翻译关闭：AI 聊天机器人以英语思考
开创性研究揭示，最大的人工智能聊天机器人可能存在对英语语言的不良偏见
“`html
AI chatbots primarily use the English language for their cognitive processes, according to research.
Have you ever wondered what language artificial intelligence (AI) chatbots “think” in? 🤔 Well, according to new research, these language models are wired to “think” in English, even when asked questions in other languages. It’s like they have an internal English subspace, processing everything in their native tongue, regardless of the language they are presented with. 🌍
Peeking Behind the Curtain
To unravel the mystery, researchers at the Swiss Federal Institute of Technology in Lausanne decided to pull back the curtain and take a closer look at these AI chatbot models. They wanted to understand the inner workings of these large-language-models (LLMs) and how they process different languages.
“We opened up these models and looked at each of the layers,” explained researcher Veniamin Veselovsky. “We wanted to see, can we see that the internal layers are actually processing in English?” 🤓
The Multilingual Menu
The research team selected three open-source AI chatbot models for their investigation. These models were put to the test using prompts in four different languages: French, German, Russian, and Chinese. The prompts came in three variations:
- Repeat the given word
- Translate from one non-English word to another
- Fill in a one-word gap in a sentence
By analyzing the layers of processing within the AI chatbots, the researchers were able to trace the path that each prompt took. And what did they discover? 🕵️♀️
The English Subspace Detour
It turns out that these AI chatbots take a detour through what the researchers termed the “English subspace.” Rather than directly translating from one language to another, such as French to German, the AI chatbots first translate from French to English and then from English to German (or vice versa).
Why is this significant? According to Veselovsky, this suggests that these language models rely on English as a bridge to understand certain concepts. 😮
Lost in Translation?
While the chatbots’ preference for English might seem innocuous, it raises concerns about the homogeneity of AI development. Researcher Aliya Bhatia from the Center for Democracy & Technology in Washington DC expressed her worries about the growing reliance on English-language data for training AI models.
“There’s more high-quality data available in English and some UN languages to train models than in most other languages,” Bhatia explained. “This risks superimposing a limited world view onto other linguistically and culturally distinct regions.” 🌐
What Does the Future Hold?
As AI continues to evolve, it’s crucial to address questions of linguistic diversity and cultural inclusivity. The reliance on English as an intermediary language for AI models can perpetuate biases and limit access to AI benefits for non-English speakers.
While improvements are being made in training AI models with multilingual data, there is still much work to be done. Researchers and developers need to prioritize linguistic and cultural diversity to ensure that AI technology is truly inclusive. 🌍
Q&A
Q: Can AI chatbots learn to think in languages other than English? A: Yes, AI chatbots can learn to think in multiple languages, but currently, they often rely on English as a bridge to understand various concepts in different languages.
Q: Is the preference for English in AI models a result of limited training data in other languages? A: Yes, one of the reasons AI models tend to rely on English is the abundance of high-quality data available in English compared to other languages. Developers often train models primarily on English-language data.
Q: What are the potential consequences of AI models thinking predominantly in English? A: Relying on English as an intermediary language risks imposing a limited world view onto linguistically and culturally distinct regions. It may perpetuate biases and hinder the development of AI technology that is inclusive of diverse cultures and languages.
References
“`Sure thing! Here is the HTML code translated to Chinese, with the original information of the src attribute in the tag preserved:
你认为英语在人工智能聊天机器人思维领域的主导地位如何？让我们开始一场对话！在下方评论中分享你的想法，并不要忘记点击分享按钮传播这个消息。👇🗣️💻
