Polina Kolozaridi, head of the DH Center at ITMO University and senior researcher at EUSPb, has prepared a review of Eugene Charniack's book The Intellectual History of Artificial Intelligence: AI and Me published by the Gaidar Institute Press in 2025.
Eugene Chernyak's book is a basic text on the history of the field. In my opinion, it is inferior to the book Deep Learning: An Introduction to Neural Networks[1]. However, Chernyak's book focuses on the conflicts, mistakes, and questions that arose among developers and scientists. As a result, the history of AI turns out to be seamless, although in some chapters the seamlessness resembles chaos.
On the other hand, the book ties together different stories: from Minsky's frames[2] to GPT-4.
Stories from the past (for example, the Shayki robot or the SHRDLU program) are described in a more interesting way than modern ones. Old robots and programs from the 1970s need to be studied because they are used by Lucy Sachman and other social researchers of AI. Their examples show that “acting according to the rules” and “creating rules” have long been two different tasks for developers.
Rules are not a random topic; they are one of the main themes of the book. After all, if rules can be clearly stated, they can be worked with, applied, and changed in different cases. And if machines can form the rules and tasks by which they operate, then the problem of AI control is not far-fetched. Chernyak carefully avoids the question of how the rules of modern machines are structured, although he complains that the creators of GPT-4 do not reveal anything about the inner workings of their developments.
The question of the subject of history, “who sets the tasks and why,” is rarely asked from within the engineering history of technology. “The party said ‘we must!’, and the Komsomol replied ‘yes!’”
Chernyak describes how tasks arose at the level of department heads and scientists or were formed around the automation of specific practices: from playing chess to recognizing proteins.
In my opinion, this is a normal stage: when the history of technologies is written by their creators, they, like commanders describing wars, do not ask questions about who is fighting whom and for what. But this is better than generalizations that make their authors look like Caesars, capable of both winning and describing the Gallic War.
In this sense, Chernyak's book is useful and modest. It is not very suitable for those who are hearing about transformers or image recognition for the first time. But if you are familiar with the reality behind each of these terms, you can read Chernyak's story.
At the same time, it is worth looking at related descriptions such as those by Matteo Pasquinelli[3]. Although his view is not particularly modest, there are not many social histories of AI written yet, and only one has been translated into Russian.
In addition to the rules, the book also has other recurring themes. First and foremost, these are the connections and disconnections between different conceptual frameworks (symbolic/non-symbolic AI), the universality of models, understanding probability theory, the role of video cards and memory, and so on. I repeat, there is no coherence in these themes yet, but Eugene Chernyak's book provides a good context for understanding how they work. It is also written without fanfare and includes a timeline and references.
[1]Arkhangelskaya, E. V., Kadurin, A. A., Nikolenko, S. I. (2020) Deep Learning: Immersion in the World of Neural Networks. St. Petersburg: Piter.
[2]Minsky, M. (1979) Frames for Representing Knowledge. Moscow: Energiya.
[3]Pasquinelli, M. (2023). Measure and Impose: A Social History of Artificial Intelligence. Moscow: Individuum.