The Evolution of AI

  The early development of artificial intelligence(AI) began in the 1950s, and its origin lies in Alan Turing’s Turing test. Turing’s concept of intelligence was based on the machine’s capability to think and whether its behavior was ‘indistinguishable’ from that of the human being. In the 1960s, artificial intelligence research shifted to algorithm and rule-based approaches. This approach led to the emergence of ‘expert systems’, referring to software systems that mimic expertise in a particular field. 

  As we entered the 1990s, machine learning began to gain attention. Machine learning aimed to develop algorithms that self-educate themselves from data. During this time, algorithms such as the Support Vector Machine (SVM) were developed. SVM was recognized as a powerful tool to solve complex nonlinear classification problems. 

 

Breakthrough points of AI 

  AI has increased the opportunity to apply previously imagined theories to computers due to advances in hardware. For instance, Convolutional Neural Network(CNN) is a concept that was published in 1989. In 2012, with the development of hardware, Alex Net, the first image classification was constructed, showing it effective. He started to conceive not only computer vision but also various natural nets. Gradually, many operations such as Generative AI, natural language processing, and reinforcement learning were made. 

 

Birth of GPT

  Representative natural language processing models before GPT include Recurrent Neural Network (RNN) based models such as Long Short-Term Memory (LSTM). These models were useful for processing sequence data, but they were limited in remembering historical information for long periods of time. 

  Transformer-based models such as GPT, have overcome this limitation. Transformer utilizes the ‘attention’ mechanism to consider the relationship between all words within a sentence, which improves contextual understanding and generation ability. This difference can be seen in the paper Attention is All You Need. 

  Models after GPT became more capable of understanding and generating natural language like humans, used in various fields, and influencing human life. Especially in the corporate environment, it was used as a chatbot or customer service bot to improve customer service. Like a tutor and a medical service bot, it made daily human life more convenient. 

 

Understanding AGI: The Present State of AI

  Artificial General Intelligence (AGI) is an artificial intelligence that can perform all kinds of cognitive tasks by imitating human intelligence. AGI is not meant to just perform specific tasks, but also to have the ability to think, learn, and solve problems like humans. Currently, there are shortages in fully implementing these general AIs, but research is constantly progressing. 

 

The Future of AI: Scholar’s Perspectives 

  What will the next generation of AI look like? Some claim that AI will completely go beyond human intelligence, referring to it as superintelligence, while others argue that AI cannot perfectly imitate the human brain, suggesting advances in AI will be made in cooperation between humans and machines. 

  The development of AI is changing our lives. It is still ongoing. You, Postechians, your passion and creativity can open up new possibilities.