Yandex is advancing the capabilities of its YandexGPT by training the AI to recognize and translate text from images, audio, and video files. The company is actively seeking translators to assist in this ambitious project, with plans to hire approximately ten specialists to train the model.
According to sources from Vedomosti, Yandex has posted vacancies for AI trainers who are proficient in English and possess professional translation skills. These experts will play a crucial role in refining the generative model by providing reference translations and evaluating the quality of the AI’s output. The translation tasks will cover a wide range of specialized fields, including jurisprudence and physics, making the role particularly challenging.
The primary challenge, as highlighted by experts, lies in the fact that training modern language models requires translations that are semantically accurate rather than literal. This is a shift from the traditional approach and is necessary because neural networks often default to literal translations, which may not always convey the intended meaning.
Yandex’s efforts to enhance its neural network are part of a broader strategy to improve its AI-driven services. In June, the company unveiled a new version of its machine translation service, powered by YandexGPT. By July, YandexGPT was already being used to correct and enhance text within Yandex Keyboard, demonstrating the company’s commitment to integrating AI advancements across its platforms.
This new initiative to train YandexGPT to handle complex translations from multimedia sources marks a significant step forward in Yandex’s AI development, promising more versatile and accurate language processing tools in the near future.
Source: Vedomosti