Neural Network and Artificial Intelligence Application in Health Care, Labour, Sport, and Education.



Main topics of Illuminoo Enlights, the conference on AI achievements.

Bohdan Broviy

Written by:




The event took place near the Dutch town called Haarlem, at Bloomingdale beach. The beach is surrounded by Zuid Kennemerland National Park’s dunes from one side, and the Atlantic Ocean’s cold and windy North Sea from the other.




Guests and Speakers


Despite the weather, the event was held in the warm atmosphere of free and informal communication, speakers’ presentations were short and succinct, each one was followed by a break so that guests could discuss topics, talk to speakers, and get acquainted.

The organizing company’s business partners, AI researchers and developers, and entrepreneurs were among guests.

Session I:


Reflective Intelligence for Personal AI


Hosted by Artur d’Avila Garcez, scientific advisor at Illuminoo


Mohammed Shameer Iqbal, Research Assistant at Acadia University, Nova Scotia, opened the conference.

He talked about systems that can have a long-term learning plan using multi-modal deep neural networks. How neural networks can accumulate knowledge and how this knowledge can be transferred to another area of neural network understanding. How the network learning systems are built.


Mohammed’s speech prepared us for the scientific and heavy rhythm of the conference, but then there were examples of the theory application.



Mohammed Shameer Iqbal




Research Assistant at Acadia University


Program development manager at Singolar


River Level Prediction


The story about a system that predicted river level and flow change by collecting multidimensional data from different sources, so that the system that accumulated enough knowledge could make predictions based on previously revealed patterns, served as an example.



Visualization of Facial Expression Predictions




The development of a neural network that was trained by large volumes of similar data represented by photographs of people with different facial expressions, emotions served as a second example. This network has accumulated enough knowledge to predict and visualize various emotions of one person having only one photograph.


Deep Learning in Human vision

Mohammed explained how Deep Learning layers work. Everything is very simple and quite difficult at the same time. In order to perceive objects as humans do, AI must work as human vision. The first layer distinguishes image pixels and groups them according to a visual characteristic, the second one distinguishes pixel groups and classifies them as angles and outlines of the objects, the third layer can see separate parts of an object, for example, a person’s lips, eyes, ears. The fourth layer perceives a human face as a group of objects that consist of angles and elementary particles – pixels.



Следующим примером была презентация работы системы многомодовой глубоких нейронных сетей, а именно - создание аудиовизуальной ассоциативной памяти, которая может различить произносимые звуки, узнать в них определенную цифру и показать эту цифру на экране, и наоборот - воспроизвезти название цифры видя только ее рисунок.




Tarek Richard Besold

Researcher at the University of Bremen, Germany.


We found out about Human-like Computing from his presentation. These are systems that have a similar way of thinking as humans, they use the same input data for processing as well as produce results of data processing in the form understandable for humans.


Recent advances in machine learning and cognitive computing are pointing the way to broader opportunities of users’ access to the information, Tarek uses the term “Data Science 2.0” to describe this phenomenon


Data science has made great strides in processing large amounts of data (“Big Data”) and extracting basic information structures.


Recent advances in machine learning and cognitive computing are pointing the way to broader user rights to the information access and opportunities which we’d never dared dream before.


Tarek Richard Besold



Wissenschaftlicher Mitarbeiter (Postdoktorand) bei University of Bremen


The main thesis of the presentation was the term “Comprehensibility” which is symbolic knowledge clarity as one of the major differences between logical and statistical/neural approaches in computer science.

The problem of symbolic knowledge comprehensibility is correct assessment of this knowledge, programmes won’t be able to learn and gain experience without a proper assessment of data operations.



In order to assess the correctness of their actions, the criteria for assessing the work done are introduced in the system, – they are similar to an ordinary test – and AI itself decides whether the correct result of computation was reached or not.


Like humans, such programmes can evaluate the results of their actions and correct their own mistakes.


The score ranges from 0 to 100 points. If a programme’s result is over 80 points, we can assume that the chosen algorithm for data processing was applied successfully.


Artur Garcez is Professor of Computer Science and Director of the Research Centre for Machine Learning at City, University of London.


He talked about the scientific research in neural-symbolic computing, exemplified the neural-symbolic methodology, and highlighted several key developments which had enabled its application in practice.


Artur d’Avila Garcez


Professor of Computer Science at City, University of London


Artur calls the current changes in the subject area of neural-symbolic computing research Artificial Intelligence Revolution for a reason. He gives a list of activities, where AI working examples and best practices have been already applied:



education (adaptive learning);

finances (time series forecast);

security (image and speech recognition);

health (monitoring – IoT, meds design);

telecom (infrastructure data analysis);

games (interactive learning);

transport (logistics optimization).











There are two approaches to AI development:



Symbolic AI: a symbolic system has everything necessary for general intelligence


• Sub-symbolic AI: the intelligence emerges from the brain structure (neural networks)



The part of Artur’s lecture was dedicated to the explanation of Machine Learning (ML) principles, most of which concerned Deep Learning – ML based on AI learning by means of neural networks with the use of large data volumes.


It was exemplified that such networks can not only recognize objects with the application of logic programming basics, but also perform logical operations with them.






The basics of image and video streaming data semantic analysis were exemplified.
I.e. the understanding of data structures in the human sense. It can be achieved by applying multi-level deep nets that can create logical connections and perform calculations, the possibility of which was previously far from being real.


Session II:

The Game of Artificial Life.


Hosted by Jeroen Arendsen, design lead at Illuminoo

Lars Immerthal,

Witten/Herdecke University, Germany.


Unlike the first one, the second part of the presentation was philosophical – the topics were clear for the majority of guests and embraced the global issues of AI development in the human environment.

Dr. Lars Immerthal


Management Consultant – Independent Advisor and Research Fellow at the University of Witten-Herdecke


Dr Lars gives an example of “Scaffolding” that was first used in the construction of railway bridges. After the end of construction, it was demolished, leaving the ready metal structure. Similarly, a human being may become scaffolding for a more advanced form of Intelligence existence. Will we be able to control our creation, and do we need to? All these questions are mostly rhetorical, and it’s not yet possible to find the answer


The comparison of Information Technologies with primitive crystal forms was an interesting topic of the speech.


There are viruses that have a crystal structure, though, according to the scientific community, crystals are, by definition, considered to be an inanimate lifeform.


The fact that viruses can exist in the form of crystals seemed to be a proof that they are ordinary proteins, the representatives of inanimate matter. But the viruses contain the same structures that the genes of living creatures have.


If the presence of genes is a sign of a living creature, viruses are living matter. Of course, a lot depends on how we will define the term “life”. If we consider viruses from such a point of view, they are living creatures just as, for instance, elephants or human beings.



When can Artificial Intelligence be Thought Sapient?


Like in the case with crystalized viruses that have signs of life, a question arises here – When can artificial intelligence be thought sapient? Like in the case with animals, we can call certain animals sapient, it all depends on our definition of intelligence presence. We will get the answer to this question only when AI can gain experience, learn from its mistakes, develop itself, and be self-aware.


Agali Mert,

Rehabilitation doctor at CIRAN, the author of “Dit Spel Verandert je Leven” (“This Game will Change your Life”).


He talked about the examples of neural network introduction into human activities, such as Siri, Tesla Autopilot, etc. These systems can, to a certain extent, make decisions that previously lay with humans.

Agali Mert


Rehabilitation Physician at Ciran


What will Happen when AI Makes the Wrong Decision?

We’ve got used that AI, e.g. Siri, often makes mistakes. But what will happen if a person’s life depends on the AI decision, as, for example, with an autopilot having basic intelligence considered sufficient for making important decisions. The autopilot’s mistake is just a matter of time. How will we classify such a decision?


In the context of game design, we should introduce AI as an equal player to the decision-making system. How this player is perceived or taken by people depends not only on the people’s psychology, the environment in which the interaction occurs, but also on the role of AI, the new player. Successful introduction of artificial intelligence is a matter of intelligent design.


It is the humans who create AI, and only they decide what it will be like. Like with programmes, there are viruses, intentionally malicious software, and they are created to harm a computer, programmes, and eventually a human.


But we know that the majority of existing programmes are useful for people, and their utility can’t be compared to their shortcomings, they can make mistakes, but they do this unintentionally, and we are ready to admit such players to our life.



User Experience for Artificial Intelligence

Jeroen Arendsen is a lead UX designer at Illuminoo and a social care worker at Sherpa.

He talked about basic design principles that Illuminoo applies in its personal agents.



Jeroen Arendsen



Healthcare professional, Interaction Designer and Entrepeneur



Great Opportunities for Users’ Imagination and Gamification

Regarding interaction design, the most interesting approach here is “freedom” of mind. Since some AI programmes don’t have precise application instructions, there’s no need to make users follow the same thought-out script each time. It’s better to spend less effort on the interaction design and provide users with the opportunity to use the app in a way most interesting for them, to constantly devise new ways of interaction. To leave space for imagination and game.


Life Metaphor for AI

It is an approach of comparing AI to the life metaphor, what does it mean? – It is possible to apply life metaphors to the AI design, for example, AI is life, individual apps are living creatures, living creatures have subspecies, everyone has its lifecycle. They can grow, die, survive, reproduce, work. They have DNA, they can copy one another, inherit the parents’ features, and learn from their parents. They adapt to the environment, and the most effectively adaptable ones survive.

Session III:

Illumino’s Agents in the Field


Hosted by Leo de Penning, CEO of Illuminoo


In the course of this session, some of Illuminoo’s partners talked about the developed programmes operating in the areas of sport, business, health care, and education.


Some of the NOOMI agents are already populating the world. They are developing and adapting to our everyday life. They are starting to understand our personal needs, behaviour, and are trying to help us with useful information that will help us understand the world around us, make right decisions, learn from one another, and improve our life.




Magic Mirror

The agent developed at Illuminoo


Sylvia Loos




Sylvia Loos is a care adviser for people with severe disabilities at Sherpa.

Sylvia Bosma caregiver at Sherpa

Consultant Care and Support at Sherpa



Sylvia Bosma




They talked about the interactive Magic Mirror app that was developed for people with mental disabilities and how the first user of the app feels the effect thanks to the cooperation of Sherpa and Illuminoo.




Caregiver at Sherpa 


Magic Mirror developed by Illuminoo works with video data from a computer web camera and shows changing outlines of objects. Thus, it is possible to get a lot of various visual effects, while interacting with an “interactive mirror”.

The app is aimed at helping people with a low level of mental abilities development, it gives the opportunity to become self-aware in the interactive game. Despite its simplicity, the app has already produced positive results. After several lessons already, positive changes in the patients’ condition are observed. Unfortunately, we can’t show the photographs from the app presentation for privacy and ethical reasons.



AI-based Apps for Sport

Basketball, Tennis, Football, Volleyball, Baseball


Martijn Brouns, talked about the apps that were developed at his company and the area of their application.



Martijn Brouns





AI-based apps for sport can make predictions for the game strategy change, analyse streaming video data, make basic predictions of the game strategy and outcome.



In most sport games, the owner of the best game strategy wins in the context of equitable or almost equitable power distribution. AI-based sport apps know the game rules which are set in the system and are sort of restricting constants. But it’s not possible to learn to play well understanding the rules. In order to play well (or make game predictions), it is necessary to have experience in human understanding of the word. In machine understanding, experience is analysis of multidimensional large volume data – Big Data – with further processing and connection establishment, pattern detection. Patterns are impacts of various factors on the course of the game, the more of them AI knows, the more accurate predictions it can make in the area of their operation.


Energy industry


TESS (Trading Energy Support System)


Gas and electricity price predictions


Charel Hakkert is Head of Eneco Trading. He talked about a trading system of gas and electricity price prediction called TESS that Illuminoo developed for Eneco to support individual traders in making right decisions on energy purchase and sale.




Charel Hakkert





Head of Asset Backed Trading at Eneco Energy Trade


The app collects market data, weather conditions in the region to take into account all factors that may influence the price of gas and electricity. Together with the analysis of a large amount of information, the app builds interrelations, begins to understand what conditions may cause an increase or a decrease in the energy price.




A good trader costs a lot of money as they bring more. Their job consists in the analysis of all possible factors that affect or may affect the price of a product. As a matter of fact, a part of this job can be given to such agents as TESS, this way will be more efficient.


Coming to such conclusions, it is possible to assume that even good traders will be jobless in the end. But it’s not true, good traders should be involved in the design of price prediction systems, control of their operation, and making decisions on investment. In the future, when competitive companies start arming themselves with such apps, more and more resources for system design will be needed, thus we free employees from machine work – data collection and analysis – and transfer them to a new plane – design and development of new, more accurate methods of AI apps work.



Illuminoo Enlights 2017

The market is constantly changing; each consumer, employer, country feels that. It’s not enough to react to the trends, we need to predict and understand them as well. Artificial intelligence is a matter of time, not film directors’ fantasies. And the sooner we accept this, the better chances we have to succeed in this field. The representatives of the companies that attended the conference understand this and do their best to join efforts and be leading AI developers at early stages of development.


Edsson is a partner in software development in several areas and projects where neural networks, AI systems, complicated optimization algorithms are used. Thanks to the partnership with Illuminoo and the conference organized by them, we managed to get acquainted with engineers, researchers, and businessmen that realize the value of technology and push the market towards the application level. Thanks to new technologies, we make the world better.


Bohdan Broviy

Written by:

Edsson on Facebook

Edsson on Facebook