ChatGPT’s Translation Skills Parallel Most Human Translators

A new study published in IEEE Transactions on Big Data found that ChatGPT's GPT-4 model performs at the level of junior and medium-level human translators, marking potentially the first time an AI algorithm has reached human-level translation quality. Only senior translators with 10+ years of experience and professional certification clearly outperformed the AI models.

• →GPT-4 achieved translation quality comparable to junior and medium-level human translators across multiple language pairs.
• →Senior translators with 10+ years experience still outperformed AI models, producing only 1.83 major errors versus GPT-4's 3.71 errors per 200 sentences.
• →AI models tend to be overly literal while human translators sometimes overinterpret ambiguous text.
• →Both humans and AI struggled more with uncommon language combinations like Chinese to Hindi.
• →This represents a major paradigm shift in AI capabilities, with potential implications for the translation industry.