NASA’s new dark energy space telescope can also detect killer asteroids
technologyreview.com·1h ago
AI Research
arXiv CS.AI·Jun 15·5 min read
Can professional translators identify machine-generated text?
TL;DR
Professional translators often struggle to identify texts generated by artificial intelligence (AI). In an experiment, 69 translators evaluated three short stories—two by ChatGPT-4o and one by a human—rating the likelihood of AI authorship.
✦ Why It Matters
Understanding how translators perceive AI-generated text can inform better training and tools for content evaluation.
Key Takeaways
- Train translators on identifying AI-generated content.
- Focus on indicators like burstiness and narrative consistency.
- Evaluate subjective impressions in translation assessments.
- Explore implications for synthetic-text editing in translation.
How It Works
Translators assessed stories based on their familiarity with narrative structures and linguistic features. They rated the likelihood of AI authorship, using analytical skills to identify patterns typical of machine-generated text.