Simon Hoyle, iCetana Senior Vice President - Europe, answers SND's quick fire questions
Simon Hoyle, iCetana Senior Vice President - Europe, speaks to SND about Machine Learning and developments in technology including what may arise in the future given its potential.
How significantly has machine learning changed in recent years?
Machine Learning allows the development of systems that understand, learn, predict, adapt and operate autonomously. The combination of extensive parallel processing power, advanced algorithms and massive data sets to feed the algorithms have unleashed this 4th industrial revolution.
Machine Learning is gradually reshaping the world and making our lives easier. It is slowly becoming an integral part of our daily lives, and this happens because Machine Learning has become far more accessible to non experts in this field. For example, nowadays a software engineer without any acknowledge in Machine Learning can develop a model to classify house prices.
In my opinion, the democratisation of Machine Learning is the way in which this field has been changed so far in the recent years.
How rapidly has the technology developed in this field and what new possibilities may arise as a result?
In the last years, several commercially ML frameworks were developed, including the well-publicised libraries like Tensorflow by Google, Caffe2 by Facebook and CNTK by Microsoft. Recently, we also saw the major cloud providers Amazon Web Services and Google Cloud Services release ML as a Service platforms and graphics processor unit machines optimised for machine learning work.
In the hardware side, IBM has worked to release quantum computing processors, and Google announced this year the Tensor Processor Units (TPUs). Both technologies can lead to innovations by using ML to solve real-world problems. So I think new possibilities are in the direction of those technologies.
To what extent is machine learning changing the ways in which data is gathered?
ML changed the way in the data is obtained from social media, websites and other sources to help in the make decision task (e.g, in financial institutions), to help in the understanding of how some disease spread out (e.g., healthcare) and to create better recommendation systems to increase the user experience (e.g., video streaming provider), among others.
For example, Netflix employs ML to recommend movies by using the user profile experience, Facebook provides intelligent advertisements based on the user search, and a wearable can tell you when you need to drink water.
What are the prerequisites for utilising machine learning technology?
Nowadays, to create basic things in ML, you just need to know some of the modern programming languages (e.g., Python or R), have access to a good computer (with GPU, to deal with a large amount of data) and Internet access if you would like to use cloud technology.
However, if you would like to go deep in ML, you need to have a good mathematics and ML basis.
How reliable is machine learning technology? How easily can mistakes occur?
The reliability of ML depends on the understanding of the problem domain and the efforts in the features engineering step to represent the problem with the most significant key information. By having those clear, the reliability of ML might be increased.
In this sense, mistakes might happen if the model was not designed properly. Indeed, achieving a 100% safe ML system is impossible. However, with good development practice, it is possible to develop a reliable ML system.
Furthermore, efforts in fields such as adversarial learning and ML interpretability start to play an important role to increase the reliability of ML systems.
What industries benefit the most from this technology?
ML has risen to prominence in the recent years. Many new products today have incorporated ML, and it seems that more and more people appreciate these advancements.
ML is very much a reality as products for a variety of uses — surveillance, assistance, biometrics, speech recognition and manufacturing, among others. Many business establishments are also now adopting the technology.
How is machine learning implemented in financial services?
With more available computing power and more accessible ML technologies, such as cloud technologies and frameworks, there are more uses cases of ML in finance than ever before.
Currently, ML is applied many phases in the financial institutions, from providing loans approval, to devise new business opportunities, to deliver risk management to provide investment prediction and even to detect banking fraud as it is taking place. One of the keys in the development of this technology is it’s ability to give you more information today to utilise in this case video security surveillance than you ever had before – so there is a step change to create improved capability in the use of CCTV particularly in real time.
How is machine learning making the lives of ordinary people safer?
There has never been a more appropriate time to offer a step change in surveillance technology, its people and processes. The challenges faced in today’s society in Europe and beyond requires a more pro-active approach to the use of surveillance and everyone has a part to play. ML has been applied in surveillance systems for city CCTV schemes, healthcare, finance, among others to help the users to feel safer. For example, fraud credit card detection ensures the user credit safety, fingerprint authentication provides more security and privacy, and disease prediction helps to forecast the disease spread out.
ML is making the surveillance smarter by changing the way in which video is analysed. For example, face recognition is applied in several systems to user's authentication and to search government databases, and anomalies detection is applied to help in the security management. Being able to prevent escalation – provide medical assistance immediately or catagorising and managing safety risks all provide a safer environment through ML surveillance.
In healthcare, data is collected and is then integrated into ML to deliver more suitable patient care. With the advent of the wearables, which provide direct access to the patient, ML can lead life-saving situations.
In my opinion, MML techniques are interacting with us in all parts of our daily lives to make them safer.
What machine learning method is most frequently used?
There are several important techniques in ML; here I will emphasize two of them.
The first one is XGBoost (Extreme Gradient Boosting), which recently has been dominating applied ML and Kaggle competitions for structured or tabular data. This implementation is based on gradient boosted decision trees and it was designed for speed and performance.
The second one is Deep Learning (Deep Neural Networks), which has achieved tremendous success in the last few years, enhances the way artificially intelligent devices absorb data, somehow imitating the way human neurons transmit information around our bodies. It enables the software to learn to identify recurring patterns in sound, images, videos and another kind of data. Fields such as computer vision, finance and speech recognition have to take advantage of this technique to make fantastic, intelligent systems.
What sets iCetana apart from it competitors?
iCetana’s technology in ML has been patented in territories across the globe making it unique in the way it is able to provide machine learnt information applicable to 100s and 1000’s of feeds in real time and at scale.
It is a paradigm shift in the application of CCTV information to manage risks and therefore allows the Security Risk Manager and other professionals to capitalise on new or existing CCTV infrastructure providing a real return on productivity and risk management.
How easily can iCetana be intergrated with preexisting systems?
iCetana’s server based system can be easily integrated into an existing or new VMS, direct to the cameras or the NVR/DVR without impacting upon any of the pre-requisite requirements for camera, video quality or storage.
You were awarded the Milestone Beyond Security Partner of the Year Award...Tell me more about that?
We were delighted to receive this award from one of our valued VMS partners Milestone who recognised iCetana’s achievements in highlighting the capability of CCTV to provide much more than just security.
We are taking the use of surveillance to a much higher level in the management of risks to an organisation: disruptions in manufacturing or operating processes, serious illness/accidents or health and safety incidents and significant damage to property through water or fire damage are just some examples where the system has been used to reduce risk exposure and go beyond security.