登录
B 热考内容 时事新闻
KEY→P37
范畴:人与社会 主题:科学技术 体裁:说明文 难度:★★★★ 建议用时:7 分钟 评价:___/5
A social media conversation is popular. It's about counting the number of “r”s in “strawberry”. ChatGPT thinks there are two “r”s in the word “strawberry”. In fact, other AI models also have trouble with this basic counting task.
How can these powerful AI chatbots have trouble with such a basic question? It comes down to their unique way of understanding text.
When it comes to understanding text, humans and AI are quite different. For us, “strawberry” is easily understood as a string of characters(一串字母). Therefore, it's simple to count the letters. AI models, on the other hand, don't see words in the same way. They rely on a process called tokenization. This process breaks down text into smaller units, or tokens. These can be entire words, parts of words, or even single characters. That depends on the model's training. For example, “strawberry” might be tokenized as “straw” and “berry”. It might even be marked as a single token if the word is often in the training data.
When you ask an AI chatbot to count letters, such as “r”s in “strawberry”, it won't examine the characters themselves. Instead, it sees larger parts, like “straw” and “berry”, but misses the detailed breakdown humans easily see.
In our AI-driven world today, we increasingly rely on large language models for quick answers and ideas. These powerful tools have their limitations(局限性), and they don't always get the details right.
Understanding how AI works—and its limitations—helps us to use these tools more effectively.
The strawberry test is a good example of how important it is to check and sometimes double-check information from AI models. Even the “smartest” models can make simple mistakes.
( )1. How do AI models understand text?
A. They see words as a string of characters.
B. They rely on tokenization to break down text.
C. They count the letters directly.
D. They understand words like humans.
( )2. What can we infer from the passage?
A. AI models will never make mistakes in the future.
B. We should always trust the information from AI models.
C. AI models are not useful at all because they have limitations.
D. We need to be careful when using information from AI models.
( )3. What does the underlined word “entire” in Paragraph 3 probably mean?
A. Difficult. B. Easy. C. Whole. D. Separate.
( )4. 新考法·行文方式 How does the author introduce the topic of how AI models understand text?
A. By telling a story. B. By making comparisons.
C. By listing numbers. D. By asking questions.
( )5. The passage mainly tells us ______.
A. AI models are very powerful and useful
B. AI models have trouble counting letters
C. humans and AI understand text differently
D. we should understand how AI works and its limitations
KEY→P37
范畴:人与社会 主题:科学技术 体裁:说明文 难度:★★★★ 建议用时:7 分钟 评价:___/5
A social media conversation is popular. It's about counting the number of “r”s in “strawberry”. ChatGPT thinks there are two “r”s in the word “strawberry”. In fact, other AI models also have trouble with this basic counting task.
How can these powerful AI chatbots have trouble with such a basic question? It comes down to their unique way of understanding text.
When it comes to understanding text, humans and AI are quite different. For us, “strawberry” is easily understood as a string of characters(一串字母). Therefore, it's simple to count the letters. AI models, on the other hand, don't see words in the same way. They rely on a process called tokenization. This process breaks down text into smaller units, or tokens. These can be entire words, parts of words, or even single characters. That depends on the model's training. For example, “strawberry” might be tokenized as “straw” and “berry”. It might even be marked as a single token if the word is often in the training data.
When you ask an AI chatbot to count letters, such as “r”s in “strawberry”, it won't examine the characters themselves. Instead, it sees larger parts, like “straw” and “berry”, but misses the detailed breakdown humans easily see.
In our AI-driven world today, we increasingly rely on large language models for quick answers and ideas. These powerful tools have their limitations(局限性), and they don't always get the details right.
Understanding how AI works—and its limitations—helps us to use these tools more effectively.
The strawberry test is a good example of how important it is to check and sometimes double-check information from AI models. Even the “smartest” models can make simple mistakes.
( )1. How do AI models understand text?
A. They see words as a string of characters.
B. They rely on tokenization to break down text.
C. They count the letters directly.
D. They understand words like humans.
( )2. What can we infer from the passage?
A. AI models will never make mistakes in the future.
B. We should always trust the information from AI models.
C. AI models are not useful at all because they have limitations.
D. We need to be careful when using information from AI models.
( )3. What does the underlined word “entire” in Paragraph 3 probably mean?
A. Difficult. B. Easy. C. Whole. D. Separate.
( )4. 新考法·行文方式 How does the author introduce the topic of how AI models understand text?
A. By telling a story. B. By making comparisons.
C. By listing numbers. D. By asking questions.
( )5. The passage mainly tells us ______.
A. AI models are very powerful and useful
B. AI models have trouble counting letters
C. humans and AI understand text differently
D. we should understand how AI works and its limitations
答案
1.B 2.D 3.C 4.B 5.D
