登录
根据短文内容，从下面的选项中选出可以填入空白处的最佳选项。选项中有两项为多余选项。
What Are “Large World Models”?
What is the next big thing in AI? Many people believe it’s world models. AI pioneer Fei-Fei Li’s World Labs has raised 230 million to build“large world models”. 41. _________. To understand world models, let’s first look at human mind models. Our brains collect useful information from our senses, like sight and sound, and turn it into clear understandings of the world. This helps us form mind models, which we use to make predictions. 42. _________. For example, a baseball player can easily predict a ball’s direction and swing without thinking, thanks to their mind model of the game.
World models are inspired by these mind models. They also make predictions about what will happen, using current(当下的) information and past data. Though the idea has existed for years, it only became popular when applied to generative(生成) video.
Most AI-generated videos look strange. You might see arms and legs twist unnaturally. This is because newly-made models can predict what happens like a basketball bouncing(弹跳) but not why. Similarly, language models don’t truly understand word meanings. 43. _________.
To achieve this, world models are trained on large data, including photos, videos, and texts. 44. _________.
Better video generation is just the start. 45. _________. For example, French computer scientist Yann Lecun described how a world model could clean a dirty room: it would plan steps like cleaning the floor, washing dishes, and taking out trash—not by following past patterns, but by understanding how to turn a messy space into a clean one.
A. People can use AI to talk with their friends easily.
B. The goal is to help them understand how the world works and reason about the results of actions.
C. And DeepMind, a famous AI company, is also spending much in this field.
D. Researchers say world models could one day handle complex tasks in both digital and real worlds.
E. Many students like to learn AI in their free time.
F. World models, however, can know basic reasons, making video generation more natural.
G. These predictions then shape how we understand and react to the world around us.
What Are “Large World Models”?
What is the next big thing in AI? Many people believe it’s world models. AI pioneer Fei-Fei Li’s World Labs has raised 230 million to build“large world models”. 41. _________. To understand world models, let’s first look at human mind models. Our brains collect useful information from our senses, like sight and sound, and turn it into clear understandings of the world. This helps us form mind models, which we use to make predictions. 42. _________. For example, a baseball player can easily predict a ball’s direction and swing without thinking, thanks to their mind model of the game.
World models are inspired by these mind models. They also make predictions about what will happen, using current(当下的) information and past data. Though the idea has existed for years, it only became popular when applied to generative(生成) video.
Most AI-generated videos look strange. You might see arms and legs twist unnaturally. This is because newly-made models can predict what happens like a basketball bouncing(弹跳) but not why. Similarly, language models don’t truly understand word meanings. 43. _________.
To achieve this, world models are trained on large data, including photos, videos, and texts. 44. _________.
Better video generation is just the start. 45. _________. For example, French computer scientist Yann Lecun described how a world model could clean a dirty room: it would plan steps like cleaning the floor, washing dishes, and taking out trash—not by following past patterns, but by understanding how to turn a messy space into a clean one.
A. People can use AI to talk with their friends easily.
B. The goal is to help them understand how the world works and reason about the results of actions.
C. And DeepMind, a famous AI company, is also spending much in this field.
D. Researchers say world models could one day handle complex tasks in both digital and real worlds.
E. Many students like to learn AI in their free time.
F. World models, however, can know basic reasons, making video generation more natural.
G. These predictions then shape how we understand and react to the world around us.
答案
41.C 42.G 43.F 44.B 45.D
解析
我们可以根据上下文逻辑逐一匹配选项：
1. 41空：前文提到AI先驱李飞飞的实验室筹集2.3亿美元研发大世界模型，C选项“知名AI公司DeepMind也在该领域投入大量资源”是对前文各大企业布局该领域的顺承补充，衔接自然。
2. 42空：前文提到人类大脑形成心智模型后会用它来做出预测，G选项“这些预测随后塑造了我们理解、回应周围世界的方式”承接前文中的“make predictions”，和后文棒球运动员的举例逻辑连贯。
3. 43空：前文指出普通AI生成模型只能预测事件发生的表象、无法理解背后的原因，F选项“然而世界模型可以知晓基础原理，让视频生成更加自然”通过转折突出世界模型的优势，符合上下文逻辑。
4. 44空：前文提到世界模型会使用大量图片、视频、文本数据进行训练，B选项“其目标是帮助它们理解世界的运行规律，推理行为带来的结果”点明训练的目的，衔接通顺。
5. 45空：前文指出更优质的视频生成只是世界模型的初步作用，后文举例说明世界模型可以完成打扫房间这类现实复杂任务，D选项“研究者表示未来世界模型可以处理数字世界和现实世界中的复杂任务”引出后文的举例，逻辑匹配。
其余A、E两个选项和本文介绍大世界模型的主题无关，为多余项。
1. 41空：前文提到AI先驱李飞飞的实验室筹集2.3亿美元研发大世界模型，C选项“知名AI公司DeepMind也在该领域投入大量资源”是对前文各大企业布局该领域的顺承补充，衔接自然。
2. 42空：前文提到人类大脑形成心智模型后会用它来做出预测，G选项“这些预测随后塑造了我们理解、回应周围世界的方式”承接前文中的“make predictions”，和后文棒球运动员的举例逻辑连贯。
3. 43空：前文指出普通AI生成模型只能预测事件发生的表象、无法理解背后的原因，F选项“然而世界模型可以知晓基础原理，让视频生成更加自然”通过转折突出世界模型的优势，符合上下文逻辑。
4. 44空：前文提到世界模型会使用大量图片、视频、文本数据进行训练，B选项“其目标是帮助它们理解世界的运行规律，推理行为带来的结果”点明训练的目的，衔接通顺。
5. 45空：前文指出更优质的视频生成只是世界模型的初步作用，后文举例说明世界模型可以完成打扫房间这类现实复杂任务，D选项“研究者表示未来世界模型可以处理数字世界和现实世界中的复杂任务”引出后文的举例，逻辑匹配。
其余A、E两个选项和本文介绍大世界模型的主题无关，为多余项。
