Now it's 2024, but it seems like 2022, 2 years ago, is the “first year of generation A I,” and people who wrote papers and created works using ChatGPT are also around, and I feel more “moving” than when the internet came out into the world. People working on new businesses now expect it to become a big name in 10 years.
Well, shamefully, I didn't really understand “generative AI,” and I found this book when I was going around the bookstore's new edition section thinking I'd study something. What I was unexpectedly convinced by reading in some book is the phrase “AI does not take away people's jobs; people who master AI do (take away) the jobs of people who can't master it.” Starting with this book, I would like to continue studying “generative AI” and become my own hands and feet.
Below are the points I was concerned about.
・We are living in a completely different world from just a little while ago, and unbelievably many things have happened surrounding a certain technology (generative AI) revolution, and it is even said that it is a turning point in human history (p3) Therefore, in this book, we have decided to focus on the technical basics of generative AI, which is thought to be universally important in the future, and topics of social impact that will surely occur (p6)
・Generative AI is a type of artificial intelligence technology that can create new sentences, images, sounds, etc., and current generative AI is realized by deep learning (deep learning), and deep learning indicates a method for learning a deep neutral network that imitates a deep neutral network that is particularly close to the human brain among the elemental technologies of artificial intelligence called machine learning (p18)
・The impact of generative AI is so great that changes are occurring at a speed that is incomparable to conventional general-purpose technology, and if I (the author) in the first half of 2022 were brought to the present day of 2023, I can't imagine how surprised I would be when I saw headlines on the internet, articles, SNS, and newspapers (p27)
・The ultimate goal of AI researchers is the realization of artificial general intelligence (AGI: Artificial General Intelligence), and human intelligence is truly “general-purpose”, capable of performing various things such as language use, visual processing, movement, and game play (p31)
・ChatGPT was realized by performing fill-in-the-blank problem learning, supervised fine tuning, and reinforcement learning based on human feedback on a large scale (p66)
・What should humans do when they encounter a situation that cannot be answered by their own knowledge alone? Surely they will try to obtain necessary information using search engines, and the same goes for language generation AI. In this way, technology that makes language generation AI respond by referring to external knowledge is called RAG (Retriveval Augmented Generation = generation enhanced by search); to put it simply, this RAG is implemented in Bing chat and Google Bard, which are search engine technology used by AI rather than humans (p95)
・Most occupations that are less likely to be affected by AI are those that perform manual labor that moves hands and feet, and occupations that are susceptible are occupations that require advanced judgment and creative thinking, such as engineers, researchers, designers, etc., and so-called white collar occupations. As a final conclusion, 80% of all occupations are affected in some way, and about 20% of them are labor It is said that it will be affected by a level that completely replaces AI (p104)
・The shorter the training required, or the lower the wage, the less likely it is to be affected by AI; conversely, occupations with longer training periods and higher wages are more susceptible to AI; however, this does not apply to occupations that require advanced skills that really take time to learn, (p105)
・If you describe a theory that is somewhat convincing at the moment, most of what has been regarded as human creative work is actually “finding valuable new combinations from past experiences,” and generative AI can now find this from vast amounts of data learning (p111)
・The two fields of customer service and software development are characterized by the fact that the introduction of generative AI was progressing prior to the generative AI boom. The strength of generative AI in customer service is that interaction with customers can be semi-automated while achieving high customer satisfaction and acceleration in some cases without degrading service quality. Also, in this field, productivity is low among new employees, and despite high training costs, there is a background where the turnover rate is high, which is a strong motivation for companies to introduce generative AI (p122)
・Creativity proposed by a researcher named Bowden is 1) combinatorial creativity (= by combining existing ideas and knowledge (including subtraction), 2) exploratory creativity (= by searching for existing ideas and knowledge according to some rules or procedures), 3) innovative creativity (p157) of these, 1) and 2) are possible with generative AI, and 3) is difficult for humans (p158)
・When creating something by demonstrating creativity, I believe that not only is it simply a combination of things, but that a creative act reflecting these personal feelings, thoughts, and stories is projected, and that personal feelings, thoughts, and stories is worthy of being called “creation” (p162)
・Since work originally performed by humans will disappear, humans will be able to spend time on other activities, and if it is an activity that only humans can do, it is not routine office work or material preparation, but rather creating ideas for fundamental business reforms and rethinking how to contribute to the future of society and humans (p199)
・In terms of living standards, it is about increasing time spent socializing and time that individuals can “enjoy”; it should be games, watching videos, or reading; AI does work on behalf of humans, but it cannot be enjoyed in place of humans; free time allows humans to demonstrate creativity and originality, and is a source for creating new discoveries, inventions, and artistic expressions (p200)
・From now on, it's not easy, but it's better to be willing to act, move as much as possible, and from a higher point of view, “I think it's better to raise one's own metacognition a little more = observe one's own behavior and thought habits from a bird's-eye view and be able to do this” (p233)
・What is needed is a skill that can give good instructions to AI, prompt engineering, that is, skills to optimize instructions and instructions are necessary in order to obtain desirable output from AI (p234) how to deal with changes in society, the actions of people in the world are polarized, and it is natural for them to transform, and on the other hand, there are also people who want to do something and are not bought after all. I think the difference between the two will become more and more clear in the future (p235)
