithms work by using the end or the beginning of a word (a stem of the word) to identify the common root form of the word. This technique is very fast but can lack accuracy. For example, the stem of “caring” would be “car” rather than the correct base form of “care”. Lemmatisation uses the context in which the word is being used and refers back to the base form according to the dictionary. So, a lemmatisation algorithm would understand that the word “better” has “good” as its lemma.

Summarisation is an NLP task that is often used in journalism and on the many newspaper sites that need to summarise news stories. Named entity recognition (NER) is also used on these sites to help with tagging and displaying related stories in a hierarchical order on the web page.

How does AI relate to natural language processing?

Natural language processing – understanding humans – is key to AI being able to justify its claim to intelligence. New deep learning models are constantly improving AI’s performance in Turing tests. Google’s Director of Engineering Ray Kurzweil predicts that AIs will “achieve human levels of intelligence” by 2029.

What humans say is sometimes very different to what humans do though, and understanding human nature is not so easy. More intelligent AIs raise the prospect of artificial consciousness, which has created a new field of philosophical and applied research.

Interested in specialising in NLP?

Whether your interest is in data science or artificial intelligence, the world of natural language processing offers solutions to real-world problems all the time. This fascinating and growing area of computer science has the potential to change the face of many industries and sectors and you could be at the forefront. 

Find out more about NLP with an MSc Computer Science with Artificial Intelligence from the University of York.

What you need to know about blockchain

Blockchain technology is best known for its role in fintech and making cryptocurrency a reality, but what is it? 

Blockchain is a database that stores information in a string of blocks rather than in tables, and which can be decentralised by being made public. Bitcoin, one of the most talked about and unpredictable cryptocurrencies, uses blockchain as does Ether, the currency of Ethereum. 

Although cryptocurrencies have been linked with criminal activity, blockchain’s mechanism of storing data with time stamps provides offers transparency and traceability. Although central banks and financial institutions have been wary of the lack of regulation, retailers are increasingly accepting Bitcoin transactions. It’s said that Bitcoin founder, Satoshi Nakamoto, created the cryptocurrencies as a response to the 2008 financial crash. It was a way of circumnavigating financial institutions by saving and transferring digital currency in a peer-to-peer network without the involvement of a central authority.

Ethereum is a blockchain network that helped shift the focus away from cryptocurrencies when it opened in 2015 by offering general purpose blockchain that can be used in different ways. In a white paper written in 2013, the founder of Ethereum, Vitalik Buterin, wrote about the need for application development beyond the blockchain technology of Bitcoin, that would lead to attachment to real-world assets such as stocks and property. Ethereum blockchain has also provided the ability to create and exchange non-fungible tokens (NFTs). NFTs are mainly known as digital artworks but can also be digital assets, such as viral video clips, gifs, music, or avatars. They’re attractive because once bought, the owner has exclusive rights to the content. They also protect the intellectual property of the artist by being tamper-proof.

There has recently been a lot of hype around NFTs because the piece Everydays: The First 5000 Days by digital artist Beeple (Mike Winkelmann) sold for a record-breaking $69,346,250 at auction. That’s the equivalent of 42,329 Ether, which was what Vignesh Sundaresan, owner of Metapurse, used to purchase the piece that combines 5,000 images created and collated over 13 years. NFTs may seem like a new technology but they’ve actually been around since 2014.

IOTA is the first cryptocurrency to make possible free micro-transactions between Internet of Things (IoT) objects. While Ethereum moved the focus away from cryptocurrency, IOTA is looking to move cryptocurrency beyond blockchain. By using a Directed Acyclic Graph called the Tangle, IOTA manages to rid any need for miners, allows for near-infinite scaling, and removes fees entirely.   

How blockchain works

Blockchain applications are many and varied including the decentralisation of financial services, healthcare, internet browsing, real estate, government, voting, music, art, and video games. Blockchain solutions are increasingly utilised across industries, for example, to provide transparency in the supply chain, or in lowering administrative overheads with smart contracts.  

But how does it actually work? Blockchain uses several technologies including distributed ledger technology, digital signatures, distributed networks and encryption methods to link the blocks of the ledger for record-keeping. Information is collected in groups which make up the blocks. The blocks have certain capacities which, once filled, become chained to the previously filled block. This creates a timeline because each block is given a timestamp which cannot be overwritten.

The benefits of blockchain are seen not just in cryptocurrencies but in legal contracts and stock inventories as well as in the sourcing of products such as coffee beans. There are notoriously many steps between coffee leaving the farm where it was grown and reaching your coffee cup. Because of the complexity of the coffee market, coffee farmers often only receive a fraction of what the end-product is worth. Consumers also increasingly want to know where their coffee has come from and that the farmer received a fair price. Initially used as an effective way to cut out the various middlemen and streamline operations, blockchain is now being used as an added reassurance for supermarket customers. In 2020, Farmer Connect partnered with Folger’s coffee in using the IBM blockchain platform to connect producers with customers. A simple QR code helps consumers see how the coffee they hold in their hand was brought to the shelf. Walmart is another big name providing one of many case studies for offering transparency with blockchain by using distributed ledger software called Hyperledger Fabric.

Are blockchains hackable?

In theory, blockchains are hackable, however the time and resources – including a vast network of computers – needed to achieve a successful hack are beyond the average hacker. Even if a hacker did manage to simultaneously control and alter 51% of the copies of the blockchain in order to gain control of the ledger and make their own copy the majority copy, each block would then have different timestamps and hash codes (the cryptographic algorithm). The deliberate design of blockchain – using decentralisation, consensus, and cryptography – makes it impossible to alter the chain without it being noticed by others and irreversibly changing the data along the whole chain.

Blockchain is not invulnerable to cybersecurity attack through phishing and ransomware but it is currently one of the most secure forms of data storage. Permissioned blockchain adds an additional access control layer – actions performed only by identifiable users allow access. These blockchains are different to both public blockchains and private blockchains.

Are blockchains good investments?

Currencies like Bitcoin and Ether are proving to be good investments both in the short-term and the long-term; NFTs are slightly different though. A good way to think about NFTs is as collector’s items in digital form. Like anything that’s collectable, it’s best to buy something because you truly admire it rather than because it’s valuable, especially in the volatile cryptocurrency ecosystem. It’s also worth bearing in mind that the values of NFTs are based entirely on what someone is prepared to pay rather than any history of worth – demand drives price.

Anyone can start investing but as most digital assets like NFTs can only be bought with cryptocurrency, you’ll need to purchase some, which you can easily do with a credit card on any of the crypto platforms. You will also need a digital wallet in which to store your cryptocurrency and assets. You’ll be issued with a public key, which works like an email address when sending and receiving funds, and a private key, which is like a password that unlocks your virtual vault. Your public key is generated by your private key which makes them a pair and adds to the security of your wallet. Some digital wallets like Coinbase also serve as crypto bank accounts for savings. Although banks occasionally freeze accounts with relation to Bitcoin transactions, they are becoming more accustomed to cryptocurrencies. Investment banks such as JP Morgan and Barclays even show interest in the asset class despite the New York attorney general declaring “Play by the rules or we will shut you down” in March 2021.

Are blockchain transactions traceable?

In a blockchain, each node (a bank of computers) has a complete record of the data that has been stored on the blockchain since it began. So for example, the data held by a Bitcoin is the entire history of its transactions. If one node presents an error in its data, the thousands of other nodes help by providing a reference point for the error so it can correct itself. This architecture means that no single node in the network has the power to alter information held within it. It also means that the record of transactions in each block that make up Bitcoin’s blockchain is irreversible. This also means that any Bitcoins extracted by a hacker can be easily traced by the transactions that appear in the wake of the hack.

Blockchain explorers allow anyone to see transactions happening in real-time.

Learn more about cryptocurrencies and blockchain

Whether you’re interested in improving cybersecurity or becoming a blockchain developer, looking for enhanced expertise in data science or artificial intelligence, specialist online Master’s degrees from University of York cover some of the hottest topics in these areas.

Discover more and get a step ahead with the MSc Computer Science with Data Analytics or the MSc Computer Science with Artificial Intelligence.

The next step in machine learning: deep learning

What is deep learning?

Deep learning is a sector of artificial intelligence (AI) concerned with creating computer structures that mimic the highly complex neural networks of the human brain. Because of this, it is also sometimes referred to as deep neural learning or deep neural networks (DNNs). 

A subset of machine learning, the artificial neural networks utilised in deep learning are capable of sorting much more information from large data sets to learn and consequently use in making decisions. These vast amounts of information that DNNs scour for patterns are sometimes referred to as big data.

Is deep learning machine learning?

The technology used in deep learning means that computers are closer to thinking for themselves without support or input from humans (and all the associated benefits and potential dangers of this). 

Traditional machine learning requires rules-based programming and a lot of raw data preprocessing by data scientists and analysts. This is prone to human bias and is limited by what we are able to observe and mentally compute ourselves before handing over the data to the machine. Supervised learning, unsupervised learning, and semi-supervised learning are all ways that computers become familiar with data and learn what to do with it. 

Artificial neural networks (sometimes called neural nets for short) use layer upon layer of neurons so that they can process a large amount of data quickly. As a result, they have the “brain power” to start noticing other patterns and create their own algorithms based on what they are “seeing”. This is unsupervised learning and leads to technological advances that would take humans a lot longer to achieve. Generative modelling is an example of unsupervised learning.

Real-world examples of deep learning

Deep learning applications are used (and built upon) every time you do a Google search. They are also used in more complicated scenarios like in self-driving cars and in cancer diagnosis. In these scenarios, the machine is almost always looking for irregularities. The decisions the machine makes are based on probability in order to predict the most likely outcome. Obviously, in the case of automated driving or medical testing, accuracy is more crucial, so computers are rigorously tested on training data and learning techniques.

Everyday examples of deep learning are augmented by computer vision for object recognition and natural language processing for things like voice activation. Speech recognition is a function that we are familiar with through use of voice-activated assistants like Siri or Alexa, but a machine’s ability to recognise natural language can help in surprising ways. Replika, also referred to as “My AI Friend”, is essentially a chatbot that gets to know a user through questioning. It uses a neural network to have an ongoing one-to-one conversation with the user to gather information. Over time, Replika begins to speak like the user, giving the impression of emotion and empathy. In April 2020, at the height of the pandemic, half a million people downloaded Replika, suggesting curiosity about AI but also a need for AI, even if it does simply mirror back human traits. This is not a new idea as in 1966, computer scientist Joseph Weizenbaum created what was a precursor to the chatbot with the program ELIZA, the computer therapist.

How does deep learning work?

Deep learning algorithms make use of very large datasets of labelled data such as images, text, audio, and video in order to build knowledge. In its computation of the content – scanning through and becoming familiar with it – the machine begins to recognise and know what to look for. Like the human brain, each computer neuron has a role in processing data, it provides an output by applying the algorithm to the input data provided. Hidden layers contain groups of neurons.

At the heart of machine learning algorithms is automated optimisation. The goal is to achieve the most accurate output so we need the speed of machines to efficiently assess all the information they have and to begin detecting patterns which we may have missed. This is also core to deep learning and how artificial neural networks are trained.

TensorFlow is an open source platform created by Google, written in Python. A symbolic maths library, it can be utilised for many tasks, but primarily for training, transfer learning, and developing deep neural networks with many layers. It’s particularly useful for reinforcement learning because it can calculate large numbers of gradients. The gradient is how the data is seen on a graph. So, for example, the gradient descent algorithm would be used to minimise error function and would be represented graphically as the gradient at its lowest possible point. The algorithm used to calculate the gradient of an error function is “backpropagation”, short for “backward propagation of errors”.

One of the most used deep learning models in reinforcement learning, particularly for image recognition, Convolutional Neural Networks (CNN) can learn increasingly abstract features by using deeper layers. CNNs can be accelerated by using Graphics Processing Units (GPUs) because they can process many pieces of data simultaneously. They can help perform feature extraction by analysing pixel colour and brightness or vectors in the case of grayscale.

Recurrent Neural Networks (RNNs) are considered state of the art because they are the first of their kind to use an algorithm that lets them remember their input. Because of this, RNNs are used in speech recognition and natural language processing in applications like Google Translate.

Can deep learning be used for regression?

Neural networks can be used for both classification and regression. However, regression models only really work well if they’re the right fit for the data and that can affect the network architecture. Classifiers in something like image recognition, have more of a compositional nature compared with the many variables that can make up a regression problem. Regression offers a lot more insight than simply, “Can we predict Y given X?”, because it explores the relationship between variables. Most regression models don’t fit the data perfectly, but neural networks are flexible enough to be able to pick the best type of regression. To add to this, hidden layers can always be added to improve prediction.

Knowing when to use regression or not to solve a problem may take some research. Luckily, there are lots of tutorials online to help, such as How to Fit Regression Data with CNN Model in Python.

Ready to discover more about deep learning?

The University of York’s online MSc Computer Science with Artificial Intelligence from the University of York is the ideal next step if your career ambitions lie in this exciting and fast-paced sector. 

Whether you already have knowledge of machine learning algorithms or want to immerse yourself in deep learning methods, this master’s degree will equip you with the knowledge you need to get ahead.

What is machine learning?

Machine learning is considered to be a branch of both artificial intelligence (AI) and computer science. It uses algorithms to replicate the way that humans learn but can also analyse vast amounts of data in a short amount of time. 

Machine learning algorithms are usually written to look for recurring themes (pattern recognition) and spot anomalies, which can help computers make predictions with more accuracy. This kind of predictive modelling can be for something as basic as a chatbot anticipating what your question may be about to something quite complex, like a self-driving car knowing when to make an emergency stop

It was an IBM employee, Arthur Samuel, who is credited with creating the phrase “machine learning” in his 1959 research paper, “Some studies in machine learning using the game of checkers”. It’s amazing to think that machine learning models were being studied as early as 1959 given that computers now contribute to society in important areas as diverse as healthcare and fraud detection.

Is machine learning AI?

Machine learning represents just a section of AI capabilities. There are three major areas of interest that use AI – machine learning, deep learning, and artificial neural networks. Deep learning is a field within machine learning, and neural networks is a field within deep learning. Traditionally, machine learning is very structured and requires more human intervention in order for the machine to start learning via supervised learning algorithms. Training data is chosen by data scientists to help the machine determine the features it needs to look for within labelled datasets. Validation datasets are then used to ensure an unbiased evaluation of a model fit on the training data set. Lastly, test data sets are used to finalise the model fit.

Unsupervised learning also needs training data, but the data points are unlabelled. The machine begins by looking at unstructured or unlabelled data and becomes familiar with what it is looking for (for example, cat faces). This then starts to inform the algorithm, and in turn helps sort through new data as it comes in. Once the machine begins this feedback loop to refine information, it can more accurately identify images (computer vision) and even carry out natural language processing. It’s this kind of deep learning that also gives us features like speech recognition. 

Currently, machines can tell whether what they’re listening to or reading was spoken or written by humans. The question is, could machines then write and speak in a way that is human? There have already been experiments to explore this, including a computer writing music as though it were Bach.

Semi-supervised learning is another learning technique that combines a small amount of labelled data within a large group of unlabelled data. This technique helps the machine to improve its learning accuracy.

As well as supervised and unsupervised learning (or a combination of the two), reinforcement learning is used to train a machine to make a sequence of decisions with many factors and variables involved, but no labelling. The machine learns by following a gaming model in which there are penalties for wrong decisions and rewards for correct decisions. This is the kind of learning carried out to provide the technology for self-driving cars.

Is clustering machine learning?

Clustering, also known as cluster analysis, is a form of unsupervised machine learning. This is when the machine is left to its own devices to discover what it perceives as natural grouping or clusters. Clustering is helpful in data analysis to learn more about the problem domain or understand arising patterns, for example, customer segmentation. In the past, segmentation was done manually and helped construct classification structures such as the phylogenetic tree, a tree diagram that shows how all species on earth are interconnected. From this example alone, we can see how what we now call big data could take years for humans to sort and compile. AI can manage this kind of data mining in a much quicker time frame and spot things that we may not, thereby helping us to understand the world around us. Real-world use cases include clustering DNA patterns in genetics studies, and finding anomalies in fraud detection.

Clusters can overlap, where data points belong to multiple clusters. This is called soft or fuzzy clustering. In other cases, the data points in clusters are exclusive – they can exist only in one cluster (also known as hard clustering). K-means clustering is an exclusive clustering method where data points are placed into various K groups. K is defined in the algorithm by the number of centroids (centre of a cluster) in a set, which it then uses to allocate each data point to the nearest cluster. The “means” in K-means refers to the average, which is worked out from the data in order to find the centroid. A larger K value is an indication of many, smaller groups, whereas a small K value shows larger, broader groups of data.

Other unsupervised machine learning methods include hierarchical clustering, probabilistic clustering (including the Gaussian Mixture Model), association rules, and dimensionality reduction.

Principal component analysis is an example of dimensionality reduction – reducing larger sets of variables in the input data without losing variance. It is also a useful method for the visualisation of high-dimensional data because it ranks principal components according to how much they contribute to patterns in the data. Although more data is generally helpful for more accurate results, it can lead to overfitting, which is when the machine starts picking up on noise or granular detail from its training data set.

The most common use of association rules is for recommendation engines on sites like Amazon, Netflix, LinkedIn, and Spotify to offer you products, films, jobs, or music similar to those that you have already browsed. The Apriori algorithm is the most commonly used for this function.

How does machine learning work?

Machine learning starts with an algorithm for predictive modelling, either self-learnt or programmed that leads to automation. Data science is the means through which we discover the problems that need solving and how that problem can be expressed through a readable algorithm. Supervised machine learning requires either classification or regression problems. 

On a basic level, classification predicts a discrete class label and regression predicts a continuous quantity. There can be an overlap in the two in that a classification algorithm can also predict a continuous value. However, the continuous value will be in the form of a probability for a class label. We often see algorithms that can be utilised for both classification and regression with minor modification in deep neural networks.

Linear regression is when the output is predicted to be continuous with a constant slope. This can help predict values within a continuous range such as sales and price rather than trying to classify them into categories. Logistic regression can be confusing because it is actually used for classification problems. The algorithm is based on the concept of probability and helps with predictive analysis.

Support Vector Machines (SVM) is a fast and much-used algorithm that can be used for both classification and regression problems but is most commonly used in classification. The algorithm is favoured because it can analyse and class even when there is a limited amount of data available. It groups data into classes even when the classes are not immediately clear because it looks at the data three-dimensionally and uses a hyperplane rather than a line to separate it. SVMs can be used for functions like helping your mailbox to detect spam.

How to learn machine learning

With an online MSc Computer Science with Data Analytics or an online MSc Computer Science with Artificial Intelligence from University of York, you’ll get an introduction to machine learning systems and how they are transforming the data science landscape. 

From big data to how artificial neurons work, you’ll understand the fundamentals of this exciting area of technological advances. Find out more and secure your place on one of our cutting-edge master’s courses.

 

What is neuromorphic computing?

Compared with first-generation artificial intelligence (AI), neuromorphic computing allows AI learning and decision-making to become more autonomous. Currently, neuromorphic systems are immersed in deep learning to sense and perceive skills used in, for example, speech recognition and complex strategic games like chess and Go. Next-generation AI will mimic the human brain in its ability to interpret and adapt to situations rather than simply working from formulaic algorithms. 

Rather than simply looking for patterns, neuromorphic computing systems will be able to apply common sense and context to what they are reading. Google famously demonstrated the limitations of computer systems that simply use algorithms when its Deep Dream AI was trained to look for dog faces. It ended up converting any imagery that looked like it might contain dog faces into dog faces.

How does neuromorphic computing work?

This third generation of AI computation aims to imitate the complex network of neurons in the human brain. This requires AI to compute and analyse unstructured data that rivals the highly energy-efficient biological brain. Human brains can consume less than 20 watts of power and still outperform supercomputers, demonstrating their unique energy efficiency. The AI version of our neural network of synapses is called spiking neural networks (SNN). Artificial neurons are arranged in layers and each of the spiking neurons can fire independently and communicate with the others, setting in motion a cascade of change in response to stimuli.

Most AI neural network structures are based on what is known as von Neumann architecture – meaning that the network uses a separate memory and processing units. Currently, computers communicate by retrieving data from the memory, moving it to the processing unit, processing data, and then moving back to the memory. This back and forth is both time consuming and energy consuming. It creates a bottleneck which is further emphasised when large datasets need processing. 

In 2017, IBM demonstrated in-memory computing using one million phase change memory (PCM) devices, which both stored and processed information. This was a natural progression from IBM’s TrueNorth neuromorphic chip which they unveiled in 2014. A major step in reducing neuromorphic computers’ power consumption, the massively parallel SNN chip uses one million programmable neurons and 256 million programmable synapses. Dharmendra Modha, IBM fellow and chief scientist for brain-inspired computing, described it as “literally a supercomputer the size of a postage stamp, light like a feather, and low power like a hearing aid.”

An analogue revolution was triggered by the successful building of nanoscale memristive devices also known as memristors. They offer the possibility of building neuromorphic hardware that performs computational tasks in place and at scale. Unlike silicon complementary metal oxide semiconductors (CMOS) circuitry, memristors are switches that store information in their resistance/conductance states. They can also modulate conductivity based on their programming history, which they can recall even if they lose power. Their function is similar to that of human synapses.

Memristive devices need to demonstrate synaptic efficacy and plasticity. Synaptic efficacy refers to the need for low power consumption to carry out the task. Synaptic plasticity is similar to brain plasticity, which we understand through neuroscience. This is the brain’s ability to forge new pathways based on new learnings or, in the case of memristors, new information.

These devices contribute to the realisation of what is known as a massively parallel, manycore supercomputer architecture like SpiNNaker (spiking neural network architecture). SpiNNaker is the largest artificial neural network using a million general purpose processors. Despite the high number of processors, it is a low-power, low-latency architecture and, more importantly, highly scalable. To save energy, chips and whole boards can be switched off. The project is supported by the European Human Brain Project (HBP) and its creators hope to model up to a billion biological neurons in real time. To understand the scale, one billion neurons is just 1% of the scale of the human brain. The HBP grew out of BrainScaleS, an EU-funded research project, which began in 2011. It has benefitted from the collaboration of 19 research groups from 10 European companies. Now with neuromorphic tech evolving fast, it seems the race is on. In 2020, Intel Corp announced that it was working on a three-year project with Sandia National Laboratories to build a brain-based computer of one billion or more artificial neurons.

We will see neuromorphic devices used more and more to complement and enhance the use of CPUs (central processing units), GPUs (graphics processing units) and FPGA (field programmable gate arrays) technologies. Neuromorphic devices can carry out complex and high-performance tasks – for example, learning, searching, sensing – using extremely low power. A real-world example would be instant voice recognition in mobile phones without the processor having to communicate with the cloud.

Why do we need neuromorphic computing?

Neuromorphic architectures, although informed by the workings of the brain, may help uncover the many things we don’t know about the brain by allowing us to see the behaviour of synapses in action. This could lead to huge strides in neuroscience and medicine. Although advances in neuromorphic processors that power supercomputers continue at unprecedented levels, there is still some way to go in achieving the full potential of neuromorphic technology.

A project like SpiNNaker, although large-scale, can only simulate relatively small regions of the brain. However, even with its current capabilities, it has been able to simulate a part of the brain known as the Basal Ganglia, a region that we know is affected in Parkinson’s Disease. Further study of the simulated activity with the assistance of machine learning  could provide scientific breakthroughs in understanding why and how Parkinson’s happens.

Intel Labs is a key player in neuromorphic computer science. Researchers from Intel Labs and Cornell University were able to use Intel’s neuromorphic chip, known as Loihi, so that AI could recognise the odour of hazardous chemicals. Loihi chips use an asynchronous spiking neural network to implement adaptive fine-grained computations in parallel that are self-modifying, and event driven. This kind of computation allows this level of odour recognition even when surrounded by ‘noise’ by imitating the architecture of the human olfactory bulb. The neuroscience involved in the sense of smell is notoriously complex, so this is a huge first for AI and wouldn’t be possible with the old-style transistors used in processing. This kind of discovery could lead to further understanding around memory and illnesses like Alzheimer’s, which has been linked to loss of smell.

Learn more about neuromorphic computing and its applications

Artificial intelligence is already helping us to make strides in everyday life from e-commerce to medicine, finance to security. There is so much more that supercomputers could potentially unlock to help us with society’s biggest challenges. 

Interested to know more? Find out about the University of York’s MSc Computer Science with Artificial Intelligence.

 

Why big data is so important to data science

What is big data?

Big data is the term for the increasing amount of data collected for analysis. Every day, vast amounts of unsorted data is drawn from various apps and social media, requiring data processing. 

Creating data sets for such a volume of data is more complex than creating those used in traditional data sorting. This is because the value of the data needs defining; without a definition it is just a lot of detail with no real meaning. Despite the term only relatively recently coming into everyday usage, big data has been around since the 1960s with the development of the relational database. It was the exponential rise in the amount and speed of data being gathered through sites like Facebook and YouTube that created the drive for big data analytics amongst tech companies. The ‘Three Vs’ model characterises big data by volume, variety, and velocity (with veracity and variability sometimes being added as fourth and fifth Vs). Hadoop appeared in 2005 offering the open-source framework to store big data and analyse it. NoSQL, the database for data without a defined structure, also rose in stature around about this time. From that point, big data has been the major focus of data science.

What is big data analytics?

Big data analytics is the sorting of data to uncover valuable insights. Before we had the technology to sort through huge volumes of large data sets using artificial intelligence, this would have been a much more laborious and slower task. The kind of deep learning we can now access through data mining is thanks to machine learning. Data management is much more streamlined now, but it still needs data analysts to define inputs and make sense of outputs. Advances like Natural Language Processing (NLP) may offer the next leap for data analytics, NLP allows machines to simulate the ability to understand language in the way that humans do. This means machines can read content and understand sentences rather than simply scanning for keywords and phrases.    

In 2016, Cisco estimated annual internet traffic had, for the first time, surpassed one zettabyte (10007 or 1,000,000,000,000,000,000,000 bytes) of data. Big data analysis can run into data sets reaching into terabytes (10004) and petabytes (10005). Organisations store these huge amounts of data in what are known as data lakes and data warehouses. Data warehouses store structured data with data points relating to one another that has been filtered for a specific purpose. These offer answers to fast SQL (structured query language) queries, which stakeholders can use for things like operational reporting. Data lakes contain raw data that has not yet been defined, drawn from apps, social media, and Internet of Things devices that await definition and cataloguing in order to be analysed.

The data flow of usable data usually involves capture, pre-processing, storage, retrieval, post-processing, analysis, and visualisation. Data visualisation is important because people tend to grasp concepts quicker through representations like graphs, diagrams, and tables.

What is Spark in big data?

Spark is a leading big data platform for large-scale SQL databases that leads to machine learning. Like Hadoop before it, Spark is a data processing framework, but it works faster and allows stream processing (or real-time processing) as opposed to just batch processing. Spark uses in-memory processing making it 100 times faster than Hadoop. Whereas Hadoop is written only in Java, Spark is written in both Java and Scala, but implementation is in Scala. With less lines of code, this speeds up processing significantly. 

Both Hadoop and Spark are owned by Apache after Spark was acquired from University of California, Berkeley’s AMPLab. Using the two in tandem leads to the best results – Spark for speed and Hadoop for security amongst other capabilities.

How is big data used?

Big data is important because it provides business value that can help companies lead in their sector – it gives a competitive advantage when used correctly.

Increasingly, big data is being used across a wide range of sectors including e-commerce, healthcare, and media and entertainment. Everyday big data uses include eBay using a customer’s purchase history to target them with relevant discounts and offers. As an online retailer, eBay’s use of big data is not new. Yet, within the retail sphere, McKinsey & Company estimate that up to 30% of retailers’ decision-making when it comes to pricing fails to deliver the best price. On average, what feels like a small increase in price of just 1% translates to an impressive 8.7% increase in operating profits (when we assume no loss in volume). Retailers are missing out on these kinds of profits based on a relatively small adjustment by not using big data technologies for price analysis and optimisation.

In healthcare, apps on mobile devices and fitness trackers can track movement and sleep, diet, and hormones creating data sources. All this personal data is fed into big data analysis for further insights into behaviours and habits related to health. Big data can also provide huge strides in some of healthcare’s biggest challenges like treating cancer. During his time as President of the United States, Barack Obama set up the Cancer Moonshot program. Pooling data from genetically sequenced cancer tissue samples is key to its aim of investigating, learning, and maybe finding a cure for cancer. Some of the unexpected results of using these types of data, includes the discovery that the antidepressant, Desipramine, has the capability to help cure certain types of lung cancer.

Within the home, energy consumption can certainly be managed more efficiently with the predictive analytics that a smart meter can provide. Smart meters are potentially part of a larger Internet of Things (IoT) – an interconnected system of objects, which are embedded with sensors and software that feeds data back and forth. This data is specifically referred to as sensor data. As more ‘Things’ become connected to one another, in  theory, the IoT can optimise everything from shopping to travel. Some buildings are designed to be smart ecosystems, where devices throughout are connected and feeding back data to make a more efficient environment. This is already seen in offices where data collection helps manage lighting, heating, storage, meeting room scheduling, and parking.

Which companies use big data?

Jeff Bezos, the founder of Amazon, has become the richest man in the world by making sure big data was core to the Amazon business model from the start. Through this initial investment in machine learning, Amazon has come to dominate the market by getting its prices right for the company and the customer, and managing its supply chains in the leanest way possible.

Netflix, the popular streaming service, takes a successful big data approach to content curation. It uses algorithms to suggest films and shows you might like to watch based on your viewing history, as well as understanding what film productions the company should fund. Once a humble DVD-rental service, Netflix enjoyed 35 leading nominations at the 2021 Academy Awards. In 2020, Netflix overtook Disney as the world’s most valuable media company. 

These are just some of the many examples of harnessing the value of big data across entertainment, energy, insurance, finance, and telecommunications.

How to become a big data engineer

With so much potential for big data in business, there is great interest in professionals like big data engineers and data scientists who can guide an organisation with its data strategy. 

Gaining a master’s that focuses on data science is the perfect first step to a career in data science. Find out more about getting started in this field with University of York’s MSc Computer Science with Data Analytics. You don’t need a background in computer science and the course is 100% online so you can fit it around your current commitments. 

The future of artificial intelligence

Artificial intelligence (AI) is the machine learning of tasks that we associate with the human brain – things like problem-solving, perceiving, learning, reasoning, and even creativity. AI has grown exponentially in recent years. The Covid-19 pandemic, in particular, highlighted the need for AI systems and automation that could respond swiftly to reduced numbers of workers. 

For organisations that had gone through a digital transformation, AI and associated emerging technologies were already being integrated into business processes. However, for many, Covid was the turning point that highlighted the need for AI solutions to be included in their business models. The AI cloud is a cutting-edge concept that will help make AI software more accessible to businesses by bringing together cloud computing and a shared infrastructure for AI use cases.

Healthcare offers many successful AI case studies, most recently for diagnosing and tracking Covid-19 using rapidly gathered big data, but also increasingly in things like cancer diagnostics or detecting the development of psychotic disorders. Other sectors that use real-world AI applications include the military, agriculture, manufacturing, telecommunications, IT and cybersecurity, and finance. AI art, or neural network art, is a genre in its own right. Holly Herndon, who has a PhD in Music and Acoustics from Stanford’s Centre for Computer Research, uses AI technology in her work.

What are the risks of AI?

Science fiction writers have long been fascinated by the idea of AI taking over. From Blade Runner to The Terminator, the fear is that the machines will start to think for themselves and rise up against humans. This moment is known as the ‘singularity’, defined as the point in time when technological growth overtakes human intelligence, creating a superintelligence developed by self-directed computers. Some people believe that this moment is nearer than we think.

In reality, AI offers many benefits, but the most obvious risks it currently poses are in relation to personal data privacy. In order for deep learning to take place, AI needs to draw information from large amounts of data that must come from people’s behaviours being tracked – their personal data. The Data Protection Act 2018, which enacted the general data protection regulation (GDPR), was brought in to ensure that people have to opt in to having their data gathered and stored, rather than having to make the request to opt out. Previously, businesses and organisations were able to simply use their customers’ data without permission.

Some of us may feel suspicious about our data being collected and yet, many of the applications we use are constantly gathering information about us, from the music we like and the books we read to the number of hours we sleep at night and the number of steps we walk in the day. When Amazon makes suggestions for what you might like to read next, it’s based on your purchasing and browsing history. A McKinsey & Company report from 2013 stated that 35% of Amazon’s revenue comes from recommendations generated by AI. AI is also instrumental in the way that LinkedIn helps both people to find jobs and companies to find people with the right skill set.

The more we allow our actions to be tracked, in theory, the more accurately our behaviours can be predicted and catered to, leading to easier decision making. New technologies like the Internet of Things (IoT) could help make this data even more interconnected and useful – a fridge that has already made a shopping order based on what you have run out of, for example.

Can AI be ethical?

There are certainly big questions around ethics and AI. For example, artificial neural networks (ANNs) are a type of AI that uses interconnected processors which mimic the human brain’s neurons. The algorithm for an ANN is not determined by human input. The machine learns and develops its own rules with which to make decisions, and which are usually not easily traceable by humans. This is known as black box AI because of its lack of transparency, which can have legal as well as ethical implications. In healthcare, for instance, who would be liable for a missed or incorrect diagnosis? If used in self-driving car insurance, who would be  liable for a wrong turn of the wheel in a crash?

When it comes to data analytics, there is also the issue of bias: because human programmers define datasets and write algorithms, this can be prone to bias. Historically, the field of data science has not been very diverse, which can lead to demographics being underrepresented and even inadvertently discriminated against. The more diverse the programming community, the more unbiased the algorithms, therefore the more accurate and useful AI applications, can become.

A popular example of problematic use of AI is deepfakes, imagery that has been manipulated or animated so that it appears that someone (usually a politician) has said or done something they haven’t. Deepfakes are linked to fake news and hoaxes which spread via social media. Ironically, just as AI software can clone a human voice or recreate the characteristic facial expressions of an individual, it is also key in combating fake news because it can detect footage that is a deepfake.

What are the challenges in using artificial intelligence?

Machine learning relies on data input from humans. A machine cannot initially simply start to think for itself. Therefore, a human – or a team of humans – has to pinpoint and define the problem first before presenting it in a computable way. 

A common example of what an AI robot cannot do – which most humans can do – is to enter a kitchen and figure out where all the items needed to make a cup of tea or coffee are kept in order to make a hot drink. This kind of task requires the brain to adapt its decision-making and improvise based on previous experience of being in an unfamiliar kitchen. AI currently cannot develop the data processing systems to spontaneously do this, but it is a situation that the neural networks of a human brain can naturally respond to.

What problems can AI solve?

Artificial Intelligence is mainly suited to deep learning which demands the scanning and sifting through of vast amounts of data looking for patterns. These algorithms developed through deep learning can, in turn, help with predictions. For instance, understanding a city’s traffic flow throughout the day and synchronising traffic lights in real-time can be facilitated through AI implementation. AI can also strategise. One of the milestones in the machine learning of AI systems was Google’s DeepMind AlphaGo beating the world’s number one Go player in 2017, Ke Jie. Go is considered to be particularly complex and much harder for machines to learn than chess.

On the practical side, AI can help reduce errors and carry out repetitive or laborious tasks that would take humans much longer to carry out. In order to increase the use of AI responsibly, the UK government launched the National AI Strategy in March 2021 to help the economy grow via AI technologies. Some of the challenges that are hoped to be addressed are tackling climate change and improving public services. 

In conclusion, AI has huge potential, but ethical, safe and trustworthy AI development is reliant on direction from humans. 

If you’re interested in understanding more about artificial intelligence, our MSc Computer Science with Artificial Intelligence at the University of York is for you. Find out how to apply for the 100% online course.