午夜剧场

Generative AI, including ChatGPT, is attracting significant attention, and bookstores are filled with books enlightening people about generative AI. Since I am publishing a new book in this environment, I decided to explain the risks and impacts of generative AI in addition to its benefits. In fact, generative AI is an incomplete technology, and companies must face its risks and impacts head-on when using it for business operations.

Yukichi Fukuzawa left us with the words, "In the world of belief, there is much deception," and the mechanism of generative AI is simply connecting words probabilistically to create plausible-sounding sentences. Naturally, there are many mistakes, and one should not blindly believe the output of generative AI. Furthermore, generative AI has the problem of bias. For example, in the case of ChatGPT, it has been pointed out that it generates liberal-leaning text. This is said to be because OpenAI, which builds and operates ChatGPT, removed texts containing slanderous expressions from the training data to avoid generating slanderous content. It is speculated that this is because texts written by non-liberals contained relatively more slanderous expressions, and as a result, the learning of non-liberal texts decreased. In other words, it is easy to bias generative AI. For example, when having generative AI create a cooking recipe, stealth marketing could occur, such as including a specific seasoning from a specific manufacturer. Additionally, generative AI is an attractive technology for authoritarian states. Since generative AI easily provides the answers we want to know, we tend to be satisfied with the results and stop bothering to investigate the information that formed the basis of the answer. If an authoritarian state provides its citizens with a generative AI that excludes information inconvenient to the state, those citizens will be trapped in an information space that is convenient for the state.

Now, the aforementioned quote by Yukichi Fukuzawa, "In the world of belief, there is much deception," has a continuation: "In the world of doubt, there is much truth." In the era of generative AI, it is unfortunately expected that situations will be filled with disinformation and misinformation, and as a result, humans will be required to have the ability to discern correct information from a jumble of data. I hope this book will be of some help in that regard.

Ichiro Sato

Chuko Shinsho La Clef

240 pages, 968 yen (tax included)

*Affiliations and titles are as of the time of publication.