1 . A new study examined scientists’ peer reviews, or researchers’ official statements on others’ work, across multiple AI-related conferences. At one such conference, those peer reviews used the word “meticulous” — a buzzword often associated with generative A.I., like ChatGPT — almost 3,400 percent more than the previous year. Other major conferences showed similar patterns. In other words, many researchers were handing, at least, parts of their peer review over to A.I.
What’s going on in science is a slice of a much bigger problem. Any viral post on social media now almost certainly includes A.I.-generated elements. There are synthetic videos for children on YouTube, like music videos about parrots where the birds have eyes within eyes, singing in an unnatural voice. The narratives make no sense, and characters appear and disappear randomly.
As a neuroscientist, this worries me. Isn’t it possible that human culture contains within it cognitive micronutrients — things like reasonable sentences, narrations and character continuity — that developing brains need? Einstein supposedly said: “If you want your children to be intelligent, read them fairy tales. If you want them to be very intelligent, read them more fairy tales.” But what happens when a child is consuming mostly A.I.-generated waste? We find ourselves in the middle of a vast developmental experiment.
A.I.’s cultural pollution is driven by a desire to fill the Internet’s appetite for content as cheaply as possible, which in turn pollutes our culture. And despite public appeals to act against it, A.I. companies are dragging their feet because it goes against the industry’s bottom line to have detectable products, which they fear might weaken the model’s performance, although there is no current evidence.
To deal with this general refusal to act, we need a Clean Internet Act. Perhaps the simplest solution would be to force built-in watermarking to A.I. generated outputs, like patterns not easily removable. Just as the 20th century required action to protect the shared environment, the 21st century is going to require actions to protect a different but equally critical resource: our shared human culture.
1. What can we learn from the first two paragraphs?| A．Children should avoid social media. |
| B．A.I. writings are unreliable. |
| C．Synthetic elements make no sense. |
| D．A.I. pollution is widespread. |
| A．To promote experiments on reading. |
| B．To connect intelligence and fairy tales. |
| C．To show concerns over cognitive input quality. |
| D．To contrast stories with A.I. generated content. |
| A． Increasing the expense. |
| B．Adding permanent labels. |
| C．Creating various patterns. |
| D．Building more A.I. models. |
| A．A.I. Garbage Is Polluting our Culture |
| B．A.I. Companies Change Tech Industry |
| C．A.I. Performance Continues to Decline |
| D．A.I. Products Impact Scientific Research |
2 . In recent news, it has been reported that Hollywood executives have decided to cancel an $800 million investment in a new studio. This decision comes after the emergence of a groundbreaking technology called Sora, which has sent shockwaves throughout the film industry.
Sora, developed by OpenAI, has quickly become a focal point of discussion within Hollywood. Unlike previous AI video generators, Sora has the ability to generate one-minute-long videos based on textual prompts (提示) while maintaining visual quality and consistency. It can switch between shots, adjust compositions and accurately combine video content with relevant background themes, resulting in incredibly realistic and virtual videos.
The introduction of Sora is just the beginning of a shocking shift in the film industry. Its ability to generate high-quality videos based on textual prompts raises questions about the future of traditional movie-making processes. The potential to replace certain job roles and disturb the livelihoods of some individuals is just a small part of the overall impact.
Looking ahead, the innovation of productivity tools and the potential for smaller teams to create Hollywood-level movies at a lower cost may completely transform the way films are produced. This shift could bring about new production methods, viewing experiences, and even business models within the film and entertainment industry.
This breakthrough technology has raised concerns within the film industry, as it has the potential to disturb the traditional movie production process and cause certain job roles’ losses. However, this technology also presents opportunities for innovation and cost-effective production methods. The film industry is on the edge of a transformative era, and only time will tell how Sora and similar technologies will shape its future. It likely won’t be long before audiences get to see film productions made with Sora. It can be stated with certainty that the film industry has entered a new era of rapid development in AI technology.
1. What is paragraph 2 mainly about?| A．The impact of Sora. | B．The concern of Sora. |
| C．The potential of Sora. | D．The advantage of Sora. |
| A．Textual creation. | B．Individual investment. |
| C．Relevant job chances. | D．Video content development. |
| A．Culture and business models. | B．Film production cost and methods. |
| C．Function and viewing experiences. | D．Innovation and background themes. |
| A．Objective. | B．Dismissive. | C．Unclear. | D．Doubtful. |
3 . Building artificial intelligences that sleep and dream can lead to more dependable models, according to researchers who aim to mimic (模仿) the behavior of the human brain.
Concetto Spampinato and his research members at the University of Catania, Italy, were looking for ways to avoid a phenomenon known as “disastrous forgetting”, where an AI model trained to do a new task loses the ability to carry out jobs it previously excelled at. For instance, a model trained to identify animals could learn to spot different fish species, but then might lose its ability to recognize birds. They developed a method of training AI called Wake-Sleep Consolidated Learning (WSCL), which mimics the way that our brains reorganize short-term memories of daily learning when we are asleep.
Besides the usual training for the “awake” phase, models using WSCL are programmed to have periods of “sleep”, where they analyze awake data from earlier lessons. This is similar to human spotting connections and patterns while sleeping.
WSCL also has a period of “dreaming”, which involves novel data made from combining previous concepts. This helps to integrate previous paths of digital “neurons (神经元)”, freeing up space for future concepts. It also prepares unused neurons with patterns that will help them pick up new lessons more easily.
The researchers tested three AI models using a traditional training method, followed by WSCL training. Then they compared performances for image identification. The sleep-trained models were 2 to 12 percent more likely to correctly identify the contents of an image. They also measured an increase in how much old knowledge a model uses to learn a new task.
Despite the results, Andrew Rogoyski at the University of Surrey, UK, says using the human brain as a blueprint isn’t necessarily the best way to boost AI performance. Instead, he suggests mimicking dolphins, which can “sleep” with one part of the brain while another part remains active. After all, an AI that requires hours of sleep isn’t ideal for commercial applications.
1. WSCL was developed to help improve AI’s ______.| A．reliability | B．creativity | C．security | D．popularity |
| A．Generate new data. | B．Process previous data. |
| C．Receive data for later analysis. | D．Save data for the “awake” phase. |
| A．The application of WSCL. | B．The benefits of AI research. |
| C．The findings of the research. | D．The underlying logic of WSCL. |
| A．Cautious. | B．Prejudiced. | C．Pessimistic. | D．Unconcerned. |
4 . Our species’ incredible capacity to quickly acquire words from 300 by age 2 to over 1, 000 by age 4 isn’t fully understood. Some cognitive scientists and linguists have theorized that people are born with built-in expectations and logical constraints (约束) that make this possible. Now, however, machine-learning research is showing that preprogrammed assumptions aren’t necessary to swiftly pick up word meanings from minimal data.
A team of scientists has successfully trained a basic artificial intelligence model to match images to words using just 61 hours of naturalistic footage (镜头) and sound-previously collected from a child named Sam in 2013 and 2014. Although it’s a small slice of a child’s life, it was apparently enough to prompt the AI to figure out what certain words mean.
The findings suggest that language acquisition could be simpler than previously thought. Maybe children “don’t need a custom-built, high-class language-specific mechanism” to efficiently grasp word meanings, says Jessica Sullivan, an associate professor of psychology at Skidmore College. “This is a really beautiful study, ” she says, because it offers evidence that simple information from a child’s worldview is rich enough to kick-start pattern recognition and word comprehension.
The new study also demonstrates that it’s possible for machines to learn similarly to the way that humans do. Large language models are trained on enormous amounts of data that can include billions and sometimes trillions of word combinations. Humans get by on orders of magnitude less information, says the paper’s lead author Wai Keen Vong. With the right type of data, that gap between machine and human learning could narrow dramatically.
Yet additional study is necessary in certain aspects of the new research. For one, the scientists acknowledge that their findings don’t prove how children acquire words. Moreover, the study only focused on recognizing the words for physical objects.
Still, it’s a step toward a deeper understanding of our own mind, which can ultimately help us improve human education, says Eva Portelance, a computational linguistics researcher. She notes that AI research can also bring clarity to long-unanswered questions about ourselves. “We can use these models in a good way, to benefit science and society, ” Portelance adds.
1. What is a significant finding of machine-learning research?| A．Vocabulary increases gradually with age. |
| B．Vocabulary can be acquired from minimal data. |
| C．Language acquisition is tied to built-in expectations. |
| D．Language acquisition is as complex as formerly assumed. |
| A．Facilitate. | B．Persuade. | C．Advise. | D．Expect. |
| A．Its limitations. | B．Its strengths. | C．Its uniqueness. | D．Its process. |
| A．Doubtful. | B．Cautious. | C．Dismissive. | D．Positive. |
5 . China launched a Long March 2F carrier rocket on December 14,2023 from the Jiuquan Satellite Launch Center in Northwest China’s Gobi Desert, sending a reusable experimental spacecraft into orbit. The test vehicle is scheduled to stay in orbit for a certain period of time and then return to its preset landing site in China. During the orbital flight, it has been tasked with verifying (核实) reusable technologies and space science experiments, which will both be used as technical support for the peaceful use of space.
It did not reveal the details of the mission and the spacecraft, such as launch time and specific plans, or publish pictures of the rocket’s liftoff or scenes inside the ground control hall. The mission is the third that China has made public related to reusable experimental spacecraft.
The country’s first orbital test of a trial vehicle took place in September 2020, and the craft was in orbit for just under two days. The second test started in August 2022 and the spaceplane stayed in the Earth’s orbit for 276 days before landing in May 2023．The second test’s success marked a major breakthrough in China’s reusable spacecraft technology, which is aimed at providing a convenient and affordable way to travel between Earth and outer space, the Jiuquan center said after the experimental spaceplane’s landing.
In recent years, advances in science and technology have reignited (重新激起) the space industry’s enthusiasm for reusable spaceships, especially robotic spaceplanes such as the Boeing X-37B that are smaller, cheaper, and less complex in their design, production and operation.
According to experts, reusable spacecraft will have a wide range of applications, including space tours for civilians, transporting astronauts, resupplying space stations, and placing satellites into orbit at a lower cost than through traditional methods.
1. What is revealed about the third test spacecraft?| A．The tasks it performed. | B．The details of its mission. |
| C．The pictures taken of its launch. | D．The scenes in the ground control hall. |
| A．To make outer space travel more available. |
| B．To mark a major breakthrough in flight speed. |
| C．To have the spaceplane stay in the orbit for 276 days. |
| D．To get the spacecraft to return to its preset landing site. |
| A．The reusable products. | B．The potential applications. |
| C．The experimental spacecraft cost. | D．The individual estimate. |
| A．Great Progress in Space Exploration |
| B．Details of the Third Spaceplane Mission |
| C．A Wide Range of Applications of Reusable Spacecraft |
| D．Reusable Spacecraft Launched to Orbit on Experimental Mission |
6 . A new study reveals that pigeons (鸽子) can tackle some problems just like artificial intelligence, enabling them to solve difficult tasks that might challenge humans. Previous research has theorized that pigeons employ a problem-solving strategy, involving a trial-and- error approach, which is similar to the approach used in AI models but differs from humans’ reliance on selective attention and rule use. To examine it, Brandon Turner, a psychology professor at the Ohio State University, and his colleagues conducted the new study.
In the study, the pigeons were presented with various visual images, including lines of different widths and angles, and different types of rings. The pigeons had to peck (啄) a button on the right or left to indicate the category to which the image belonged. If they got it correct, they received food; if they were wrong, they received nothing. Results showed that, through trial and error, the pigeons improved their accuracy in categorization tasks, increasing their correct choices from about 55% to 95%.
Researchers believed pigeons used associative learning, which is linking two phenomena with each other. For example, it is easy to understand the link. between “water” and “wet”. “Associative learning is frequently assumed to be far too primitive to. explain complex visual categorization like what we saw the pigeons do,” Turner said. But that’s exactly what the researchers found.
The researchers’ AI model tackled the same tasks using just the two simple mechanisms that pigeons were assumed to use: associative learning and error correction. And, like the pigeons, the AI model learned to make the right predictions to significantly increase the number of correct answers. For humans, the challenge when given tasks like those given to pigeons is that they would try to come up with rules that could make the task easier. But in this case, there were no rules, which upsets humans.
What’s interesting, though, is that pigeons use this method of learning that is very similar to AI designed by humans, Turner said. “We celebrate how smart we are that we designed artificial intelligence: at the same time, we regard pigeons as not clever animals,” he said.
1. What is the purpose of the new study?| A．To test a theory. | B．To evaluate a model. |
| C．To employ a strategy. | D．To involve an approach. |
| A．Draw circles. | B．Correct errors. | C．Copy gestures. | D．Identify images. |
| A．They are of equal intelligence. |
| B．They are good at making rules. |
| C．They respond rapidly to orders from humans. |
| D．They employ simple ways to get things done. |
| A．Pigeons’ trial-and-error method is revealed |
| B．Pigeons outperform humans in tough tasks |
| C．“Not smart” pigeons may be as smart as AI |
| D．AI models after pigeons’ learning approach |
7 . How Can We Trust AI If We Don’t Know How It Works
Alien minds, in the form of artificial intelligence (AI) systems, surround us in daily life, powering facial recognition, determining creditworthiness, and even creating poetry and code.
Many AI systems, particularly those based on deep learning neural networks, have trillions of parameters (参数), making their decision-making processes opaque (不透明的) — a phenomenon known as the AI explainability problem.
Unlike humans who can justify their decisions based on ethical principles, AI lacks the ability to explain its decision-making, failing the predictive requirement for trust. Consider situations like the “Trolley Problem,” where a self-driving car must make a difficult decision without the capacity to provide a human-like explanation.
Trust relies not only on explainability and predictability but also on ethical or moral motivations. Humans expect others to act not just as they assume they will, but also as they should.
The integration of AI into vital systems, such as electric grids and military operations, heightens the importance of trust.
In conclusion, AI lacks the predictability and ethical elements that would make it inherently trustworthy. Addressing the challenges of explainability and alignment is essential to ensure that AI systems earn and maintain human trust in critical applications. Further research in this area is needed to pave the way for trustworthy AI systems in the future.
| A．As a result, understanding why AI systems make specific decisions is challenging, raising concerns about their trustworthiness. |
| B．However, AI operates with a fixed way of thinking molded by training data. |
| C．Addressing and resolving the issue before it reaches a critical point becomes crucial. |
| D．These are contexts where undesirable behavior could lead to severe consequences. |
| E．To enhance trust, proposals have been made to involve humans in AI decision-making, either in the loop or on the loop. |
| F．Due to the lack of unpredictability, people find it impossible to interpret the specific factors influencing Al-driven decisions. |
| G．However, the trustworthiness of these AI systems faces significant challenges due to their limited explainability and unpredictability. |
8 . Artists everywhere are getting “understandably nervous” about recent advances in artificial intelligence. Last month, a winner of an art prize at the Colorado State Fair “sparked a violent protest” when he posted the news and explained that he’d created his image using an AI program. Critics quickly accused 39-year-old Lance Allen of cheating. To be fair, Allen had won in the digital art category and made no secret of how the image had been produced. But the rules of art making are clearly changing.
Allen’s creative process, to be clear, “was not a push-button operation, ”said Jason Blain in Forbes. He claims to have spent 80 hours on his entry, first on fine-tuning his text prompts (提示), then by touching up the final image using Photoshop and similar tools, then arranging to print the image on canvas. He made the finished product using AI much as a photographer creates an image using a camera.
But Allen, a tabletop game developer, is awed by AI’s capabilities and urges artists and illustrators to welcome the technology rather than fight it. “Art is dead,” he says. “AI won. Humans lost.” A more inspiring lesson to take from his victory, though, is that image generators are likely to “expand the appreciation for and creation of art” by opening the field to people, like him, who could never draw anything as detailed as his award-winning image. “If anything, we will have more artists,” and as the technology progresses, “we might see the emergence of art styles that none have seen before.”
You can’t blame traditional artists if they’re unhappy. Image generators work their magic, after all, by analyzing the aesthetics (美学) of millions of pre-existing images. One of the most complicated image generators “makes crystal clear just how destructive this technology will be,” said Loz Eliot in New Atlas. Given a specific prompt, it can produce an image of just about anything you can imagine and even follow the style of a favorite artist’s work. Its arrival marks “an incredible popularization of visual creativity” while aiming “a knife to the heart of anyone who’s spent decades improving their artistic techniques hoping to make a living from them.”
1. Why are artists getting nervous about AI recently?| A．A winner of an art prize used AI. | B．Lance Allen cheated in the art competition. |
| C．The digital art will soon dominate. | D．There will be great changes in art creation. |
| A．It was no easy work for Allen even with Al. | B．Allen worked as a photographer creating an image. |
| C．AI played a key role in Allen’s art creation. | D．Although with AI, Allen’s creation counted a lot. |
| A．Human has been beaten by AI. | B．AI will make art more popular. |
| C．Greater artists and new art styles will appear. | D．AI enables amateurs to win art competitions. |
| A．It works by analyzing images created by human. |
| B．It can produce images beyond people’s imagination. |
| C．It makes artists’ long-time effort meaningless. |
| D．It makes it impossible for artists to make a living. |
The Tianzhou 7 robotic cargo ship is scheduled
Tianzhou 7 will become the
Pang Zhihao, an expert on space exploration technology said that the Tianzhou 7
The Long March 7 model is a liquid-fueled rocket
10 . Sonia Kleindorfer was a new director of the Konrad Lorenz Research Center in Austria. At her institute, Konrad Lorenz was a famous Austrian zoologist who could correctly name each kind of goose, which made Kleindorfer feel a certain amount of pressure. “I can do five, but when the next five come, I start to have a mental meltdown,” she says. So she contacted a more technically-minded colleague and asked him: Could he write a program to distinguish these faces?
He said, yes, but he’d need a database of geese photos to work with. Kleindorfer got her team out there, snapping pictures of the geese from every angle. After building the database, they wrote a piece of facial recognition AI that could ID a goose, by looking at specific features of its beak (喙). After a couple of years, the team reports that their goose recognition software is now about 97% accurate.
“Geese have such drama—there are archrivals (劲敌), and jealousy and retribution (报答),” Kleindorfer says. To find out how faces figured into this drama, she presented the geese with full-sized pictures of themselves, their partners, or another member of the flock. She showed evidence that geese seemed to recognize photos of their partners and friends, but not themselves. For further study, Sonia Kleindorfer hopes birdwatchers will someday be able to snap a picture of a goose, ID it, and share its location with scientists. But she adds, just remember, her new research suggests that bird watching goes both ways: Geese can remember faces too. “If you are ever not kind to a goose,” she warns, “that goose may find you again.”
Kleindorfer thinks that facial recognition is going to play a really important role in conservation and ecology. “We need more computer scientists trained in behavioral ecology and we need more conservation scientists trained in computer science,” she says. “But working together, I think we can do this.”
1. What was the problem of Sonia Kleindorfer at work?| A．She was always stressed. | B．She was not able to count the geese. |
| C．She suffered a mental illness. | D．She couldn’t recognize all the geese. |
| A．By snapping pictures of the geese. | B．By presenting the full-sized pictures. |
| C．By identifying the beaks in the photos. | D．By building the database of the geese. |
| A．Birdwatching is a dangerous activity. |
| B．Geese can locate and find human beings. |
| C．Geese have the ability of facial recognition. |
| D．Birdwatchers can snap a picture of a goose. |
| A．Science training. | B．Computer science. |
| C．Photo-taking skills. | D．Cross-subject study. |
