There have been many predictions about whether AI and advanced generative AI will take over human jobs. However, the impact of generative AI is beginning to become apparent. The latest language model, GPT4, has reached a level that it can pass the bar exam and the national medical exam. It has also reached a level that it can pass Google’s coding test. The impact of AI appears to be beginning to be felt among engineers, researchers, and designers. In China, the use of image generative AI has reduced illustrators’ salaries by one-tenth. Occupations likely to be affected by this impact appear to be those requiring advanced judgment and creative thinking. Furthermore, this impact will be felt most strongly in highly educated, highly skilled, and highly paid occupations.
A tool to solve work and language challenges is emerging. This is generative AI. ChatGPT is a prime example. Large-scale language models have the ability to realize unprecedented conversational communication with humans. The emergence of this language model has made it possible to gather information through conversation with AI. This generative AI can work 24 hours a day, without rest. AI never tires. To obtain output from this language model generative AI, an appropriate “prompt” is presented. The instructions entered by humans are called “prompts.” Even when requesting the same answer, the output can be completely different simply by changing the prompt. For example, current image generation AI can be instructed in the text, “I want you to generate an image of this painting,” and it will produce professional-quality illustrations and images indistinguishable from real photographs, just as desired. Including terms like “Van Gogh” and “Sharaku” in the prompt will result in more detailed output.
Generative AI is a labor-complementary technology. There is a growing belief that the skillful use of generative AI can make existing work more productive, comfortable, and of higher quality. Whether labor-complementary generative AI leaves room for human intervention depends on the original complexity of the task. When tasks are complex enough, it is humans who will use generative AI to increase efficiency. There will always be a need for people with the discerning eye to verify the programs created by AI and determine whether they are correct. In other words, we need to train people who can control the technology itself and who can responsibly select the final output. Understanding the fundamental technology behind system design is essential, and in addition to knowledge of AI, a sense of ethics is also essential. It seems that education in the future will develop people who can freely manipulate AI, rather than be used by it.
