Nanite in Unreal Engine 5
The Unreal Engine 5 sets new standards in terms of photorealistic graphics. New rendering technologies like Nanite play a decisive role in this. But…
Everything about ai
Technology
Objective
The term artificial intelligence is omnipresent today. But what exactly is AI? How does it work? These questions will be explored in this article.
Definition
John McCarthy, one of the pioneers in the field of artificial intelligence, defined the term in 1955 as follows: "The goal of AI is to develop machines that behave as if they have human intelligence." These intelligent machines, in our current understanding, are usually computer systems that behave intelligently. Intelligence refers to the ability of AI to solve tasks that require intelligence, recognize and understand human language, recognize images, and much more. However, most AIs can typically only solve a trained task at a high level, where they already surpass humans by far. All current AI systems are referred to as weak AIs. An artificial intelligence that can solve multiple tasks at a high level is called general AI by experts. This form of AI does not yet exist, but it is the ultimate goal of research in this field.
Creation of an AI
When creating artificial intelligence, there are various approaches and methods. The current favorite in AI research is called "Machine Learning." In this approach, software autonomously writes and optimizes its own code. This code powers many modern AI services, such as Netflix's recommendation algorithm.
Another approach involves neural networks, which date back to 1943. In a neural network, an algorithm creates multiple layers connected through neurons or nodes. Information is exchanged through these connections. Generally, the architecture of the networks includes at least one input layer, one hidden layer, and one output layer. In summary, neural networks are algorithms that can self-optimize. When neural networks are combined with machine learning, as described above, it is referred to as deep learning. This methodology is responsible for the massive AI boom in recent years and is primarily used in image recognition and autonomous driving. Next, we come to "Generative Adversarial Networks," or GANs. This term refers to neural networks (also called agents) that control and improve each other. Both are trained with the same dataset. One of the agents starts by generating content similar to the dataset, while the other compares it to the dataset. If the second agent detects a fake, the fake agent is forced to improve. These methods are used, for example, in the creation of deepfakes. All of the described approaches are united by the so-called "Blackbox Problem." How an artificial intelligence reaches its result is often not understandable. This is primarily due to the highly complex and multi-layered neural networks, which, like a black box, lie between the input and output.
AI-Training
Training is an essential part of modern AI technology. However, the way in which learning and training occur can vary greatly from one AI to another. Below are some well-known learning methods.
First, there's "Supervised Learning". In this method, the training data for the AI is prepared beforehand by humans. For example, certain objects in photos are labeled, and the AI learns what patterns it should look for in the data. This type of training is primarily used in autonomous driving and facial recognition.
Another learning method is "Unsupervised Learning". In this case, the AI is provided with a large dataset that does not contain any labels. Unlike supervised learning, the AI must independently search for patterns and relationships within the data.
Next, let's look at "Reinforcement Learning", also known as the "trial and error method." In this approach, the AI is "rewarded" when it successfully completes a task. If it misses the target, it either doesn't receive a reward or is "punished."
"Transfer Learning" involves applying the skills that an AI has learned to a related problem. For example, image recognition AIs are used by medical professionals for targeted cancer diagnosis. This training method bridges the gap from specialized abilities to flexibility and general AI.
Now, let’s consider "Imitation Learning". Here, a human demonstration is used as training material. An example could be a recording of a human player playing a video game.
Lastly, there is "Few-Shot Learning". In this training method, an AI learns a new skill based on a few examples (photos, videos, etc.).
Areas of application
The application of AI technology can be divided into four major areas: fundamental research, industry, end consumers, and art. Fundamental research mainly focuses on robot-controlled processes, machine learning, and the development of digital assistants. The use of AI in industry is primarily in supply chain management, maintenance, and sales and marketing. Online retailers like Amazon and search engines like Google use AI, for example, to offer their customers suitable products. End consumers use digital assistants and translators powered by artificial intelligence. AI is also making its way into the arts, particularly in image generation and modification.
Conclusion
AI technology is now an integral part of our digitalized world. Tech companies like Google and Facebook use artificial intelligence to tailor search results and news feeds to users. End consumers use AI daily on their smartphones when they interact with one of the countless digital assistants. However, technological progress also has its dark sides. Totalitarian states around the world use facial recognition AIs for total surveillance and crime prevention. Additionally, deepfakes are increasingly being used by fraudsters and spies.
Questions & Wishes
We hope you like our article and would like to invite you to share your thoughts and questions on the topic with us. If you have any questions, comments or feedback on the content of this article, please don't hesitate to let us know in the comments section. We're always happy to hear from our readers and engage in meaningful discussions about game development.
Just ask us anything you want to know and we'll do our best to provide the answers you're looking for. Thank you for your support and we look forward to hearing from you!
6 tips for good storytelling
This article will list and explain some tips for good storytelling.
Unity DOTS: Our top 5 features
In this article we share our experiences with Unity's Data Oriented Technology Stack (DOTS) while working on our DOTS project “Recursive Reckoning”.…
Write comment