- AI can solve all problems.
Currently, each type of AI has been developed for use in clear objectives. For example, Face Recognition AI is created to identify identities by face. Will not be able to read the fingerprint, etc. AI learns from data sets When the data set changes, the mechanics of AI may also change, such as using NLP to translate languages, which each language has a different grammar structure. Makes it impossible to use the same AI. Therefore, AI cannot solve all problems As for the question that What kind of problems does AI solve? The simplest answer is a problem that has no information before. And don’t know what information to use as a model Including problems that have no clear cause and effect May not be able to use AI to solve problems such as the spread of emerging diseases Can not find an answer with AI how to treat, etc. So the conclusion is that AI can not solve all problems. And not every problem has to be solved by AI
- AI is a ready-to-use technology.
AI technology is available in 2 forms, which are ready-made AI and newly developed AI. The difference is that ready-made AI is learned from one data set. That has a large enough amount, while AI that needs to be developed Will learn from the data set of one of the problems. The word “ready” here means that there is no need to train again. Algorithm currently used has a Library (meaning an algorithm called Algorithm) to choose from. Use a variety of Train Model is to call Algorithm to learn with information. In the case of using ready-made AI, the program will select Algorithm and learn from the data set that has already been selected. Causing the user to not be able to modify or open the code to understand the mechanism of the system
Choosing whether to use ready-made AI or not Depends on the need And the characteristics of ready-made AI sample data such as Image Processing that can analyze cars on the road Makes it possible to know What type of car is each car, what brand and how fast? In this case, the data set required is the car and the speed of the car on the road. Therefore, data sets are not specific data sets. But should be a large set of data, large enough for the Train Model. Therefore, technology companies that can capture images And the movement of vehicles on the road Therefore, this AI can be made ready-made and can be used for general road use as well.
On the other hand, in the case of creating a specific AI Arising from the data set Or certain non-general factors But is a specific information such as forecasting model for different types of clothing stores, which must understand the cycle of each type of clothing Before creating a model for forecasting sales, such as certain clothing types, can be sold throughout Some types are sold during festivals. Some countries are fashion products that sell only one lot, etc. In addition, other factors such as promotions, sales channels. Different types of stores and locations are all factors that influence the development of the model. This model of sales prediction is suitable for creating a specific model. And further developed into an intelligent ordering system Therefore, it can be said that AI is created especially for this problem.
- AI learn by yourself Even with new factors involved
At present, most AI is in the type of Limited Memory and learning by using prototype data to create Train Model (Model) and use that Model whenever the results do not answer. Or to the new Train Model. AI will learn according to the same mechanism that was designed. By using specified factors Or is a factor in memory only. Therefore, if behavior changes Or if new information is generated, AI will not be able to learn by itself, or even AI in the future will need some basic mechanisms. In order to determine the learning objectives of AI. And this is where we come to say, “Anyway, humans are smarter than AI because humans are the creators of AI.”
The learning of AI and the purpose of learning is not the same. Humans define learning frameworks, such as inputting factors, selecting algorithms, determining target variables, and so on. Train model is the operation by computer system. Model is created according to the objectives set. Therefore, when new factors occur May result in objective change Resulting in the need to adjust the learning of AI
- Just being processed is AI.
Processing is part of AI, but not all. But some types of processing are not AI, because by definition, AI must be able to learn Technically, most programs will learn logic. For example, if a number equals 1, then add that number to +2, but if that number equals 2, then add that number to +3 and so on. Programming like this is called Rule-Based If-else writing, which is the basic principle of programming. But learning logic like this is not considered to be AI but is considered as Part of AI development. For example, Chatbot type Rule-Based Chatbot is not AI because it must type the correct information as required. And there is no learning mechanism, for example. However, said that Rule-Based is not AI but is part of AI development because AI does not have only the processing. But there are still actions Which needs to be programmed To show the results
- AI developers must finish PhD only.
Is a very misunderstanding Because the development of AI has many technical parts, every person who studies computer science Can be an AI developer because it is a science that is especially studied for computer systems In which basic subjects such as Software Architecture are subjects studied at the undergraduate level Therefore, it is not necessary to finish a Ph.D. Can develop AI
However, this belief arose because in the past The key people involved in the history of AI development, most of them are university professors. And graduated with a doctorate degree That is so Because in the past, AI science was only known in the research industry, but now the development of AI can be carried out easier. Even ordinary people can develop simple AI such as machine learning model etc.
In addition, the misconception about AI development is that it is only necessary to learn about Modeling. In fact, each type of AI has different elements, such as NLP, which focuses on defining words for AI to learn. Image Processing focuses on Place image In order for the computer to understand the structure of the image, etc. Therefore, in the group of AI developers, there will be personnel with diverse expertise combined to develop into intelligent mechanisms.
- AI was born to compete for work
Go back to the time of the 1st Industrial Revolution during the steam engine era. Is the origin of the train The transportation system has changed. The question is whether the arrival of trains makes life easier or not. The answer is yes, because if today we don’t have trains. We still don’t have cars. No plane We would have to travel on foot. Comparable to other technologies, including AI that has been developed from human challenges. In response to the lives of humans, on the other hand, may also affect certain groups of people as well. But with the main objective, AI was not created to compete for human work. But was created to be a human tool
Will there be unemployed because of true AI? Or must answer that true in some parts But not all Technological development can have a negative impact on human behavior, such as the coming of YouTube, resulting in fewer live TV viewers. Therefore, the use of AI technology also affects work.
But on the other hand, the advent of AI also leads to new careers as well, such as the career of AI developers, intelligent marketers who can analyze customer behavior with AI. Genius for customers Which in the past, the marketing team Will have the same promotions over and over again like Mother’s Day promotion Birthday promotion Year-end promotions, for example, allowing customers to start attaching images to those promotions But when the system is used to classify customers with the Clustering Model, a type of Machine Learning Model, traditional promotions are no longer effective.
Therefore, the marketing team must adapt. And think of new tricks To present to customers in each group More than that When having customer analysis like Personalization Causing everyone to see different promotions It turns out that marketers have to work harder. In order to look for things that are more responsive to customers
In this case, taking advantage of AI does not reduce the amount of the marketing team. Instead, adding work to the marketing team to do more detailed work with AI as a tool. At the same time If the marketing team does not learn, does not understand how the Clustering Model works, does not learn to create data views. That marketer will no longer be the person who is more suitable for marketing work using AI.
A misconception about AI results in the failure of the AI development program because it is unable to fully utilize the AI. According to Gartner, 2018 – 2019 there are a number of projects that use AI failures. Up to 85%. Therefore, if you wish to use AI, you must first understand the operation of AI because the investment in AI must consider the ready ecosystem, including related personnel. with