压轴题02 阅读理解C、D 篇
命题预测
分析近几年高考英语阅读理解 C、D 篇可知，人工智能类说明文是高频压轴题材，选材
贴合时代热点，语篇多来自英美科技媒体、科研报告、高校研究发布，主题聚焦 AI 技术原
理、应用场景、伦理争议、社会影响、未来发展等。文章逻辑性强、长难句多、专业术语常
见，侧重考查信息定位、逻辑推理、主旨概括等高阶能力。2026 年高考仍会将人工智能类作
为 C、D 篇核心考查方向，命题更关注 AI 与教育、医疗、生活、科研、环保、版权等领域
的结合，强调辩证思考与实际应用。
高频考法
1. 推理判断题
2. 标题归纳题
3. 细节理解题
4. 词义猜测题
5. 主旨大意题
6. 观点态度题
一、人工智能类说明文基本规律及解题要领
高考人工智能类阅读通常无标题，结构稳定，逻辑清晰，一般分为四部分：
首段：开门见山引出 AI 核心话题 —— 新技术、新模型、新争议、新研究。
背景：介绍 AI 发展现状、传统技术局限、现实需求或争议起源。
主干：详细说明 AI 工作原理、功能特点、实验数据、应用场景、优势与问题。
结尾：总结 AI 价值、未来前景、现存挑战、专家观点或社会反思。
二、人工智能类说明文的解题技巧
1. 抓结构，快速把握主旨
用略读法浏览首尾段 + 各段首尾句，圈出 AI 相关核心词（model/algorithm/LLM/chatbot/robot 等）。
人工智能类文章常见说明思路：
技术介绍型：原理→特点→优势→应用→局限
社会争议型：现象→正方观点→反方观点→作者态度
研究发现型：实验目的→过程→数据→结论→展望
2. 定位标志词，精准破解细节与推理
题干关键词：人名、机构、专业术语、数字、时间、转折词。
长难句处理：先找主句谓语，剥离定语、状语、插入语，抓核心意思。
答案原则：原文同义替换 / 合理概括，不主观臆断。
3. 重点关注观点态度与引语
文中researchers/experts/developers/authorities 所说的话，常是观点题、推理题题眼。
把握情感词：positive/negative/concerned/skeptical/optimistic/cautious。
4. 紧盯转折逻辑，锁定核心信息
AI 类文章高频转折词：however / but / yet / while / although / on the other hand 转折后往往是作
者真正观点、核心问题、重要结论，是命题重灾区。
5. 熟悉选项陷阱，快速排除干扰
正确选项：原文信息同义改写、合理归纳。
干扰项：
张冠李戴（把 A 的功能安到 B）
偷梁换柱（改变程度、范围、对象）
无中生有（原文未提及）
以偏概全（只讲局部当主旨）
6. 标题归纳技巧（
AI
类专用）
必须包含AI/technology/model 等核心概念。
范围适中，不夸大、不缩小。
常见格式：AI + 功能 / 争议 / 未来 / 应用。
02 人工智能类
1.（2026·青岛·一模）
Artificial intelligence (AI) researchers have long dreamed of tools to supercharge science-asking novel
questions, designing and running experiments. Recently, large language models (LLMs) have made
discoveries that some AI developers claim have inched us closer to that future. But how do you test
whether an AI model can truly do science?
For answers, researchers turn to benchmarks (基准): standardized sets of questions or tasks that help
measure an AI’s efficiency and reliability and compare it against other models. But the complexity of
science makes assessing their aptitude especially challenging. As Hao Peng, a computer scientist at the
University of Illinois Urbana-Champaign, puts it: “Models have all this knowledge. Do they know how
to use it?”
Dozens of new science-focused benchmarks have emerged over the past year to answer that question,
but scientists have yet to settle on a single best approach. One of the most popular, published in Nature,
is Humanity’ s Last Exam (HLE). It uses 2500 questions drawn from “ the frontier of human
knowledge” to put LLMs through their paces. One, for example, asks how many types of sensory
receptors the human skin contains. “ We wanted a diverse dataset that only experts who have been
working on a field for a long time can answer,” says Long Phan, a research engineer with the HLE’s
developer.
Since the HLE first appeared as a preprint in January 2025, the benchmark has become an important
proving ground for LLMs and HLE scores are now a common talking point for AI companies seeking to
highlight the capabilities of their products. At the HLE’s launch, the leading developer OpenAI’s AI
model won the best score at a mere 8.3%. Earlier this month, Google claimed that its latest reasoning
model for science, called Gemini 3 Deep Think, had achieved a new record HLE score of 48.4%.
But some scientists argue that many of the HLE’s questions test for little-known or even useless
knowledge, rather than an ability to do meaningful research. A Nature editorial accompanying the HLE’s
publication also raised this issue: “We think that more scientists should be asking: What would it take
to develop an AI benchmark that truly measures expert-level thinking?”
1. What does the underlined word “aptitude” in paragraph 2 mean?
A. Knowledge.
B. Performance.
C. Intelligence.
D. Progress.
2. What does Long Phan stress about HLE?
A. Its topic diversity.
B. Experts’ involvement in it.
C. The expertise of its dataset. D. Its data-backed popularity.
3. What is paragraph 4 mainly about?
A. HLE’s role as a key AI test.
B. Companies’ use of HLE.
C. HLE scores of leading AI models.
D. The process of HLE’s launch.
4. By sharing its view, the Nature editorial aimed to ________.
A. back the current testing B. express concern over HLE
C. propose a workable solution D. predict future AI benchmarks
2.（2026·聊城·一模）
In the age of large language models (LLMs) and generative AI, we are witnessing an unprecedented
transformation in how knowledge is produced, spread and consumed.
LLMs, we are told, make us more efficient, simplify complex work, automate boring tasks and allow
us to focus on what matters. But as we feel surprised at their capabilities, a pressing concern emerges:
Are these models genuinely boosting efficiency, or are they eroding our capacity for independent thought,
judgment and critical reflection?
Efficiency is not a neutral term. The current narrative around generative AI treats efficiency as
progress. It suggests that the faster something is done, the better. But faster is not always better. And not
everything that can be automated should be.
The popular belief is that LLMs allow humans to assign repetitive work to machines and reserve
their energy for more reflective tasks, but the opposite is often true. As the more intellectual labor —
writing, summarizing and decision-making, for example — is handed over to AI, the less we will engage
with it ourselves. Instead of reserving our thoughtfulness for higher tasks, we will increasingly lose the
opportunities, and perhaps even the ability, to think critically.
So what do we really mean by “efficiency”? If it means shortening the time it takes to write a
report, perhaps we have succeeded. But if it means replacing the intellectual effort that creates depth,
coherence and reflection, then it’s not a gain; it’s a loss. The moment we accept LLMs as thought
substitutes, rather than thought aids, we begin to worsen the very conditions under which human
reasoning thrives: questioning, dialogue, uncertainty and contradiction.
There is no turning back the presence of LLMs in our lives. But we can choose how to live with
them. The question is not whether they will think for us, but whether we will let them define what it
means to think at all. Efficiency, in the true sense, should not be about doing more with less thought. It
should be about doing better, with deeper attention, stronger ethics and sustained human insight.
5. What does the underlined word “eroding” in paragraph 2 mean?
A. Changing. B. Improving.C. Destroying.
D. Expanding.
6. What do LLMs lead to, according to paragraph 4?
A. We get more reflective labor.
B. We do independent thinking less.
C. We engage in more repetitive tasks.
D. We reduce our work efficiency indeed.
7. What does the author advocate about our using LLMs?
A. Putting efficiency first.
B. Reducing intellectual effort.
C. Achieving more with less time.
D. Increasing human engagement.
8. What is the author’s purpose in writing the text?
A. To describe the fast development of LLMs.
B. To reflect on the negative effects of LLMs.
C. To question the necessity of pursuing efficiency.
D. To challenge the traditional definition of efficiency.
3.（2026·广州·一模）
Survey data shows that most freshmen regularly use generative AI, often treating it as “an intellectual partner”,
Professor John Hampson reported at a faculty (全体教师) meeting in Elite Technology University (ETU). Students
most commonly use it to understand difficult concepts, search, generate study materials, and edit writing.
Interestingly, the lowest reported use is for generating text.
Meanwhile, students are using faculty office hours and the speaking and writing centers less. In last year’s
computer science courses, scores on problem sets increased, yet exam scores declined. “This is concerning,” noted
Hampson. “If they were using AI as a study pal, they weren’t absorbing as much as they might think.”
Students want clearer AI policies, and Hampson advised faculty to carefully consider and share what level of
use they permit, the reasoning behind it, how to cite use of AI, and examples of what’s permissible. He also
encouraged department-wide discussions to best prepare students for a workplace where they will need to know
how to write or code with its assistance. “I also believe that students need to learn to write and code unaided, to
develop critical thinking skills, their agency as citizens, and also meaning — making the ideas that help them
understand their own lives,” he added.
Some professors expressed concerns about how AI use is impacting students’ mental health and learning.
Professor George Wilson noted that students are often highly competitive, and “it’s important to create rules so that
competition leads to healthy behaviors that make them better educated people.” While some suggested more one-
on-one time with students, others noted that budget restrictions would make that difficult.
Professor Poly Burnett observed that lecture attendance is also down. She urged faculty to make lectures
something students genuinely want to attend. She also noted that many teachers are making small changes, in hopes
of continuing teaching as they’ve previously taught. “We actually have to see this less as a problem and more as an
opportunity,” Burnett suggested. “How can ETU lead in rethinking how we teach, how we learn... and have our
students be benefiting and being at the leading edge of that?”
9. What does the author imply about the survey findings by using “interestingly” in paragraph 1?
A. They indicate a promising trend.
B. They contradict a common assumption.
C. They capture the faculty’s interest.
D. They require further investigation.
10. Which of the following changes is mentioned in paragraph 2?
A. Students are interacting more with others.
B. AI use has led to better learning outcomes.
C. Exam scores rose while homework scores fell.
D. Students are using off-line academic services less.
11. Why does Hampson emphasize students writing and coding without AI?
A. To clarify acceptable uses of AI in coursework.
B. To prepare students for future workplace demands.
C. To ensure students develop essential human capacities.
D. To improve students’ long-term academic performance.
12. What is Burnett’s suggestion to the faculty?
A. Make lectures more entertaining.
B. Let students take the leading role.
C. Take the chance to reform education.
D. Adjust teaching slightly to AI challenges.
4.（2026·苏北七市·二模）
Generative AI tools have exploded in popularity, enabling users to create text, images, music and video in
seconds. But behind the innovation lies a controversial issue: the unauthorized use of copyrighted material to train
these models and the uncredited reproduction of protected works in AI-generated content.
For artists, writers and filmmakers, the rise of generative AI feels like a threat. Many AI systems are trained on
vast datasets scraped from the internet—including novels, paintings, songs and films—without permission or
compensation to the original creators. When AI produces new content that closely mimics the style or even specific
elements of copyrighted works, it often does so without attribution, leaving creators feeling their labor is exploited.
Take visual artists as an example. A digital painter might spend years developing a unique style, only to find
AI tools can replicate that style instantly. Some artists have filed lawsuits against AI companies, arguing that
training models on their work without consent violates copyright law. They demand fair compensation and clearer
rules on how AI can use creative content.
Tech companies, however, argue that AI training falls under "fair use," a legal doctrine that allows limited use
of copyrighted material without permission for purposes such as education, research or innovation. They claim AI
transforms the original material into something new, thus not infringing on copyright. Yet this argument fails to
address the core concern of many creators: that AI profits from their work without giving anything back.
The debate is far from settled. Governments around the world are struggling to update copyright laws to keep
pace with AI. The European Union’s AI Act requires transparency about training data, while the US Copyright
Office has refused to grant copyright to purely AI-generated works. In China, new regulations mandate labeling AI-
generated content to prevent deception.
As generative AI continues to evolve, finding a balance between innovation and protection will be critical.
Without clear rules, both creators and AI developers will face uncertainty. The goal should be to foster AI progress
while ensuring that those who create original work are respected and rewarded.
13.What problem is mainly discussed in Paragraph 1?
A. The rapid development of AI technology.
B. The lack of legal protection for AI users.
C. The illegal use of copyrighted material by AI.
D. The difficulty in creating original content.
14.Why do many artists oppose generative AI?
A. AI makes their works less popular.
B. AI copies their styles without permission.
C. AI reduces the value of creative jobs.
D. AI fails to produce high-quality content.
15.What do tech companies claim about AI training?
A. It should be strictly banned by law.
B. It belongs to the category of fair use.
C. It needs full permission from creators.
D. It has nothing to do with copyright.
16.What can be inferred from the last two paragraphs?
A. Global laws on AI copyright are consistent.
B. AI-generated works can get copyright in the US.
C. China requires clear labels for AI content.
D. Innovation should come before copyright protection.
5.（2026·运城·一模）
New research challenges the widespread belief that artificial intelligence (AI) is driving a major rise
in global greenhouse gas emissions Scientists from the University of Waterloo and the Georgia Institute
of Technology analyzed U.S. economic data alongside estimates of how frequently AI tools are used
across different industries. Their aim was to understand what might happen to the environment if AI
adoption increases along its current path.
According to the U.S. Energy Information Administration, 83 percent of the nation’ s economic
activity relies on petrol, coal and natural gas. These fuels release greenhouse gases when burned. The
researchers noted that total energy use from AI in the U.S. matched the electricity consumption of
Iceland, yet this amount remained insignificant when viewed at national or global levels.
“It is important to note that the increase in energy use is not going to be uniform. It’s going to
be felt more in the places where electricity is produced to power the data centers,” said Dr Juan
Moreno-Cruz, a professor at the School of Environment, Enterprise and Development at the University of
Waterloo and Canada Research Chair in Energy Transitions. “If you look at that energy from the local
perspective, that’s a big deal because some places could see double the amount of electricity output and
emissions. But at a larger scale, AI’s use of energy won’t be noticeable.”
“For people who believe that the use of AI will be a major problem for the climate and think we
should avoid it, we’re offering a different perspective,” Moreno-Cruz added. “The effects on climate
are not that significant, and we can use AI to develop green technologies or to improve existing ones.”
To develop their findings, environmental economists Moreno-Cruz and Dr Anthony Harding reviewed
a variety of economic sectors, the types of jobs within those sectors, and the share of tasks that could
potentially be performed by AI. Moreno-Cruz and Harding intend to apply the same approach to
additional countries in order to understand how AI adoption may affect environmental outcomes across
different regions of the world.
17. What is the primary goal of the research?
A. To promote the development of green AI. B. To measure energy consumption worldwide.
C. To warn about AI’s growing energy demands.D. To assess AI’s potential environmental effects.
18. What can be said about AI energy consumption in the U.S.?
A. It contributes to petrol-based activities.
B. It will soon reach the global emission target.
C. It has small influence at the national level.
D. It exceeds Iceland’s electricity consumption.
19 What do researchers plan to do next?
A. Extend their research to more countries.
B. Shift focus to AI’s economic advantages.
C. Develop AI applications to stop emissions. D. Reduce the energy use of AI in data centers.
20. Which of the following is the main idea of the text?
A. AI technology drives greenhouse gas emissions.
B. AI energy consumption urgently needs regulating.
C. Data centers emit more than previously estimated.
D. AI’s impact on climate is much smaller than believed.
6.（2026·郑州·一模）
AI technology has long been able to recognize patterns in music preferences and create personalized
playlists. Now, a new AI system has taken this a step further by analyzing how people listen to music
and identifying their unique “listening styles”. This advancement changes how music streaming services
tailor playlists to individual users, making them more enjoyable.
Music recommendation algorithms (算法) have been highly effective at suggesting new songs and
artists. But Dr. Emily Carter, a music data scientist at the University of Music and Technology, notes that
these algorithms often use a one-size-fits-all approach that doesn’ t record the slight differences of
individual listening behavior. To better understand and satisfy individual preferences, researchers need to
analyze each user’s unique listening patterns.
To develop and train their AI, the researchers collected data from over 50 million listening sessions
and fed it into a neural network. They tested the system by seeing how well it could distinguish between
different users’ listening habits. The system was given 100 listening sessions from each of about 3,000
known users and 100 new sessions from an unknown user. The AI looked for the best match and
identified the unknown user 86% of the time, according to a study presented at the International Society
for Music Information Retrieval (ISMIR).“We were quite surprised by the accuracy,” says Alex Johnson,
a doctoral student in Carter’s lab and the lead author of the study. A non-AI method was only 28%
accurate.
“ The work is innovative,” says Dr. Sarah Kim, a music researcher. “ Personalized music
experiences could transform how we interact with music platforms.”
The researchers are aware of the privacy impact of their system, which could potentially identify
users based on their listening habits. In theory, similar systems could also analyze other behaviors, such
as the types of podcasts ( 播客) people listen to or the timing of their music consumption. ISMIR
organizers found the study impressive but questionable, and accepted it on condition that the researchers
detail the privacy risks. Carter says they have decided, for now, not to release the software publicly.
21. What advancement of AI is mentioned in paragraph 1?
A. Protecting people’s privacy.
B. Recognizing music patterns.
C. Tailoring personalized playlists.
D. Improving music streaming quality.
22. What does Carter say about the music recommendation algorithms?
A. They consider listening styles.
B. They renew networks constantly.
C. They recommend popular songs.
D. They ignore individual preferences.
23. What is the main concern about the new AI system?
A. Its technical weaknesses in analyzing data.
B. Its inability to distinguish between users’ habits.
C. Its limited accuracy compared to non-AI methods.
D. Its potential privacy risk from tracking listening habits.
24. How do ISMIR organizers feel about the new AI system study?
A. Careful.
B. Disappointed.
C. Favorable. D. Uninterested.
7.（2026·西安·3 月）
An AI-powered robot was able to separate a gall bladder (胆藏) from the liver of a dead pig in
what researchers claim is the first realistic surgery by a machine with almost no human intervention.
The robot is powered by a two-tier AI system trained on 17 hours of video containing 16,000
motions performed by human surgeons during operations. When put to work, the first layer of the AI
system watches video, monitors the surgery and issues plain-language instructions, while the second AI
layer turns each instruction into 3D tool motions. In all, the gall bladder surgery requires 17 separate
tasks. The robotic system has performed the operation eight times, achieving 100 percent success in all of
the tasks.
“Current surgical robotic technology has made some procedures less invasive,but risks haven't really
dropped from previous laparoscopic(使用腹腔镜的)surgeries by human surgeons,” says team member Axel
Krieger at Johns Hopkins University in Maryland. “This made us look into what is the next generation
of robotic systems that can help patients and surgeons.” “ The study really highlights the art of the
possibility with Al and surgical robotics,” says Danail Stoyanov at University College London.
“Incredible advances in computer vision for surgical video with the availability of open robotic platforms
for research make it possible to demonstrate surgical automation. "
But many challenges remain to make the system practical in clinical use. “ While the robot
completed the task with 100% success, it had to self-correct six times per case. For example, this could
mean a gripper (夹持器) designed to grasp an artery missed its hold on the first try, " Stoyanov said.
“There were a lot of instances where it had to self-correct, but this was all fully autonomous,”
says Krieger. “It would correctly identify the initial mistake and then fix itself. " The robot also had to
ask a human to change one of its surgical instruments for another, meaning some level of human
intervention was required. The next step,says Krieger, is to let a robot operate autonomously on a live
animal, where breathing and bleeding could complicate things. “ But with continued research, we' re
confident that we can overcome these obstacles step by step. "
25. What are the two-tier tasks that the Al system is trained to perform?
A. Giving instructions and performing motions.
B. Monitoring the surgery and issuing commands.
C. Analyzing video and choosing surgical tools.
D. Imitating human surgeons and separating tasks.
26. What breakthrough does the new robot achieve over traditional laparoscopic surgeries?
A. Minimal invasiveness with no danger.
B. Near autonomy with high success rate.
C. Low risks in complex surgical tasks.
D. Faster self-correction speed in operations.
27. What may prove challenging in a robot operation according to the last paragraph?
A. Adapting to real-time variability.
B. Identifying surgical mistakes quicker.
C. Reducing human help for crucial tasks.
D. Dealing with complicated surgeries.
28. Which of the following best summarizes the passage?
A. The Robotic Surgery: Cutting Medical Risks
B. The Robotic Surgery: Great Clinical Progress
C. The Robotic Surgery: Simplifying Surgery
D. The Robotic Surgery: Success and Ongoing Issue
8.（2026·衡阳·3 月）
Chinese scientists have uncovered the world’s first AI - powered breeding robot named GEAIR. It can cruise
autonomously and carry out cross–pollination ( 异花授粉), promising reduced breeding costs, short breeding
cycles, and improved breeding efficiency.
GEAIR has been built with a combination of two technologies: AI and biotechnology. Xu Cao, a researcher
from the Chinese Academy of Sciences, led the research team that built the robot.
Cross-pollination, also known as hybrid pollination, is the process of transferring pollen (花粉) from a flower
of one plant to another. This process helps in creating hybrid flowers of plants, also known as hybrid breeding.
The aim of hybrid breeding is to develop crop varieties with improved traits, thereby achieving enhanced yield
and quality. However, according to Xu, doing this process repeatedly is time - consuming. GEAIR can help reduce
the time and also avoid human errors.
Living up to its promised potential, the robot carried out a trial in a greenhouse. It identified a flower
accurately and extended its arm gently to complete the hybrid pollination process. The entire breeding process was
done with inch-perfect precision. The researchers also built the first “intelligent robotic breeding factory”, which
can quickly and efficiently develop new, high-quality plant varieties.
GEAIR will start a new era backed by AI and biotechnology in the breeding industry. “Our new study has
initiated an intelligent breeding model of integrated biotechnology, AI and robot labor — marking China’s
successful pioneering efforts in the construction of a closed-loop ( 闭环的) technology system for intelligent
robotized hybrid breeding,” Xu said. “It also shows the application prospects of ‘AI for science’ in the sector of
biological breeding.”
With biotechnology as its foundation, AI as empowerment, and robots as operators, this study could help
China take the lead in the race to create breeding robots that are fully autonomous and intelligent.
29. What is the primary function of the GEAIR robot?
A. To take care of human gardeners.
B. To monitor plant growth conditions.
C. To conduct hybrid pollination tasks.
D. To harvest mature crops automatically.
30. What problem of traditional hybrid breeding does GEAIR solve?
A. Lack of pollen sources.
B. Long time and mistakes.
C. High costs of hybridization.
D. A narrow range of hybrid types.
31. What can we infer about the “intelligent robotic breeding factory”?
A. It is popular worldwide now.
B. It can work without any power.
C. It mainly focuses on common crops.
D. It can enhance the diversity of agriculture.
32. What is the significance of GEAIR’s development?
A. It makes organic farming possible.
B. It lowers the cost of traditional farming.
C. It proves robots can work better than humans.
D. It shows China’s leadership in agricultural technology.
9.（2026·河南·一模）
For years, the dream future kitchen looked like something from a sci-fi film: robots turning burgers,
mechanical arms moving wildly. But at CES (International Consumer Electronics Show) 2026, industry
experts painted a different prospect. The future isn’t arriving with robots looking like us. It’s arriving
quietly, invisibly, and it’s already here.
Early smart kitchen products made a critical mistake. As Nicole Papantoniou from the Good
Housekeeping Institute put it, “A lot of people were putting smart features, which you didn’t really
need, into products.” Today’s successful ideas aren’t about adding technology for its own purpose.
They’re about friction reduction — making cooking easier without the user even noticing the intelligence
at work.
This shift is clear in the latest AI appliances. Several brands offer ovens (烤箱) with systems that
“see” what you put inside. Simply place the food in, and the machine automatically selects the best
cooking option. No buttons, no guesswork. Refrigerators are changing in a similar way. The latest AI
models have cameras that identify ingredients, track best-before dates, and suggest recipes based on what
you have. A partnership with chef Jamie Oliver brings AI-made recipes tailored to your needs. But
perhaps the most unexpected use of AI in the kitchen has nothing to do with cooking. Companies are
developing smart range hoods (抽油烟机) that use airflow to create a low-pressure zone above the pan,
trapping very small particles (颗粒) before they reach your lungs.
So will robots replace human cooks? At a CES Discussion, chef Tyler Florence gave a firm answer.
“Human-made will become the new luxury item,” he said, “Machines excel at repetitive, boring tasks.
But creativity, the human touch — these will only become precious as technology advances.”
The vision from CES 2026 is not a kitchen without cooks. It’ s a kitchen where invisible
intelligence handles the heavy work, and humans are freed to turn ingredients into meals, and meals into
memories.
33. What is the big change of today’s smart kitchen ideas?
A. Creating more robot lookalikes.
B. Reducing trouble while cooking.
C. Designing more sci-fi products.
D. Adding more complex functions.
34. How do new AI ovens simplify the cooking process?
A. They recognize food and set the right mode.
B. They bring AI-made recipes tailored to needs.
C. They suggest recipes based on what you have.
D. They use airflow to create a low-pressure zone.
35. What can be inferred from Tyler Florence’s words?
A. Human creativity will be highly valuable.
B. AI will take the place of human creativity.
C. Human-made food is more than expensive.
D. Machines are better at innovative cooking.
36. What can be a suitable title for the text?
A. AI in Kitchens: A Smart Master for Cooking
B. Smart Kitchens: More Robotic, Less Human
C. CES 2026: When Kitchens Finally Go Sci-Fi
D. Hidden AI: The New Face of Future Kitchens
10.（2026·呼和浩特·一模）
Around Christmas 50-year-old New Yorker Holly Jespersen felt unwell but hesitated to see a doctor.
She turned to ChatGPT, which advised her against visiting. Days later, with a high fever and headaches,
again using the chatbot to decide when, she finally went to urgent care and was diagnosed with influenza
A.
Holly is far from alone. According to OpenAI, over 40 million daily health-related enquiries, with
230 million weekly. In January, it announced ChatGPT Health, allowing users to upload medical records
for customized (定制的) support. The company stresses it is meant to “support, not replace” medical
care, not for diagnosis or treatment, but to help with everyday questions and pattern recognition.
Yet concerns arise. Family physician Dr. Alexa Mieses Malchuk warns that ChatGPT, like WebMD,
prioritizes being helpful over accurate. A 2023 study found ChatGPT’s cancer treatment plans contained
many errors, some hard even for experts to detect. However, newer research on colon cancer showed its
answers on symptoms and prevention were highly accurate, suggesting LLMs (大型语言模型) may assist
patient education but not clinical decisions.
Beyond accuracy, psychologists highlight anxiety risks. A 2013 study confirmed that online symptom
searches can intensify health anxiety, especially for those intolerant of uncertainty. Clinical psychologist
Elizabeth Sadock notes that ChatGPT, always available and affirming, fuels reassurance-seeking (寻求慰藉)
behavior, trapping users in a cycle of anxiety. For some patients, limiting ChatGPT use may now be part
of treatment.
Privacy is another puzzle. Biomedical informatics professor Bradley Malin acknowledges OpenAI’s
security efforts, but stresses ChatGPT Health falls outside HIPAA regulation. Patients may unknowingly
lose legal protections when their data flows from secured medical records to an unregulated third party.
Yet some see value. Dermatologist Kumar views ChatGPT Health as educational, clarifying terms like
sunscreen types, not diagnostic. He distinguishes it from WebMD’ s curated, reviewed content, while
ChatGPT’s AI may mislead.
Thus, ChatGPT Health enters America’s broken system as a double-edged sword: a round-the-clock
assistant that may empower (赋权) patients, yet risks misinforming, over-reassuring, and exposing them to
unregulated data practices.
37. Why does OpenAI launch ChatGPT Health?
A. To replace medical care totally.
B. To provide consultation timely.
C. To treat the patients early.
D. To diagnose diseases quickly.
38. What can we learn from paragraphs 3-5?
A. ChatGPT may lead to more risks than benefits.
B. ChatGPT is always available, helpful and accurate.
C. Psychologists advise people not to use ChatGPT.
D. People will have no privacy when using ChatGPT.
39. How does Kumar find ChatGPT?
A. It teaches patients some medical terms.
B. It can be used as an assistant to patients.
C. It can help more patients cure diseases.
D. It has more advantages than disadvantages.
40. What is the author’s attitude toward ChatGPT Health?
A. Enthusiastic and supportive. B. Cautious and optimistic.
C. Disapproving and negative.
D. Critical and loyal.
11.（2026·江西赣南·一模）
During a golden sunset, Sharon Wilson pointed a thermal-imaging (热成像) camera at a flagship data
centre, revealing the enormous heat its AI supercomputer had been releasing into the sky. Meanwhile, the
facility’s core product, like many other AI chatbots, kept generating floods of false or harmful content
for users worldwide. “It’s a horrible waste,” said Wilson, director of the campaign group Oilfield
Witness.
Wilson is not alone in having this concern. Scientists are watching the AI expansion with unease as
it pollutes the natural world with carbon and the digital world with dangers ranging from misinformation
to poisonous videos.
Data centres currently consume about 1% of global electricity, but that share may jump soon. Their
slice of power is projected to hit 8.6% by 2035, while the International Energy Agency (IEA) expects
data centres to account for at least a fifth of electricity-demand growth to the end of the decade.
What if AI could pay off its energy debts by saving carbon elsewhere? That idea was put forward
in an IEA report, which argued that AI applications could cut emissions (排放) by far more than data
centres produce. A research paper reached a similar conclusion after modelling cases in which AI would
help integrate solar and wind into power networks, improve battery chemistry in electric cars, and
encourage consumers to make climate-friendly choices.
The projected carbon savings carry large uncertainties-greater efficiency can lead to greater use, the
IEA warns, and rebound effects may undercut the gains, such as self-driving cars undermining public
transport. But other sectors are so polluting, the researchers say, AI would need to cut their emissions by
only a small percentage to cover its own carbon cost.
Ultimately, given the massive energy consumed by algorithms ( 算法), it is essential that AI be
employed to “do good in terms of fighting the climate crisis-designing the next generation of batteries,
tracking deforestation,” as Sasha Luccioni, climate lead at an AI firm, said, rather than “create social-
media websites filled with rubbish while data centres are still powered by coal-fired generators.”
41. What does the underlined words “this concern” in paragraph 2 refer to?
A. The shortage of AI service. B. The unreliability of AI output.
C. The release of heat by AI centers.
D. The misuse of energy by AI systems.
42. What do the IEA report and the research paper in paragraph 4 agree on?
A. AI can be a net carbon saver.
B. AI can be energy-efficient.
C. AI can provide computing power.D. AI can direct electricity distribution.
43 What is the purpose of paragraph 5?
A. To put forward an opposite position. B. To offer a more comprehensive view.
C. To add some background information D. To demonstrate the previous argument.
44. What does Sasha Luccioni argue about AI?
A. Its design calls for improvement. B. Its energy use demands restriction.
C. Its application requires wise guidance. D. Its development deserves public support.
12.（2026·天津·统考）
The question of whether artificial intelligence (AI) will take away our jobs is on many people’s minds today.
Current applications, from AI robotics performing complex surgeries to large language models like ChatGPT
writing academic essays and solving tough problems, have not only demonstrated remarkable capabilities but also
sparked significant moral concerns.
Broadly speaking, public opinion is divided. Some view AI as the ultimate tool for solving society’s most
pressing challenges, from disease to climate change. Others, however, fear that AI will overtake human
intelligence. Both views rest on a common assumption that AI possesses, or will possess, a superior form of
intelligence that could replace human decision-making. But given the fact that technology is the product of human
civilization, the challenge from AI is something we have created for ourselves as we keep pushing our own
boundaries. In other words, AI’s progress, functions and future direction are all directed by the human mind.
Therefore, before AI evolves into a potential threat, the global community must reach an agreement on the role
it is to play. More importantly, related laws and regulations must ensure that AI will benefit society and prevent it
from threatening human life. For instance, while future robots might develop a form of emotional intelligence,
enabling them to recognize, understand and express emotions in a way that is similar to humans, we must establish
clear boundaries to prevent AI copying human emotions. Without legal restrictions, AI may become a social
disaster.
The new industrial revolution, driven by AI, is an unstoppable force. This change, much like the steam and
internet revolutions that brought once-unimaginable shifts, will definitely reshape the world of work, meaning
some jobs will disappear. Yet, history repeatedly shows that humanity possesses a great capacity for adaptation.
Following each technological leap, new forms of work have emerged, often more creative and fulfilling than the
previous ones. Consequently, it’s unnecessary to worry AI will replace our jobs. While technology advances at a
rapid pace, what we need to do is to welcome the AI era rather than resisting its progress for fear of the unknown.
45. Why does the author provide examples of AI applications in Paragraph 1?
A. To compare the functions of different AIs.
B. To explain the principles of deep learning.
C. To show evidence for worries about AI.
D. To predict breakthroughs in medical fields.
46. What does the author imply about AI’s progress?
A. It will be too complex to control.
B. It depends on human innovation.
C. It will overtake human intelligence.
D. It helps human break boundaries.
47. How can we prevent AI’s potential threat?
A. By preventing it threatening humans.
B. By stopping it expressing emotions.
C. By changing global agreements.
D. By setting clear rules and laws.
48. What does the writer suggest readers do with the coming of the AI era?
A. Deal with it positively.
B. Accept it passively.
C. Respond to it randomly.
D. Defend it unconditionally.
49. Where is the passage most probably taken from?
A. A newspaper column on science.
B. A textbook on computer science.
C. An advertisement for AI software.
D. A research paper on AI development.
13.（2026·深圳·一模）
People around the globe have suffered the anxiety of waiting months to find out if their homes have
been damaged by wildfires. Now, once the smoke has cleared for aerial photography, researchers have
found a way to identify building damage within minutes. Through a system called DamageMap, a team at
Stanford University has brought an AI approach to building assessment: Instead of comparing before-and-
after photos, they’ve trained a program using machine learning to rely only on post-fire images.
The current method of assessing damage involves people going door-to-door to check every building.
While DamageMap is not intended to replace in-person damage assessment, it could be used as a
supplementary tool by offering immediate results and providing the exact locations of the affected
buildings. The researchers tested it using a variety of satellite and aerial photography with at least 92
percent accuracy.
Most computational systems now cannot efficiently classify building damage because the AI compares
post-disaster photos with pre-disaster images that must use the same satellite, camera angle and lighting
conditions, which can be expensive to obtain or unavailable. Therefore, DamageMap first uses pre-fire
photos to map the area and confirm building locations. Then, the program analyzes post-wildfire images
to identify damage through features like blackened surfaces, collapsed roofs or the absence of structures.
Structural damage from wildfires in California is typically divided into four categories: almost no
damage minor damage, major damage or destroyed. Because DamageMap is based on aerial images, the
researchers quickly realized the system could not make such detailed assessments and trained the machine
to simply determine if there was a fire damage or not.
Because the team used a deep learning technique, their model can continue to be improved by
feeding it more data. The researchers said the tool can be applied to any area suffering from wildfires
and hope it could also be trained to classify damages from other disasters, such as floods or hurricanes.
“So far our results suggest that this can be generalized, and we can keep improving it,” said lead
study author Marios Galanis, a graduate student at Stanford’s School of Engineering.
50. What is the advantage of using DamageMap?
A. It helps improve the evaluation efficiency. B. It operates automatically after self-learning.
C. It analyzes large numbers of disaster photos.
D. It takes the place of the traditional measures.
51. How does DamageMap work?
A. It identifies damage with pre-fire photos.
B. It confirms locations with post-fire photos.
C. It assesses damage through the features of buildings.
D. It maps the fire-affected area through comparing photos.
52. What would the future study focus on according to Marios Galanis?
A. Accuracy improvement. B. A wider range of application.
C. Techniques development.D. A higher speed of machine learning.
53. What does the text mainly talk about?
A. The impact of wildfires on local residents.B. Main challenges to classify structural damage.
C. Possible solutions to identify natural disasters. D. An AI system for rapid fire damage evaluation.
14.（2026·襄阳·一模）
John Hester, a retired software developer, who lives in Southern California, asked Grok 3, a large
language model, to write him code (代码) for a game that he could play on his computer last February.
Some two hours later, he had “a playable, functional game.” “It’s so amazing,” he says.
Rather than being programmed to search through a set of options, generative AI models learn from a
huge number of examples. Some video games now use generative AI. You can try a demo (演示版游戏)
called Oasis, like an AI-generated version of Minecraft. In the real game Minecraft, a map and rules
govern everything around you. Not here. Oasis, which was released in 2024, is based on a new type of
generative AI called a world model. Whatever is on the screen now feeds into the AI world model. It
predicts what you will see next based on what you’re seeing now and builds virtual environments you
can move through on the spot. Millions of hours of Minecraft gameplay videos went into training the
world model behind Oasis.
Cook, a researcher and game designer, sees some drawbacks to using generative AI to create all or
parts of video games. With generative AI, typically only big companies get to make decisions about how
the models work. Besides, using generative AI or world models to make lots of automated game content
“might lead to more boring stuff being made,” cautions Cook. A person’s creative work reflects their
experience of living in the world. And today’s generative AI can only copy what people have already
created.
Tessa Kaur, editor at The Gamer magazine, writes that AI-generated dialogue doesn’ t produce
fascinating characters. AI “simply cannot be creative enough,” she writes. When you care about game
characters, it’s “because someone took the time to craft that dialogue for you, with many rewrites and
deep thought.”
54. Why was John Hester impressed?
A. Grok 3 taught him game coding. B. He developed a new piece of software.
C. Grok 3 coded a game for him quickly.
D. He updated his computer successfully.
55. What can Oasis provide for game players?
A. Pre-programmed game scenes.
B. AI-generated virtual environments.
C. Personalized game maps and rules.
D. Numerous world model training data.
56. What do Cook and Tessa both agree with?
A. AI crafts fascinating dialogue.
B. Boring characters need AI polish.
C. Humans create vivid game content.
D. Big firms control AI game design.
57. What can be a suitable title for the text?
A. AI, Create Awesome Video Games?
B. AI, Train World Game Models?
C. Grok 3, Generate Vivid Game Roles
D. Grok 3, Beat the Original Game
15.（2026·日照·一模）
A new study by researchers at the Cluster of Excellence Science of Intelligence shows that a
combination of uncertainty and heterogeneity ( 异质性) plays a crucial role in how groups reach
agreement.
Classic models of decision-making assume that all individuals contribute equally to consensus (共识),
but in reality, groups are diverse and homogeneous in both knowledge and influence. Just as some people
are experts in a topic, some individuals have more accurate or reliable information than the rest of the
group. Others might be more “connected,” which causes their opinions to spread more widely.
These two types of diversity, namely level of knowledge and number of connections, are not
independent, as uncertainty influences how the two shape decision-making. In other words, individuals
with more initial knowledge tend to become more central and influential, helping others reduce
uncertainty, while those who interact with many others obtain more information and thus become less
uncertain over time. This dynamic allows groups to naturally remove weak or biased information and
come to reliable conclusions — as long as central individuals don’t become overconfident too quickly.
To explore these effects, the researchers built a model where individuals adjust their beliefs and
certainty dynamically as new information comes in. Uncertain individuals relied more on their peers, while
confident ones shaped the group’s direction of opinion. But position within the network mattered just as
much — highly connected agents spread their opinions widely, whether they were right or wrong.
The researchers found that a mix of perspectives wasn’ t enough to improve decisions. Groups
reached smarter and faster decisions when guided by uncertainty. When everyone had equal certainty and
connections, consensus was slow and unreliable. But in heterogeneous groups, uncertainty helped weigh
opinions, so that decisions were faster and more accurate.
In artificial intelligence and robotics, this research offers a new way to design systems that make
better collective decisions. Self-driving cars could assess not just sensor inputs, but also the confidence of
other nearby vehicles, improving safety. Many natural systems already follow the principle of adapting to
uncertainty. Schools of fish, flocks of birds, and ant colonies don’t treat all input equally but adapt
dynamically. We can use that knowledge to build better AI and improve human collaboration.
58. What do classic models of decision-making ignore?
A. Group discussion.
B. Individual difference.
C. Equal contribution. D. Interpersonal relationship.
59. What can be inferred about “knowledge” and “connections”?
A. They can be misleading.B. They can remove overconfidence.
C. They rely on central individuals. D. They interact through uncertainty.
60. How can uncertainty assist with decision-making according to the research?
A. By balancing different views.
B. By encouraging more participation.
C. By making people decisive. D. By reducing unnecessary conflicts.
61. What does the author mainly discuss in the last paragraph?
A. Choice of new research methods. B. Possible directions of AI technology.
C. Ways of adapting to uncertainty. D. Potential application of the findings.
