Hitachi to develop Generative AI that preserves the experience of retiring employees18.08.2023 | 01:23 |
The Japanese company hopes that artificial intelligence will help new employees master the intuition that comes with many years of experience.
At the moment, any Internet user can turn on ChatGPT or other generative artificial intelligence and ask him questions, but some companies are going to use artificial intelligence not only for communication, but for their professional purposes, so they are developing specialized versions of generative AI. Now Hitachi has been added to these companies.
A Japanese electronics manufacturer is developing artificial intelligence that can preserve the unique experience of retiring employees, in the hope that future employees will be able to learn from the "electronic ghosts" of retired colleagues.
Hitachi management believes that the vast professional experience that retiring employees have received cannot be contained in the directory, so artificial intelligence will take up this task. The company sees a future in which its own artificial intelligence will be trained on the basis of the important knowledge that people receive throughout their careers spent at power plants, manufacturing plants and other jobs of high complexity.
Nikkei Asia reports that Hitachi is interested in this way of training its new employees. They will be able to adopt invaluable experience and a deep understanding of high-tech equipment that allows an employee to notice changes in sound, temperature or smell in production, which can be a sign of an approaching malfunction, which can be dangerous for workers (and for Hitachi profits).
It is difficult to explain such "esoteric" concepts in the employee's instructions, so Hitachi sees the solution to this problem in its artificial intelligence. The new AI will show the trained employees, for example, various situations at the control room of a power plant or a railway station. Then it will add various predictors of malfunction, such as smoke or flashing sensors. Interns will solve problems by learning from past failures that artificial intelligence has been trained on.
Hitachi is also developing a text-based AI that looks more like ChatGPT. This AI will be trained in a similar way, allowing employees to access their database compiled based on the experience of previous employees. This will help employees to eliminate problems that arise, for example, to find possible causes of a certain series of signals and warnings at a power plant.
Generative AI became popular last year due to its ability to create content such as text and images, which in some cases can pass for the work of professionals. However, the accuracy of generative AI leaves much to be desired. These systems are prone to "hallucinations" that produce completely made-up information. Generative AI is not intelligence in the sense that we know it: the system does not know anything, but it absorbs so much data that it can build a pretty good answer to a correctly specified user query. It's like a word calculator. Should such an AI really tell the technicians at a nuclear power plant what to do?
Nevertheless, with appropriate "narrow" training, it is possible to create a model that will accurately answer questions about specific problems. Hitachi is moving in this direction, and not only it. Perhaps public chatbots will eventually fade into the background, as more and more companies are developing their own highly specialized chatbots with AI.