EN TR

Haber

Turkish Aerospace Preps the Path to the Future

Training AI Teams to Train AI Projects

Tarih: Issue 102 - December 2020

Times are certainly changing, and Artificial Intelligence (AI) or smart software solutions seem to be on the fast track to becoming ubiquitous, weather we are aware of it or not.  One could say imminent efficiency is at hand, enhancing many areas of our lives, impacting many sectors and industries. These are thought-provoking and exciting times indeed.  With the embracing of Industry 4.0 (and beyond, such as Japan’s society 5.0) we can see that areas such as the manufacturing industry widely utilizes AI systems, and this is providing benefits such as ensuring the seamless continuity of past learning and systems in changing and unpredicted environmental conditions. Artificial intelligence systems that have subsets such as machine learning, artificial neural networks, deep learning, pattern recognition, and have application areas such as autonomous planning and natural language processing stand out today as the most vital information technology element in digital transformation in the manufacturing industry, thanks to the developing hardware infrastructures as well as developing cloud solution algorithms supported by access to data. Even though it is difficult for human beings to break habits and change traditional perspectives in the century we live in, we will inevitably confront this reality with the broadening of digital transformation around the globe within decades. Some professions will disappear; on the other hand, the workforce will gain new skills and competencies, and we will witness the emergence of many new careers that we do not know today.   

Considering the technological developments and the resources allocated by countries and global companies, Artificial Intelligence, which is mainly utilized in Energy, Health, Automotive, Information Technologies, Education, and Defense & Aerospace & Security, will continue to grow rapidly. While the total investment made in artificial intelligence startups in 2010 was US$ 1.3 billion, it reached US$ 45 billion by the end of 2019.

According to the ‘Artificial Intelligence Global Market Size Report during 2019-2025’ published by Bloomberg on February 5, 2020, the size of the Global Artificial Intelligence Market in 2019 reached US$ 39.9 billion, and the market size is expected to boom by 46.2% annually to US$ 390 billion in 2025. While US$ 8.9 billion was allocated globally for artificial intelligence solutions in the field of Defense and Security in 2019, this figure is expected to rise to US$ 18.82 billion in 2025. According to data published by the EU commission, the European Union aims to increase its AI investments to € 1.5 billion by the end of 2020.

During 2010-2020, in the last decade, Turkey has invested US$ 3.4 billion in artificial intelligence projects. It was recently shared by the Minister of Industry and Technology Mustafa VARANK that TÜBİTAK has funded about 2,000 artificial intelligence projects related to AI such as deep learning, machine learning, decision support systems, e-commerce, and big data in the last decade.

According to OECD data, more than 50 countries have set their National Artificial Intelligence strategy as of the end of 2019. Turkey is not yet among these countries. However, in order for Turkey to take its place among these countries, the Presidential Digital Transformation Office, the Ministry of Industry and Technology, and TUBITAK held workshops on the National Artificial Intelligence Strategy and Artificial Intelligence Institute on February 19-22, 2020. Following such workshops, we can say that the efforts to set the National AI strategy with the participation of the public and private sectors as well as the universities have sped up, and the Minister of Technology and Industry Mustafa VARANK, who presented information about the 2021 budget activities of the Ministry of Technology and Industry at the Plan and Budget Commission on November 16, 2020, said that they would publish the National Artificial Intelligence Strategy very soon.

As part of these efforts in Artificial Intelligence, the Ministry of Technology and Industry had also published Turkey’s Industrial Revolution “Digital Turkey” roadmap in 2018 to increase the digital transformation of the country rapidly. In this road map, the focus was on Human Capital - the development of education infrastructure and the training of a qualified workforce, Technology – the development of technology and innovation capacity, Infrastructure – the strengthening of data communication infrastructure, Suppliers – the support of national technology suppliers, Users – the support of digitally transformed users, and Governance – the strengthening of corporate governance.

In line with this vision set forth by the decision-makers, our private companies and organizations in recent years have started to concentrate on investments in Artificial Intelligence and have initiated the establishment of new departments focusing solely on Artificial Intelligence and Machine Learning.

Turkish Aerospace is a pioneer in this particular area with the number of personnel reaching tens of thousands today in the Aerospace industry in Turkey, with the exports exceeding millions of dollars, adding value it reveals through innovative products it develops. Turkish Aerospace continues to increase its investments each day in support of Turkey’s future with technology, comprising artificial intelligence in particular. Well, how exactly is Turkish Aerospace preparing to build for the future now during this period where digital transformation has sped up, and Artificial Intelligence investments have become top priority?

In this interview, Turkish Aerospace Artificial Intelligence and Machine Learning Manager Güven Orkun TANIK discusses future focused activities launched within the company in the field of artificial intelligence, AI activities conducted within the scope of university-industry cooperation, training activities carried out within the company, their reflection on projects and the power of data to train artificial intelligence.

Enjoy the interview!

 

Defence Turkey: Mr. Güven Orkun TANIK, first off, we thank you for dedicating your time for this interview. Could you please tell us about yourself briefly?

Güven Orkun TANIK: First of all, thank you very much for this interview. My name is Güven Orkun Tanık. I graduated from METU Department of Computer Engineering. I also have a degree in Electrical and Electronics Engineering. I also have a MSc in computer engineering at METU. I have been continuing my postgraduate study for a long time. I hope I will have completed it within this year. After graduation, I started my career at Havelsan, worked there for about seven years, and then at Arçelik for about seven years. I have been at Turkish Aerospace for more than a year now. While working at Arçelik, I had established the company’s artificial intelligence department, and then I took part in the establishment of Turkish Aerospace’s artificial intelligence teams, and now I’m responsible for the management of this team. About one and a half years ago, our company decided to establish an Artificial Intelligence Department and took a significant step in this direction. I joined the company at the end of July 2019 or early August, and for over a year, we have been working with our team on various projects in-house and apart from the company.

Defence Turkey: You said that you started working in the field of artificial intelligence at Turkish Aerospace, a year ago. Can you share some details about the staff structure within Turkish Aerospace, and activities regarding cooperation with universities and other institutions, the internship training program, and your future vision?

Güven Orkun TANIK:  Finding manpower working in the field of artificial intelligence in Turkey is not an easy task. Hence, we usually strive to train our own teams. When I took office, the first thing we did was to establish a team by internal and external transfers and put this team through joint training and standardized the terms and concepts. Then, the team started to gain experience through training projects. So, we train our own team. We are currently open to internal and external transfers, and we will be doing so in the coming years. We have established very close cooperation with universities. This cooperation will also continue in the upcoming years. Out of these collaborations, we have made long-term cooperation with Hacettepe University. I am currently talking with you at Hacettepe University. We have a large office at Hacettepe Teknokent. The first artificial intelligence engineering department of Turkey was established at Hacettepe University last year. This was a substantial step for Turkey. This year we have seen that quite a number of artificial intelligence departments have been established in other universities. And that number will increase even more so in the following years. Our relationship with Hacettepe University was first built with the decision of our senior management to support this department, as a visionary move. With this support, we, as a company, employ the most successful 15 students in our office at Hacettepe as part-time personnel throughout the year (currently the number of staff as students is 30 in our department). You can think of it as a paid internship. They are here with us, they see and learn the work we do, and we train them. They share what they do at the university with us. Therefore, we build a very close relationship.

In addition to Hacettepe University, we also collaborate with METU, ITU, Ege, and Bilkent universities. We initiated joint studies with many universities in Turkey, and we continue to do so.

Defence Turkey: Today, artificial intelligence, machine learning, and deep learning are not merely used in the systems and subsystems of platforms in Defense and Aerospace projects. It is possible to use artificial intelligence in all critical processes within the company, from engineering design to development and production processes, from supply infrastructure management to project execution. Besides, artificial intelligence is utilized to reduce the workload of the staff and accordingly to use time more efficiently. What kind of a plan do you have regarding the integration of technical and administrative staff to AI-based systems along with the engineers within Turkish Aerospace?

Güven Orkun TANIK:  You asked a profound question; let me explain it under a few headings. The most important point is that all employees in the company must have basic AI literacy. They need to know what artificial intelligence is. I believe this is necessary to understand the limits of artificial intelligence and find potential artificial intelligence application fields. We have a training program that we have prepared jointly with the Turkish Aerospace Academy, and it will be launched soon. This will have different levels. There is an entry-level program that everyone can qualify for. There is a more detailed, advanced level for those involved in software because it is also essential to generate data in accordance with artificial intelligence in their software and in their field. We have the training to ensure that all company projects are adapted to a certain level of artificial intelligence. Furthermore, we have a training program for our own department. Apart from these, we are also preparing a training set that will enable our managers to correctly interpret opportunities, projects, suggestions, etc., with artificial intelligence.

So how do we generate projects? You mentioned management and project execution. In short, we can refer to artificial intelligence as programs and systems that enhance human capacity or reduce the workload, which we teach to make decisions like humans. In addition, to explain in-house, we need to go directly outside and identify and generate these projects. Especially at this stage, we need to define all kinds of projects that we think will be beneficial to our country, not just the company. Indeed, there are relatively basic projects; there are some that will create high value. Some will also generate relatively low value; in other words, they will not take the burden of people that much. We categorize them and choose the most valuable ones, the fastest ones, the ones in which we need to invest heavily, and we set our own roadmap accordingly. We currently have a roadmap covering the period until 2025. We have a project portfolio; we have projects that we make and plan. This does not mean it is fixed and what we will do by 2025 is unchangeable because artificial intelligence is developing so rapidly. Let me tell you how rapidly it progresses; the AI algorithms we used in a project were revealed four months ago. Another AI algorithm did not exist a year ago. Artificial intelligence is a science that progresses very, very fast. Therefore, it is not possible for us to plan clearly what we will do within five years from now. The limits of what we can do are continually changing. We update our roadmap several times every year, and we choose and execute the projects accordingly. Meanwhile, it is necessary that we conduct our own business development for project selection and execution. It is our responsibility to find stakeholders both internally and externally, meet with them, understand their requirements, raise the right questions, and produce satisfactory answers to these questions.

Defence Turkey: You have touched upon your roadmap. Could you give us some details about the pilot studies you have launched within Turkish Aerospace on this issue? 

Güven Orkun TANIK: We certainly have pilot studies. I cannot go into much detail, but I can certainly give you a brief explanation. We work on a data infrastructure system. The data infrastructure is critical because data is quite critical, particularly for two of the three main areas of artificial intelligence. Therefore, we need to create a clean data infrastructure where the data we generate within the company can be found, accessed, used, and interpreted by those who need it and have authorization. Establishing such data infrastructure may seem uncomplicated but let me give an illustration: two developers can give two different names to the same thing, and when you name it differently, you can even write such a different name in different ways. You use an abbreviation, or you write it in another way. For example, a piece has an identity number; one writes “ID No,” the other writes “Identity No,” and the other writes “No.” If you cannot even singularize its name, it is impossible to access such data and collate them. It seems quite easy, but terabytes of stored data of companies cannot be easily accessed, found, or used due to such cases. We have a data infrastructure project aiming at the remedy. This data infrastructure project currently runs on our flight and test data. We have a system where we work on the proper storage, quick access, and the processing of them. As I previously said, data is crucial to train AI projects. What we refer to as artificial intelligence does not work without data. That is why data is dubbed as oil of the digital era.

We have image processing projects among our other projects. Image processing has a wide range of applications, from production to tests in production, from facilitating the work of employees to systems on platforms. As you know, there are optical systems on the platforms, and image processing is also performed on them. What is more, the image processing is not completed only by processing the image; a decision must be made. Therefore, our second domain is image processing and decision support systems. We develop systems that interpret what you see and make decisions on it. In the meantime, we also worked on a system that measures people’s cognitive load as a training project. We worked on a system that tries to understand emotions. These training projects were developed for the team to gain experience in artificial intelligence, and they have areas of use in our prospective projects. This is one of the traditional study fields of the defense industry, i.e., to measure the cognitive load of the pilot. When we measure the cognitive load, we can reduce the problem of overloading.

There are also autonomous systems. Autonomous systems are the third pillar of artificial intelligence, which I previously touched on. They are based a little more on a simulation experiment. Because here, there is an AI agent moving in an environment, which could either be a simulated one or a real one. This agent observes its interactions with the moves it chooses. In other words, it observes what is happening in the environment and learns to behave accordingly. It is like training your pet; you reward what he does well, punish what he does wrong, or don’t give any reward, then he learns to behave in a way you like. The key is that he learns to generalize these behaviors. When he can generalize behaviors, the main force arises from there. We have a separate system for this. The frequently referred to DARPA ACE system is actually based on this.

Defence Turkey: Before moving on to hardware and software infrastructure, I would like to talk about the new composite facility you are going to launch. This will become a facility to focus on Industry 4.0 or even 5.0. When we say Industry 4.0, artificial intelligence is an integral part of it. We speak of a place where there are unmanned systems, a virtual factory established with various sensors and recognition systems, which turns into a simulation. What contribution will your Artificial Intelligence and Machine Learning Department make to this composite facility?

Güven Orkun TANIK:  The principles I just mentioned apply to any, and all, whether it is a factory or a platform. Any flying platform continually generates data during all its tasks from the roll-out to the moment it flies, and even afterward. It is possible to use this data you generate at any point. Whether it is with artificial intelligence or with modest control systems, or even with creating historical data and then making a statistical approach, it is possible to offer benefits in many ways. Almost every segment of the newly established factories you just touched on is in actuality built on a data repository. Then you add on a control system, an artificial intelligence system, or new designs to improve quality. If there is data, you can work on it; for example, it can be used to improve designs, it can be done in any field you can imagine. If there is no data, there is nothing you can do. We make our system designs, starting with data. Efficiency and performance improvements are achieved over this.

Defence Turkey: Certainly, there is a need for hardware and software infrastructure to generate such data and to store this data within the company. Do you use the systems you have developed for this purpose, or do you use foreign open-source systems?

Güven Orkun TANIK:  If we talk about hardware, the basic training and operating systems of artificial intelligence systems that are present in the hardware are supplied from abroad. If we talk about software, we develop our own software using open source libraries. We choose the libraries we use from the systems in which we have access and control over everything. One of the main reasons for this is that we are not just using them in the factory, etc., so when we need to upload them to a platform, there are certification processes to be followed. If you don’t have 100% control of the software, you cannot achieve this process, or if we are producing/developing something that is an important part of our national security, we need to have control over everything.

Defence Turkey: The defense industry companies in Turkey and other foundation companies are also performing various activities in the field of artificial intelligence. In view of that, what would you like to say about your collaborations with other companies? Besides, you also mentioned that you have efforts to hire qualified human resources for the company to reverse the brain drain. Do you have a roadmap to bring such high-skilled professionals back to the country?

Güven Orkun TANIK: With regards to the reverse brain drain, you know there are valuable skilled people who have already returned from abroad by completing their higher education programs supported by our government. There is an ongoing activity in which our company also takes part for their engagement. In addition, we have our own efforts. We have colleagues who want to return even though we have not performed any specific effort for this, and of course, we welcome them. We meet with all the other foundation companies as a joint effort. We already have close cooperation with all foundation companies. There have been preliminary talks so far in terms of artificial intelligence. There is no specific initiative launched, but I believe it will be launched very soon.

Defence Turkey: Artificial intelligence and smart systems have grown to be very popular in today’s world. So, it may even be possible to establish a new artificial intelligence cluster, perhaps under your leadership in the upcoming years. Surely, you follow the developments in the world closely. You just mentioned, with the novel algorithm that emerged 4 months ago, a completely different algorithm emerges all of a sudden and maybe you need to get organized accordingly. When you look at the global developments in this respect, what is the current position of Turkey? If we specifically speak of Turkish Aerospace, what is the status of your company in integrating Turkish Aerospace’s artificial intelligence applications to platforms? Are we at the infancy stage or are we able to progress and pick up the pace in a short period of time?

Güven Orkun TANIK: Artificial intelligence is an interesting technology. Unlike many of today’s technologies, it has developed mostly in the civilian world. Therefore, following the developments in the world is very critical for us. This technology is developing substantially in private companies and academia. It is one of our duties to accurately transfer what we deem appropriate from these developments to the defense industry. Thus, we keep up with this technology very closely. We regularly acquire new technologies, try them, test them, and learn from them, and we even try to contribute to developing new technologies. We, as Turkey, have joined the game in time. We have made investments in this regard, both as a country and as a company. The Digital Transformation Office of the Presidency of Turkey has published a roadmap regarding artificial intelligence at the beginning of this year. Our state, our government has a strategy in this field. The Presidency of Defense Industries has also been exerting efforts for years. As a country, we create focused technology networks, plan the projects for the next decade, determine the investment areas, foresee where the technology needs to be developed, and invest in these areas. So we are at the adolescence stage, getting ready to grow, let alone the infancy stage. Geoffrey HINTON’s unveiling of Deep Learning in 2012 has led to a boom in Artificial intelligence in the world because the facts remaining academic until then started to flow into the real world. Thereupon, a very rapid development period, was experienced. According to the research of GARTNER, its peak will be in 2030. In fact, it is not possible to exactly say what we will do in 2030, because as I previously mentioned, artificial intelligence is advancing dramatically. So in 2030, I hope we will be ahead of the wave, but unfortunately, it is not possible to say anything definite right now. 

Defence Turkey: Do you have a goal to apply Artificial Intelligence, Deep Learning and Machine Learning in your current projects? How will they evolve in the following years? At Teknofest Istanbul 2019, I had the opportunity to see the fully AI-driven and high-tech virtual cockpit system, which is planned to be used in the 2040s in National Combat Aircraft, with a VR helmet. You are already trying to perform studies that foresee the needs of the 2040s. What would you like to say about this?

Güven Orkun TANIK:  Artificial intelligence will be used in almost all our platforms, whether we like it or not. The reason why AI is to be used is that, as I said at the beginning of our conversation, AI has two main functions. One is to reduce the workload of people; the other is to raise people’s skills, to help to make things that they would not normally be able to do, easier. Both are very crucial on any battlefield. In other words, the more you can reduce the cognitive or physical load of a person you assign for decision making, the more focused and accurate decisions that person can make and accordingly make better moves. Likewise, any system that will make things easier to do that they cannot normally do, or a system that does it automatically, will surely be inevitable. Why will it be inevitable? Because humans can get tired, distracted, our eyes can overlook or miss something. We especially have difficulties in repetitive, dull tasks. Again, we have biological limits, our reaction time can be low, and our body has a certain G-force that it can withstand. Artificial intelligence will be of great benefit for us to overcome these limits or to reduce their restraining effect on us. Therefore, for the answer to the question of where we can integrate it, we can say to almost everything that comes to mind. If we ask where it will reach, it is a very speculative issue. I think it will go beyond the dreams of even science fiction authors. It will certainly be utilized on the national combat aircraft.

Defence Turkey: Artificial intelligence provides significant advantages for decision support mechanisms, especially in air platforms. From this perspective, during joint air operations involving air elements that are now operationally used and are a strategic power multiplier, such as F-16s, tanker aircraft, airborne early warning aircraft or UAVs, what kind of capability enhancements will artificial intelligence provide in terms of generating operational area pictures or achieving decision support in command-and-control systems?

Güven Orkun TANIK:  I just talked about simulation, for example. Data and simulation go hand in hand with artificial intelligence. Artificial intelligence can generate data from data. Generating data from data can actually add broader information to anything we see. Artificial intelligence increases the value of data. The value of standalone data is relatively low. Artificial Intelligence also is able to combine data. The more data gathers, the stronger it gets. Let’s say you have one eye, you see something, and when you have a second eye, you see the same area again, but when you combine what you see with two eyes, you start to have distance perception. If you think of data like this, the more you gather data, the more the possibilities increase. Artificial intelligence has the opportunity to gather all data by using the data infrastructure underneath. It is possible to see the future with a simulation in the field of command and control. By seeing the future, I mean the simulation has the ability to offer the most possible future to us and illustrates the results in whatever path you choose, whatever move you make. Therefore, I think this will be very crucial in the following years.

Defence Turkey: Our indigenous projects such as the National Combat Aircraft (MMU) and HürJet Projects, will become operational in the 2030s. What would you say about the plans for the use of artificial intelligence in these projects? Will artificial intelligence have a significant role in the National Combat Aircraft?

Güven Orkun TANIK: I will give a brief answer to the National Combat Aircraft issue; I cannot go into much detail. Regarding HürJet, we have a training simulator that has appeared in the media before. In this system, you can think there is an AI agent for the training of pilots. First of all, it is possible for pilots to fulfill these training and operations with more realistic virtual agents next to them. Under normal circumstances, when the pilot enters the simulator cabin in a simulator environment, the person managing the simulator positions friend and foe elements within the place where the pilot is positioned. They act according to the predefined rules taking part in the plan. For example, move around that target, engage when the enemy is within that range, lock a missile, etc. These usually operate with very simple rules. When you say artificial intelligence system, it is possible to offer a more adaptable, more dynamic, more flexible, more realistic environment. It contributes in this way to the simulator environment.

Defence Turkey: Considering unmanned systems such as the ANKA UAV and AKSUNGUR, which are expected to be included in the inventory in the near future, we also need to talk about the use of artificial intelligence in such systems. What would you say about TUSAŞ’s use of artificial intelligence in unmanned systems? Could you explain the role of artificial intelligence in unmanned systems, especially about data, image, and signal processing and its contribution to the Decision Support Mechanism?

Güven Orkun TANIK:  Let me explain it in this way, when we say UAV, there are no humans in the UAVs of the previous years, but there is a person in the background. They are controlled by people, pilots… It is not possible for such people to feel like they are inside the platform. So when you are on the platform, you can feel and see all kinds of things, such as the sound of the engine. Otherwise, there is a distance between you and the platform. Likewise, these systems have many sensors on them. They obtain data from images and signals, and they are extremely high in number. Humans are only able to pay attention to three to seven different things simultaneously; we have cognitive limitations. Sometimes there are thousands of sensors in these systems. Therefore, the most important role of artificial intelligence here is to reveal and highlight important information. AI applications allowing us to determine which of the signals is important, which one should be tracked, or those facilitate tracking - you can even think of it as drawing a circle around the person in any image - are the current usage areas on unmanned platforms. When we talk about artificial intelligence and aviation, the autonomous systems you just mentioned that can decide on their own are still the subject of the future. Current aviation certifications also do not grant permissions to fully autonomous systems. Besides, the ethical and legal aspects of the use of autonomous systems are not yet mature enough. Therefore, artificial intelligence is currently functioning as a support system.

Defence Turkey: Nowadays, ‘hybrid warfare’ has started to show up as a concept, where manned systems and unmanned systems are employed together or where unmanned systems that can be controlled from a manned platform are used. Do you have any plans and activities regarding this matter?

Güven Orkun TANIK:   Hybrid warfare will be inevitable as the systems we have previously mentioned get involved. Old systems will not disappear immediately, and these systems are now increasing people’s skills. Unmanned systems utilized and controlled from manned systems are being developed by the 

armed forces of various countries in the world. Therefore, while there is a second pilot using the weapon system in relatively old systems, for example, artificial intelligence may take over some of the roles. I cannot go into much about the plans.

Defence Turkey: One of the important issues regarding the use of Artificial Intelligence is that unmanned systems are now being used strategically in the combat environment to a great extent. On the one hand, the unmanned fighter jet concept is gradually coming to the fore. The trials in which a human F-16 pilot competed with artificial intelligence in a simulation environment hosted by DARPA in the previous period resulted in the superiority of artificial intelligence. This was actually an indication that artificial intelligence could achieve a superhuman level. Can I have your comments on this matter? What is the position of Turkey in this area? Is it possible to cooperate in the future with foreign companies performing in this specific field? How should we approach this issue?  

Güven Orkun TANIK: I think DARPA’s work was very, really inspiring. We examined this very closely and even conducted a study on it, reviewing what they did and what they could not do. It flawlessly defeated the pilot. While Artificial Intelligence defeats the pilot, the systems running behind would not actually be sufficient in a real environment if you put them on a platform as they are. Because they ignore many things, they function there with artificial intelligence and excellent knowledge. Maybe you also watched it; while the pilot was looking for the enemy fighter jet in the cabin while he was wearing a virtual reality headset, the AI system clearly knew the position of the pilot’s aircraft and acted accordingly. In fact, the pilot defended himself extremely well, but the AI system won this competition because it didn’t care much about its own safety and not observe the limits of any platform, and because it had the ability to fly precisely and its reflexes were much better than the pilot. This shows us that in the following years, as these systems mature, they will move into areas such as dogfights, where AI has not been able to enter so far and will be much more effective there. What are we doing? As a matter of fact, we have been working on this for a long time. One of the first studies we started here was a similar system. When we examined the system made by DARPA, we saw that we made many similar choices, and it made us happy, actually. We did not become distinct at certain points; we were common ground. How should we approach it? Unfortunately, it is not possible to import such technology; therefore, we should make our utmost effort to develop such a system indigenously.

Defence Turkey: Mr. TANIK, thank you very much for your time