GLOM: The Revolutionary AI Discovery That Will Change the Future

Contents:

• How to teach a computer intuitive perception
• How GLOM will change our perception of the world
• How the GLOM system works
• GLOM prospects: a breakthrough in artificial intelligence

How to Train a Computer to Be Intuitive

Geoffrey Hinton's article on GLOM theory begins with a key clarification: "This concept does not describe a functioning system, but rather a hypothetical model." GLOM is a term that refers to the unification of various elements into a single structure, thus emphasizing its uniqueness. This model can serve as a basis for further research in neuroscience and artificial intelligence, offering new perspectives for understanding cognitive processes.

Hinton aims to develop a system capable of modeling Human perception in the field of artificial intelligence. The GLOM project proposes that artificial intelligence will be able to process visual information and interpret it differently. The scientist emphasizes that the human brain contains significant vectors of neural activity that play a key role in this process. GLOM combines similar vectors, which are arrays of numbers encoding information, a fundamental element of neural networks. This innovative model could open new horizons in the development of AI, improving its ability to perceive and analyze visual data. Hinton argues that the key difference between human and machine thinking is intuition. Intuition is the ability to identify analogies and parallels between different objects. This ability is formed from an early age and develops throughout life, making human thinking more flexible and adaptive compared to the algorithmic processes of machines. Intuitive perception allows people to make quick decisions based on accumulated experience and knowledge, while machines act on the basis of predetermined rules and data.

The GLOM model demonstrates the ability to reproduce artificial intuition, which is a key aspect for adequately perceiving the surrounding world. This characteristic allows artificial intelligence to understand reality in the same way that humans perceive it. To date, no existing theory has achieved such significant results, which highlights the uniqueness and innovation of the GLOM approach in the field of artificial intelligence.

Modern theories of artificial intelligence are often based on the idea that the brain processes information in the form of images or symbols. However, the GLOM approach represents an alternative view, arguing that the brain functions using extensive vectors of neural activity. This changes our understanding of perception and information processing, offering a more complex and integrated model of brain function. This approach can lead to new discoveries in the fields of artificial intelligence and neuroscience, opening up new horizons for the development of more efficient data processing systems.

GLOM theory provides solutions to two main problems facing artificial intelligence. These challenges relate to understanding and processing information, as well as AI's ability to learn and adapt. GLOM focuses on improving these aspects, resulting in more efficient and accurate models. Implementing this theory contributes to the development of AI, making it more versatile and capable of autonomous learning in complex environments.

• Understanding the world through the lens of objects and their natural parts;
• Recognizing objects from different viewing angles.

How GLOM Will Change Our Perception of the World

Geoffrey Hinton, one of the leading experts in the field of artificial intelligence, argues that GLOM technology has the potential to significantly change the way AI perceives objects. He focuses on two important aspects: the part-whole relationship and the viewing angle. Hinton emphasizes that if GLOM is successful, its ability to perceive objects will be significantly closer to human perception than current neural networks. This technology could lead to new opportunities in the development of intelligent systems that are more adaptive and effective in solving complex problems.

In developing GLOM, Hinton integrated the best practices of the first two generations of computer vision systems. The first systems relied on part-to-whole object recognition, while the second generation applied deep learning, leveraging large amounts of data to improve recognition quality. GLOM represents a harmonious combination of these approaches, enabling greater accuracy and efficiency in computer vision tasks.

In developing GLOM, Hinton aimed to reproduce the mental shortcuts that people use to perceive the world around them. For example, given data about eyes, artificial intelligence can infer that it is looking at a face. GLOM uses a part-analysis strategy to allow AI to identify the entire image: by recognizing a specific nose, the system can associate it with a specific face. This technique facilitates a deeper understanding of object structure and improves recognition accuracy, making GLOM a promising tool in computer vision and artificial intelligence. According to Hinton, the human brain forms a "parse tree" that illustrates the hierarchical relationships between the whole object and its parts. In this model, the face occupies the top position, and its elements—the eyes, nose, ears, and mouth—act as branches. One of the key goals of GLOM is to reproduce this tree in the architecture of a neural network, which allows for more efficient processing and recognition of visual information.

How GLOM Works

GLOM uses advanced image processing techniques. Visual content, such as photographs of faces, is divided into a grid, where each cell corresponds to a specific region of the image. One cell might capture the iris, while another captures the tip of the nose. For each region, multiple layers are provided that sequentially create a prediction with a vector containing information about the content. This approach ensures high accuracy of image recognition and analysis, enabling the efficient processing of visual data in a variety of applications.

Each vector layer generates predictions, ranging from simple statements, such as "I represent the tip of the nose," to more complex ones, such as "I am part of a face in a side view." This multi-layered approach allows GLOM to more effectively identify and classify objects in an image, significantly improving the accuracy of visual data analysis.

Nick Frost, a Google Brains employee, offered an interesting analogy to explain the work of artificial neurons. Imagine a group of people discussing the same idea, where each participant makes their own unique changes. Vectors in the GLOM system function in a similar way, striving to reach a common understanding. Over time, they develop a shared understanding of the concept, which strengthens and refines their perception of the image through collective interaction. This analogy helps to better understand how collective processes in neural networks contribute to a deeper understanding and interpretation of visual information.

GLOM's Prospects: A Breakthrough in Artificial Intelligence

Geoffrey Hinton, one of the leading experts in the field of artificial intelligence, expressed confidence that GLOM could become a significant breakthrough in the field. He believes that for significant progress in the development of AI, it is necessary for the system to learn to effectively solve complex problems faced by people. GLOM will be able not only to perceive new information but also to use accumulated experience based on the analysis of previously processed data. This will allow the system to generalize, extrapolate, and experiment with different concepts, which in turn will open new horizons for artificial intelligence.

Hinton emphasizes that GLOM is currently more of a theoretical concept than a definitive solution. Chris Williams, Professor of Machine Learning at the University of Edinburgh's School of Computer Science, notes that GLOM has the potential to change paradigms in the field of artificial intelligence. However, there is currently insufficient evidence to assess its actual impact on technology. Despite the uncertainties, the GLOM concept appears promising and requires further research to unlock its full potential in AI.