I have always heard, wondered, understood, applied, and developed upon the following terms;
Pick and place, feeding machines (robots), boxing robots, assortment box-filling robots, automation, input/output, PLC, workstation, SCADA, operation research, supply chain, factory automation, etc.
The machines and systems that make all of them possible:
Belts/pulleys, reductors/variators, gearboxes/cams, AC/DC electric motors, drive cards, cascade systems, machine communication systems, industry 4.0, and IoT, etc.
And what else, and what more…
When I look back, these are the terms that I remember making all the effort to find someone who knows some of these, convincing them to teach me or flying across continents to just experience one of them.
However, these terms are unfamiliar even to freshly graduated engineers and such machines are now obsolete if we disregard some exceptions. Mechanical, electronic, computer engineering is now done by combining parts like LEGOs without getting their hands dirty. I do not want to be misunderstood, because I do not contempt them; in fact I admire and appreciate them while expressing this fact. Yet, in the past, we could only open and tamper with the hood of the car stranded on the side of the road, now we can access the motor which only has a plastic cover, through a computer.
Yes, who will be the masters of the future? The nations that have money and use all of these or the poor who produce them for cheap prices? Or will it be those who create the ‘know-how’?
I have always wondered how machines work. I would immediately take apart my toys with curiosity, and then, of course, they would either remain in pieces or the number of pieces would outnumber while collecting; my goal was to do something new. I remember when I was young, I unpacked the new arrived machines, assembled them and watched them while they were commissioned. It was such an excitement, such a pleasure. My father used to take me on business trips. There was no trip we couldn’t take to see a new machine work. Fifty years ago, we went to Scotland by plane, train, and car. The factory owner welcomed us with curiosity and was amazed to see that we were not dressed like Turks his grandfather who fought in Canakkale had told him about.
Anyway, let’s leave these memories now and come to our topic.
ARTIFICIAL INTELLIGENCE (AI)
As you can appreciate, the necessity and practice of artificial intelligence cannot be understood without knowing the reason for automation and robotization needs.
Behind all these, the human mind and practicality exists. In my opinion, the human is utilitarian, lazy, and malefic. The human has invented machines, automation, robots, and finally IoT and AI for his own benefit and convenience. But of course, it is cruel to treat humans, who are perfect beings, like a machine or a part of a facility.
This inevitable change has always been opposed because it is unknown, incomprehensible, and frightening to most. But you know, what happened to porters when pallet trucks and forklifts came out or to horse carriages when trucks came out, and what happened to cabinet workhorses when the engines were invented.
In short, new professions have always been invented (1), and employment has increased. Thanks to the increased production and productivity, income levels and the quality of life have increased. Humans have continued to chase their whims and complain.
Now, we are facing a major change again. We will either change or change!
In other words, we will modernize by putting artificial intelligence at the center of social life and shaping our education and work accordingly.
So, when did all this begin in Turkey? At a Public Conference where “Artificial Intelligence” was explained in Erzurum in 1957-1958!
If we consider our ability to recognize, understand, and interpret our environment with our senses what we define as “intelligence” among us; “artificial intelligence” is a little exaggerated. First of all, a distinction must be made between “recognition” and “judgment”. Machines and robots are now very good at observing and recognizing our environment, making patterns from them, and making predictions; because the image, recognition systems and statistical estimation methods based on big data are highly developed.
Artificial intelligence (machine learning, deep learning which is a form of it) is a highly technical, computer science-related topic. From time to time, many executives from IBM have expanded our horizons by introducing the features of Watson, an expert artificial intelligence program. If you go to Amazon today and search for an Artificial Intelligence book, there are over 30,000 books written or to be written in two years on related topics. There are 30 thousand books on “Big Data” which have not yet been fully defined. These numbers reveal the extent of the future impact of Artificial Intelligence. However those who write the subject in a way that the operators can understand, are not more than the number of fingers of a hand. After reading them, many say that something is still missing (1, 2). Many business articles written on artificial intelligence write that they do not understand the computational logic of the job well enough, but I think this is an effort to hide some moral problems. I will write about the reason when we come to it (3, 4).
I think computer scientists are looking at neuroscience and cognitive science research and trying to imitate the human mind and solve a problem given to them. For example, when you give a voice command “Go to Sariyer” to a device, outputting the route from Beykoz to Sariyer as text or voice alternately in time. How does that happen? You need to teach the machine a lot of traffic data, road data, speed limits, and the machine must detect the current situation from the data and tell you the error-free result. In the meantime, let me state that in the field of artificial intelligence, the cost of making faulty software and making software without errors is the same. So, you can’t get a better one for cheaper. In other words, artificial intelligence software has only one KPI: correct guess (5). Returning to our traffic example above, the machine needs to be constantly trained (updated), new roads and intersections, dead-ends, and changing street names need to be constantly transferred to “intelligence”. What if, suddenly, the streets become one-way, road maintenance begins, or the route changes? That’s the time when the machine (software/algorithm) makes an error. Man finds a way with his own mind. To improve the example; software, machine coding, artificial learning, artificial intelligence, whatever it is, by looking at past data, it will predict which road the municipality will repair, which road will be made a dead-end, which road will change its name. Moreover it will predict the likelihood of the road being blocked as a result of an accident and now it will give 100% accurate estimates. We will see whether it is possible for the machine to sense the behavior of the municipality.
In 1956, ten scientists who were interested in machine intelligence came together for a six-week workshop at Dartmouth College in New Hampshire. Their purpose was to discuss how machines can simulate human intelligence characteristics such as sensation, logic, decision-making, and the ability to predict the future. The organizer of the meeting, US mathematics professor John McCarty (1927-2011), used the term artificial intelligence for the first time to attract some attention when looking for a name for the conference where the results of the meeting would be discussed a year later.
In fact, even today, the software called artificial intelligence, contains McCarty’s basic idea. McCarty argued that human thinking and reasoning can be defined mathematically like gravity rules defined by short equations; consequently, he argued that the abstract moment, thought, and logical thinking could be transformed into algorithms, explanatory instructions that guide problem-solving processes. The organization that funded the workshop was the Rockefeller Foundation.
This is exactly why I say that, this job cannot be solved with the known 0/1 computer software mathematics, the “maybe, sometimes, let’s see options” should also be included in the work, fuzzy logic should be utilized much more. Of course, we do not take into account the conscience aspect of the matter. If someone connects to websites with explicit videos from their computer after midnight, and next day AI keep showing him these explicit videos, here’s what I’m talking about: atrocity!
In 1958, Cahit Arf, our world-renowned mathematician held a conference on Artificial Intelligence, in Erzurum at Atatürk University entitled “Can a Machine Think? and How Can a Machine Think?”. It made me very proud to learn that he gave such an entitled conference. At this conference, Arf attempted to convince the audience that machines could also think like humans, and concluded his speech with the following sentences (6):
“Although machines can do some work much faster than the human brain, the understanding, that is, even those whose reception capacities are large enough to fill a large hall, are much lower than the human brain in terms of variety. In contrast to the human brain being improved by itself, the machine remains as it was built. However, it is possible to design a self-evolving machine. However, in my opinion, the main difference between the human brain and the machine is that the human brain can take and process aesthetic agents and make decisions that are of aesthetic nature, and feel free to do a job or not, but the disappearance of similar qualities in the machine. What characterizes these characteristics is that they all contain an element of uncertainty, the absence of the rules to which they are infallible. There are non-human natural phenomena that have the character of uncertainty.”
Today, artificial intelligence applications are everywhere. Siri takes our commands and chats with us, making us laugh with her responses. But it also scare us with its smart answers. Alexa and Google Home also serve people in home management and shopping by voice command. Autonomous vehicles are redefining automobiles and transportation. There are automatic algorithms that can buy and sell stocks at a speed that traders cannot catch up to, and programmatic software that shows advertisements at a speed that people cannot keep up with. Before arriving home, the ambient temperature is adjusted, Google Maps offers alternative ways by checking the traffic, unnecessary e-mails fall into junk, Instagram and YouTube constantly recommend new videos according to our interests, while Netflix offers movies. All of this is software that McCarty and his colleges initially sparkled and it imitates the human thinking system.
Machines can now make medical diagnoses and prevent fraud. There is even a joke which states that: If the machine manages to perform a task that could only be done by humans before, that task does not require intelligence! Yet intelligence is a very abstract thing. What is intelligence? What is learning? After all, computers are still not far from Pablo Picasso’s thought that “They have no use, they can only respond to you” (8). But computers started to respond very quickly by performing very fast operations (data analysis that human intelligence cannot reach) on more subjects.
I decided to read some and gather what I know and write this article, when, I saw my name mentioned in a YouTube video, which belongs to a phenomenon, announcing a technology named GPT (Generative Pre-Training Transformer (3). Why was my name mentioned? My answer is coming soon.
As far as I understand, Open AI’s GPT3 Natural Language Programming technology, founded by Elon Musk and Sam Altman, is a paradigm-breaking software and will affect “business artificial intelligence” applications the most in the new period, and artificial intelligence applications and solutions of this period are more than image recognition. It is said that it will turn to language processing (9). The GPT3 can write press releases, newspaper articles, and technical manuals when it is given an intention. Also it can write these articles in the person’s own style. Moreover, GPT3 can write computer code, an algorithm. So, you no longer need an expert software developer to solve a problem, GPT3 can do it. Currently, the system is not without errors, but Microsoft, Google, Alibaba, and Facebook have immediately begun working on their versions.
As far as I see, the thing that has pushed the development of artificial intelligence applications so much is the “Commercialization”. Large companies fund basic research on the latest technology, assuming that commercially significant breakthroughs will occur in a short period. In fact, everything accelerated with the reorganization of Google, Facebook, Apple, Amazon, Microsoft, and Baidu, who have billions of data in their hands, around artificial intelligence. As I said in the beginning, nobody cares about building smart minds and creating perfect people. Artificial intelligence is seen as a tool to solve problems (10). Companies invest in artificial intelligence software to be leaders in their industry (11).
Our main business is biscuit and chocolate, with artificial intelligence applications, we can improve our business firstly by making pilot applications, and then transforming or production, business and management processes to a faster, cheaper and more efficient work culture that makes our customers even happier. The aim here is not to remove people out of the process, but to create companies of a more efficient future with human-machine cooperation. All jobs and tasks should be evaluated and the artificial intelligence software-human relations should be studied one by one, considering the conscientious dimension. Our media planning agencies have been using advertising technologies for five years and have been showing ads in the right place, at the right time, with the right budget, at the right frequency. So, these technologies are already in our lives, whether we want them or not. It is time to bring these technologies into the work in a desired and targeted manner.
We have businesses that collect digital data directly from customers. Online + offline transformation and coexistence is accelerating. Areas with high contact with people, such as large cafes and restaurant chains, are trying to adapt to changing consumer habits with the investments they make in artificial intelligence applications. Likewise, in our retail business, we need to demonstrate how we can improve our way of doing business with pilot studies to stay ahead of the competition.
For example, let’s imagine that we produce and serve in geography with a population of 4 billion. This means we can make 1.5 billion households happy with our products. What if we teach the consumption data of these households to the machines, the machine (algorithm) tells us how much of which product they can consume according to the weather temperature and mood, and we send these products to the houses weekly by drones. We can also add the products we have just released into the system, the algorithm can tells us which house will consume the new product (if there is already data, the new product is produced by micro-segmentation); we add it, and we get the payment automatically. Also we can take back those that are not consumed, this will make a programmatic demonstration to maintain the image of our brands and to promote new products. Wouldn’t it be nice?
Likewise, if the artificial intelligence-based software tells us what kind of coffee and menu they desire, at what time, in what mood, at what temperature, in the places where Godiva Cafes serve; and if we optimized our cafes and stores accordingly, prepared and showcased those coffees and the food they want with those before they arrived, and bought for them before entering the store, while they were passing by, and made automatic payments – wouldn’t it be great?
While I was writing that “So we have to dream first, then we have to see how we can do it”, the introduction of a company’s artificial intelligence sales management assistant named “Sellina” came to my mail. It was sent by sales tech expert Shubham Gupta. Here’s another new profession, I don’t know if they found my address with LinkedIn artificial intelligence, but it’s obvious that Shubham Gupta is still a real intelligence like us. Anyway, the mail is like this:
Good afternoon Murat,
If you had a conversation with an AI assistant when you walked into your office today:
Murat: Sellina, could you tell me where to focus on achieving our sales targets this month?
Sellina: Hello Murat, this is what you will do this month to meet your sales targets.
Murat: Thanks Sellina, can you tell us the best salespeople in my team?
Sellina: I’ll find in a second. Here are those who perform well.
Murat: Why Mark’s performance was not good?
Sellina: That’s what good performers do, and that’s what Mark does.
Sellina was able to provide surveillance, coaching, assistance to more than 1000 salespeople at the same time, increasing daily sales by 13 to 18%! A good start that sounds like the voice control of the voice response system and its working like an assistant, here is something that excites me … But the sales analysis programs we currently use are doing the same, although not full-time and interactive like Sellina. I guess my expectations from artificial intelligence are much higher…
What does what I have written so far tell you? It seems to me that employees who have “decision-making under uncertainty, more cooperation, more experimentation, quickly solving common problems in different ways and creativity” features in business environments will gain importance (12). To differentiate from machines, it is necessary to take open communication and initiative. “How can we do better?” Employees who think, question, and do are the biggest dream of the bosses. Fear of Speaking-Up, unwillingness, reluctance to point out problems, is the biggest threat to business environments. Because, if repetitive tasks are to be left to artificial intelligence, what differentiates the employees will be the ability to see problems, say fearlessly, and create/produce solutions.
Micro Innovator is a new word. It means always trying to do the job better. In other words, those who constantly think about how to do my job in a way that is more effective, better, faster, easier, cheaper, and more satisfying to the customer are micro innovators (13).
This is where artificial intelligence comes into play. Now, every employee needs to think about how I can do his / her job better, easier, cheaper, faster and better with the possibilities of artificial intelligence and how can I make the consumer happy. Currently few people know the detail of how a machine learns. You really need to be a “code writer” for this. But for us humans to benefit from artificial intelligence, we need to know what it does and how to use it to improve our business. Not just for using artificial intelligence software, but to really get help! Artificial intelligence requires history data, but not as much as before. In other words, those who have processed trillions of data such as Facebook, Google and YouTube until today seem to have benefited from the advantage of artificial intelligence, but now monopolies in this area will be destroyed.
For this reason, it is not artificial intelligence that will create today’s advantage, but the courage and agility of our employees. Some say that “If your employees do not use their potentials such as creativity, empathy and problem-solving techniques, then in the future, submit all your work to machines with artificial intelligence.” If our job was repetitive, routine, or predictable, we would be in a competition to automate our work urgently. There is certainly an automation revolution. What computers cannot do with machines: is to decide how to manage the company and the machines.
This is exactly why our Human Resources Department established the Yildiz Analytical Academy. Many projects have been created by our Digital Disruption groups. To make projects such as Route Optimization, Vehicle Load Optimization, Raw Material / Product Optimization, Sales Campaign Optimization, Store Segmentation, CRM Analytics, Product Demand Forecasting, Consumer Behavior Forecasting, Return Optimization, Order Creation, HR Analytics, Demand Sensing and Unique Customer we need employees who have knowledge about statistic, data analysis and machine learning.
Let me say that there should be personnel specialized in artificial intelligence, business analytics, and big data analysis, but now every employee should also have an awareness of what these technologies do, what analyzes they do and how they achieve results to benefit from new technologies. Therefore, I congratulate our team who established and are running Yildiz Analytical Academy. I think our point of view should not be just starting from the project. For example, we should put the “input” part of artificial intelligence in front of us in the attached table, diversify it and imagine which tasks can be done with artificial intelligence technologies, and then turn them into projects (7, 8).
Finally, let me touch on one more subject. It is claimed that even those who wrote these algorithms (an algorithm is a code written for a target) do not know why a specific result came out after writing it and this is called a “black box” (14). It is stated that this will bring serious moral and legal problems.
The lack of transparency of machine learning is partly due to the way the algorithm is trained, it says. In “Deep Learning”, which is a sub-method of machine learning, there is a large amount of data at one end of the representative neural network. For example, they were introducing millions of dog photos to the machine. As the data travels between the information layers of the neural network; each layer was gradually picking out more abstract features to produce the correct result for the final output layer, like separating a Chihuahua from a Miniature Pinscher. However, since this process takes place within a neural network, the researchers could not explain each abstract feature or why the network decided to extract a particular set of features. Unless this was disclosed, the job was getting harder. Since Artificial Intelligence coding imitates human cognition, it was normal to include intuition, but while the human could explain his own intuition, the machine, the software developer, could not explain it. It was therefore possible for the machine always make choices in favor of sex, a race, a nation. I think this is a big lie! Your dog carries your prejudices. If you sulk at the guest, he will bite him “crunch!” I think everyone knows well what they are doing, but they already have an excuse for the immoralities and illegal practices that will be “we don’t know, this is a black box”.
Normally researchers do not find algorithms that make errors dangerous, but the problem is the people who hide the errors, not systems that hide their errors (1, 15).
As a result, artificial intelligence is now such a large sector that there are so many sides that it requires software awareness to understand which article and which news are made for “sale”. Culture and politics also affect artificial intelligence. The myth that Cambridge Analytica (CA), which uses algorithms to target voter groups who may be affected by the negative aspects of artificial intelligence software produced micro-targeted content in the Leave EU campaign with the presidential elections of the USA and harmed democracy is still a legend. Campaigning with micro targets is legal if the data is authorized. There is no research showing that personalized ads produced by CA affect the selection outcome behaviorally. Another example is the largest Artificial Intelligence Company in China; Real-time detection and identification of moving objects from surveillance videos is a “privacy” problem for the West, whereas the Chinese people do not care at all but find it useful (16).
Despite the GPT3 artificial intelligence revolution and developments in this area, I cannot understand some people who think: “Even the smartest artificial intelligence system is problematic right now. For example, partial blocking or noise deceives facial recognition systems, self-driving vehicles can cause accidents because they do not recognize stimuli, translation systems can not identify unusual accents. It is predicted that even the highest-level neural networks are not capable of dealing with a changing situation flexibly as even a young child could. A three-year-old child can recognize a dog, use simple sentences, and figure out how to use a tablet. Ask any robot to do this, if it has not received any specific training in these tasks, you will see it fail. Artificial Intelligence can only be trained for one job. There is no such thing as General Artificial Intelligence.
Yes, machines and robots are quite unsuccessful in intuition, decision-making, bringing out something new by blending new information, because even though the machines are partially “conscious” with “coding”, the “subconscious” and “short-way” solution that feeds on emotion, that is, the acquisition of practical intelligence it seems unlikely. But even the current benefits of artificial intelligence can provide great advantages in the business environment, customer satisfaction and cost reduction. Two years ago, artificial intelligence applications did not exist that much in our lives. Also, if algorithms that can draw flexible conclusions from learning activities (the example of changing the street name of the municipality) can be written, the nature of artificial intelligence may change. Artificial intelligence is ultimately a tool designed to serve the interests of human coding. This should never be forgotten (17, 18). I closely follow business technologies and applications for analytics (19). It is said that the speed of supercomputers in the world has stagnated in recent years, which slows down artificial intelligence. But what is currently being discussed is that quantum computers are coming; Quantum computing is different from computer coding based on 0/1 we know. Quantum bits can be both 0 and 1 at the same time, according to probability, so it seems that the fuzzy logic that I mentioned at the beginning can be processed more with quantum computers. On the other hand, it is said that artificial intelligence applications will be more efficient if “cloud computing” is used (20, 21). There is also the issue that big data leads us to science without theory, which is another story (22).
I was going to tell you why my name was mentioned in the video about GPT3:
As they said, currently, the cost of the latest version of this artificial intelligence software can only be covered by one or two people in Turkey and based on the video, I am supposed to be one of them. Let me warn you: if they are going to invent something by counting on me, tell them not to. I am thankfully happy with my own intelligence.
Note: This article, which is open source, can be cited by mentioning the author. Does not require copyright.
1. Fan, Shelly. Yapay Zeka Yerimizi Alacak Mı? (Will AI Replace Us: A Primer for the 21st Century (The Big Idea Series). Translated by Ipek Gunes Cigay, Hep Kitap, 2020, p. 214-229.
2. Chojecki. P, Artificial Intelligence Business, Amazon Publication, 2020.
3. Harvard Business Review, Yapay Zeka (On AI, Analytics and the New Machine Age), /HBR’S 10 Must Read Series, Translated by: Levent Goktem, Dijital Donusum, Optimist, 2020.
4. Harvard Business Review, Translated by: Nadir Ozata, Yapay Zeka (Artifical Intelligence), Harvard Business Review Press, Optimist, 2019.
5. Brynjolfsson E. ve McAfee A., Yapay Zekanin Vaat Ettikleri, icinde Dijital Donusum, Yapay Zeka (What Artificial Intelligence Promises, in Digital Transformation, Artificial Intelligence), Optimist, 2017, pgs.19-56.
6. Arf, C., (1959) ‘Makine Dusunebilir Mi ve Nasıl Dusunebilir?’, Ataturk Universitesi , Universite Calismalarını Muhite Yayma ve Halk Egitimi Yayınlari Konferanslar Serisi No: 1 (Can Machine Think and How Can It Be? “, Ataturk University, Spreading University Studies to Neighborhood and Public Education Publications Conference Series), 1958-1959 Educational Year, Erzurum, pgs. 91-103.
7. Alpaydın, E. (2016). Machine Learning, MIT Press; (2018), Yapay Ogrenme (Artificial Learning), Bosphorus University Publications, 4th Ed.
8. Davenport H. T and Ronanki R., Gercek Dunya Icin Yapay Zeka, in: Yapay Zeka (Artificial Intelligence for the Real World, in: Artificial Intelligence, Optimist, 2019, pgs.7-29.
9. Wilson, J. H. (2020), The Next Big Breakthrough in AI Will Be Around Language, HBR 25 September; Artificial intelligence: fear or rejoice because of the Next Generation GPT-3, BBC News, Turkish, 10 August 2020.
10. Fountaine T. and McCarty B. (2019). Artificial Intelligence Powered Company, HBR, August 2019.
11. Lansiti M. and Lakhani R. Karim (2020). Competition in the Age of Artificial Intelligence, HBR, February 2020.
12. Wilson, J.H and Daugherty P.R (2019). İşbirliğine Dayalı Zeka: İnsanlar ile Yapay Zeka Güçlerini Nasıl Birleştiriyor? in: Collaborative Intelligence: How Does Human and Artificial Intelligence Unite Powers? in: Artificial Intelligence), Optimist, p.177-198.
13. Hurt, K. and Dye, M.D., Courageous Cultures, Harper Collins, 2020, pg. 224.
14. Davenport T.H., İş Dünyasında Yapay Zekanın Durumu, in: Dijital Dönüşüm, Yapay Zeka (The State of Artificial Intelligence in the Business World, In: Digital Transformation, Artificial Intelligence), Optimist, 2020, pgs.7-18.
15. Say, Cem, Yapay Zeka, Bilim ve Gelecek Kitaplığı (The Book of: Artificial Intelligence, Science and the Future), 3.Baskı, 2018, pgs. 93-105.
16. Lee, K., Yapay Zeka ve Yeni Dünya Düzeni (A.I. and the New World Order), Optimist, 2018, pg. 278.
17. Agrawal, A., Gans, J. and Goldfarb A., How to Win with Machine Learning, HBR, Sept-Oct. 2020.
18. Koç, Onur, Daha iyi Bir Dünya İçin Yapay Zeka (A.I. for a Better World), Doğan Kitap, 2018, pgs.51-58.
19. Thomas, R., The AI Ladder, O’Reilly, 2019, Pg. 16.
20. Smith, Brad, Tools and Weapons, Penguin Press, 2019, pg. 346.
21. Ghose, S., Are you ready for the Quantum Computing Revolution, Harvard Business Review, 26, September, 2020.
22. Leonelli, Sabina,“Scientific Research and Big Data.” Stanford Encyclopedia of Philosophy, Stanford University, 29 May 2020, plato.stanford.edu/entries/science-big-data/.