Crazy New World of AI and ML
Even though the authors of this article want to explain artificial intelligence with common sense, it is essential to define the basic concepts we use in book creation. Before delving into the main topic of this cool book, it is crucial to familiarize ourselves with the basic terms currently appearing on the market, where the term artificial intelligence has become as ubiquitous as, for example, the phrase mobile telephony.
And even though Roland Haenggi has horses on his property and I am working in the field of digital transformation of production facilities, there's a chance that when it comes to an understanding artificial intelligence, I may be more of a farmer, and Roland more of an expert with many years of experience. This definition of terms served as our alignment and foundation for writing the article for the upcoming book and our collaboration. Because even a different understanding of the same issues among various experts in the market can lead to additional confusion, now imagine the chaos that would arise if the authors used other terms for the same things in this book.
TIP: To understand and engage in discussions about Artificial Intelligence, it's crucial first to familiarize yourself with the basic terminology in this field. Misunderstanding or misusing these terms can lead to confusion, even among experts. Therefore, please always ensure you understand and align on terms before delving into deeper conversations or collaborations around AI.
What is AI?
Artificial intelligence is the simulation of human intelligence in various devices designed to think, learn, and solve problems like humans. Such a solution encompasses a wide range of technologies, many different algorithms, and techniques that allow computers to perform complex tasks that typically require human interaction, experience, or even knowledge about a particular problem, such as recognizing patterns, faces, or languages. In addition to all the above, artificial intelligence can make independent decisions.
We can thus classify the artificial intelligence system into two main categories:
First category: Narrow AI
They are designed to perform specific tasks without the ability to expand to a broader scope. These solutions excel at solving particular tasks, usually performing them faster than a human can, and are significantly more accurate. However, they cannot learn and adapt in areas that exceed their programmed functions.
These are examples that can be found on the market, and you likely already use some of these solutions daily and have experienced the capabilities of artificial intelligence.
Second Category: General AI
General Artificial Intelligence, or Artificial General Intelligence (AGI), refers to a form of artificial intelligence capable of understanding, learning, and applying its intelligence to numerous tasks similar to a human being. Unlike Narrow AI, Generic AI possesses cognitive capabilities that allow solutions to adapt to new situations, gain knowledge across various domains, and transfer learning from one context to another.
However, general artificial intelligence remains a theoretical concept, as no known AI solution has reached the level of human intelligence. Therefore, most solutions today fall under the Narrow AI category, which is highly specialized and limited to specific functions.
While AGI has yet to be achieved, companies are actively engaged in this field, such as OpenAI, DeepMind, and others. They strive to develop AI with more general capabilities. However, the timeline for development remains to be determined, and it may take several years or even decades to achieve this level of artificial intelligence.
TIP: When learning about Artificial Intelligence, it's crucial to understand the two main categories: Narrow AI and General AI. Narrow AI systems excel at specific tasks, such as voice or image recognition, and are commonly found in everyday applications. On the other hand, General AI, capable of understanding and learning across various tasks similar to a human being, remains a theoretical concept yet to be fully achieved. Understanding these categories will help you comprehend AI technologies' current capabilities and future potential.
What is the difference between AI and ML?
Then there is a significant distinction to understanding solutions that we all generally call artificial intelligence, even though, in some cases, it is not artificial intelligence but machine learning. Therefore, there is a critical difference between AI and ML solutions, which a layman's eye does not need to understand. On the other hand, both authors agree that if you read this text, you are not a layman but an expert in your field who would like to gain more information about new innovative solutions.
When we use AI (artificial intelligence), we talk about the latest components provided by vendors such as AWS, Microsoft, Google, etc., for image analysis, text-to-speech, speech-to-text, or ChatGPT. Therefore, artificial intelligence solutions are ready-to-use solutions that users can use relatively quickly. So, for ChatGPT, you don't need any integration, development, or training to start using the potential of this solution. If we now refer to production, these solutions allow almost immediate use. Some examples are fault detection on specific machines via sound analysis, face recognition for factory entry, safety at work solutions that warn employees to wear safety equipment through cameras, and the like. These are closed solutions that are difficult to use for different tasks than they were initially developed for. In this case, we are dealing with Narrow AI.
When it comes to ML, the understanding is more complex. As the very name suggests, we are dealing with machine learning, so these solutions and their results are based on learning. When we talk about ML, we focus on creating a machine-learning model using tools and platforms like Anaconda and Tensorflow, among many others. Machine Learning (ML) is a field of Artificial Intelligence (AI) that relies on algorithms and statistical models to enable computer systems to learn and improve their performance autonomously without being specifically programmed for this.
领英推荐
Definition of AI versus ML: Artificial Intelligence (AI) encompasses a broad range of algorithms and methodologies that enable machines to perform tasks that typically require human intelligence. In contrast, Machine Learning (ML) is a specialized subset of AI that involves the development of algorithms that can learn from and make predictions or decisions based on training and validation data. This enables ML models to excel in performing specific tasks by adapting to patterns and structures within the data.
TIP: To understand innovative solutions, it's essential to distinguish between AI and ML. AI, or Artificial Intelligence, often refers to ready-to-use solutions provided by vendors like AWS, Microsoft, or Google. They are utilized for tasks such as image analysis, text-to-speech, and more. On the other hand, Machine Learning (ML), a field within AI, involves creating models using tools like Anaconda and Tensorflow. These models enable systems to learn and improve performance autonomously. Recognizing the difference between AI and ML can help you select the appropriate solution for your needs.
A machine learning model operates in a few basic steps.
Data Collection: The complete dataset is typically partitioned into two subsets: the training and the validation or test datasets, often in a 70/30 ratio. The training dataset is utilized to train the model effectively, allowing it to learn and adjust its parameters for optimal performance.
Data Processing: The data must be processed and prepared for the machine learning model. This can involve various steps, such as cleaning data, removing unnecessary information, encoding categorical data, etc. Model Building: We then select the machine learning algorithm we want to use (e.g., linear regression, logistic regression, neural networks, etc.) and build the model.
Model Training: The model then "learns" based on our training dataset. This is the step where the model recognizes patterns in the data and adjusts based on these patterns. Model Testing: The model is tested on the validation data not used in training to check its accuracy. This data is known as the testing dataset. Model Application: If the model is accurate, we can apply it to new data to predict or classify outcomes. It is imperative to note that the partitioning of the dataset into training and validation/test subsets should be conducted through an automated, unbiased, and randomized method rather than manual human selection to ensure the integrity and representativeness of the validation/test dataset.
Machine learning is often used to solve various problems, from handwriting and speech recognition to recommendation systems, health diagnosis, stock trading, and self-driving vehicles. In manufacturing, machine learning is used for defect recognition on the product, production planning based on market changes, business decision recommendations, and the like.
The main difference is that artificial intelligence (AI) is like a big box of Lego blocks a company has built. We can use these blocks in various ways, but how we put them together is unique. On the other hand, machine learning (ML) is like a growing plant that develops on its own and specializes in solving a specific problem uniquely.
TIP: Machine learning model operates through several crucial steps: Data Collection, Data Processing, Model Building, Model Training, Model Testing, and Model Application. Choosing appropriate datasets for training and testing is essential to ensure the model's accuracy. Machine learning, used for various tasks such as defect recognition or production planning, can be likened to a growing plant that develops on its own. In contrast, Artificial Intelligence (AI) resembles a set of Lego blocks, ready to be assembled in various unique ways. Understanding this distinction can guide you to choose the right technology based on your specific needs.
Understanding whether a solution incorporates AI or ML or is just a sales gimmick?
Indeed, it is difficult to answer this question, even if you are an AI and ML solution development expert. Nevertheless, it's essential to address this aspect for further understanding, as this knowledge can greatly benefit you in reaching success in digital transformation. Distinguishing whether a company uses AI or ML can pose a significant challenge, as AI solutions and ML solutions are based on complex mathematical algorithms. However, there are subtle signs that could indicate that a company indeed uses AI or ML.
A key characteristic of AI and ML solutions is that they can learn and improve over time based on data and real-world events. Therefore, these solutions are highly adaptable. Even if it's an AI solution, it can learn from data and react differently the next time. Such systems gradually fulfill their predictions, decisions, and actions based on feedback. For solutions that only use advanced algorithmic functions and basic mathematical equations, this is hard to assert, as they are more rigid, requiring a development engineer to update a part of the code if a solution update is made.
A company will likely utilize AI or ML if its solution can handle large, unstructured, and complex datasets, providing results from these diverse entries. Solutions that require structured data similar to traditional systems could be a warning sign. Decision-making and prediction abilities, especially those predicting future events or trends, also suggest AI or ML usage. The solution's autonomy, particularly in constantly changing environments like self-driving vehicles or planes, further points to AI or ML. Finally, pattern recognition or anomaly detection, mainly when conventional algorithms overlook deviations, indicates the use of AI or ML.
But after all, it is still difficult to answer whether a company uses AI or ML or if it's just a sales gimmick because until you start using the solution yourself or without insight into what technology the company uses for the development of such a solution, or without the assessment of a technical expert, it will be almost impossible. Moreover, knowing that AI and ML help solve specific problems is essential. Ultimately, if such a solution successfully solves your problems and you find out later that it isn't based on AI or ML, it must still bring you some value.
TIP: When evaluating if a company uses AI or ML, look for key signs such as adaptability, data handling capacity, predictive capabilities, autonomy, and pattern recognition abilities. AI and ML solutions learn and improve over time, handle complex datasets, make predictions, operate autonomously, and detect patterns or anomalies. Be skeptical of solutions requiring structured data or not exhibiting these characteristics. However, remember that it can be challenging to make this determination without technical expertise or firsthand use of the solution. Ultimately, a solution's value in solving your problems is most important, irrespective of whether it is AI or ML-based. Frequently, Artificial Intelligence (AI) or Machine Learning (ML) can serve as powerful solutions or game-changers when conventional analytics-based approaches are inefficient. In such instances, AI or ML may provide the requisite edge. Consequently, it is advisable not to actively seek scenarios for applying AI or ML but to consider employing these technologies when traditional methods are ineffectual.
Join the AI Adventure
This article is just a sneak peek into an upcoming book Roland, and I are passionately working on. Our mission is to demystify the wonders of AI and bring them right to your doorstep. We're committed to publishing at least one article monthly to keep your curiosity well-fed and your intellect ever-growing.
But wait, there's more! If you're as excited about this as we are, you can stay in the loop by subscribing to this newsletter. Not only will you get early access to our articles, but you’ll also become a part of our close-knit community of AI enthusiasts!
Moreover, this journey is as much yours as it is ours. We value your insights and are eager to learn from you. If you have any feedback, questions, or thoughts about AI that you'd like us to explore, please don't hesitate to reach out. Your input could be the spark that shapes the next chapter of our book.