Playing Minecraft, a video game, could be key to creating adaptable (可调试的) artificial intelligence models that can pick up a variety of tasks the way humans do.
Steven James at the University of the Witwatersrand in South Africa and his colleagues developed a test within Minecraft to measure the general intelligence of AI models. This MinePlanner test rates an AI’s ability to ignore unimportant details while solving a complex multi-step problem.
Lots of AI training “cheats” by giving a model all the data it needs to learn how to do a job and nothing related to it, says James. Future AI models will need to deal with confusing problems, and he hopes that MinePlanner will guide that research. AI working to solve a problem in the game will see everything involved, including objects and other details that aren’t necessarily needed to solve a problem and must be ignored. It will have to survey its surroundings and work out by itself what is and isn’t needed.
The virtual test consists of 15 construction problems, and each one can be easy, medium and hard. To finish each task, the AI may need to take some steps in between, like building stairs to reach a certain height. This means the AI has to think about the whole picture and plan what to do next.
State-of-the-art planning AI models were unable to complete any of the tough problems and they only do a little better on the easier ones, suggesting there is room for improvement.
“We can’t require a human designer to come in and tell the AI exactly what it should and shouldn’t care about for each and every task it might have to solve,” says James. “That’s the problem we’re trying to address.”
1. Why is playing Minecraft important for creating adaptable AI models?A.Because it offers tasks that require human-like ability. |
B.Because it provides different situations for AI training. |
C.Because it is an easy game for AI to learn and master. |
D.Because it is a video game popular among AI scientists. |
A.AI models are trained in a dishonest way. |
B.AI models are only taught to perform simple tasks. |
C.AI models are offered all the necessary data for a task. |
D.AI models are given wrong information during training. |
A.It solved all the 15 construction problems. |
B.It performed poorly in handling hard tasks. |
C.It was completely unable to deal with any task. |
D.It performed excellently in solving easy problems. |
A.Adaptabe ability. | B.Computing speed. |
C.Communicative skills. | D.Data-processing power. |
相似题推荐
【推荐1】Modern Europeans came from three major groups of ancient humans, not two as was thought before, according to a study published on Wednesday.
Until now, it was widely believed that Europeans evolved (进化) from two prehistoric groups. One was early farmers who moved into Europe from the Middle East about 7,500 years ago. The other was local hunter-gatherers who had lived in Europe for more than 40,000 years.
But a new study in the journal Nature says there was a third group in the mix: people from northern Eurasia. They lived in today’s Russia and northern Asia. The finding means that northern Eurasians contributed to the human genes (基因) both in Europe and North America.
Their influence on the Americas has been proved by previous studies which showed that they reached modern-day Alaska in the US more than 15,000 years ago. They crossed an “ice bridge” that connected islands in the Bering Strait, a narrow passage of water between Asia and North America, at the time.
Researchers collected genetic information in nine ancient humans’ bones. The remains were found in Sweden, Luxembourg and Germany. They were one farmer from about 7,000 years ago and eight hunter-gatherers who lived about 8,000 years ago, before the coming of agriculture.
The researchers compared the information with the gene pool of 2,345 present-day people living all over the world. They found almost all Europeans have ancestry from all three of those ancient groups.
The ancient northern Eurasians contributed up to 20% of the genetics of Europeans, although this was the smallest percentage among the three ancestral groups.
People in northern Europe, especially the Baltic states, have the highest percentage of western European hunter-gatherer ancestry. Up to 50% of the DNA of Lithuanians of northeast Europe comes from this group.
Southern Europeans had more of their genetic ancestry from the ancient farmers. Up to 90% of the DNA of Sardinians of Italy can be traced back to (追溯到) these early European immigrants.
Looking ahead, the researchers plan to find out when the ancient northern Eurasians arrived in Europe.
1. Ancient people from northern Eurasia _____.A.brought agriculture into Europe | B.reached Europe about 7,000 years ago |
C.were hunter-gatherers in northern Asia | D.were also ancestors of modern Europeans |
A.About 7,500 years ago. | B.About 8,000 years ago. |
C.More than 15,000 years ago. | D.More than 40,000 years ago. |
A.By analysing genes. | B.By visiting ancient sites. |
C.By doing medical experiments. | D.By comparing studies in different periods. |
A.Hunter-gatherers in western Europe. | B.Hunter-gatherers in northern Europe. |
C.Ancient farmers from the Middle East. | D.Ancient farmers from northern Eurasia. |
【推荐2】Faced with an attempt by a new chatbot to imitate (模仿) his own songs, the musician Nick Cave delivered a strong response: It was “an absolutely horrible attempt”. He understood that AI was in its babyhood, but could only conclude that the true horror might be that “it will forever be in its babyhood”. While a robot might one day be able to create a song, he wrote, it would never grow beyond “a kind of burlesque (滑稽的模仿)”, because robots-being composed of data-are unable to suffer, while songs arise out of suffering.
Fans of Cave and his band will agree that his music is inimitable, but that doesn’t mean they would necessarily be able to tell the difference. A few days before Cave’s remarks, experts were asked to distinguish between four genuine artworks and their AI imitations. Their conclusions were wrong five times out of 12, and they were only unitedly right in one of the four picture comparisons.
These are party games, but they point to an unfolding challenge that must be managed as a matter of urgency because, like it or not, Al art is upon us. The arrival of the human-impersonating ChatGPT might have increased general awareness, but artists across a wide range of disciplines are already exploring its potential, with the dancer Wayne McGregor and London’s Young Vic Theatre among those who have created AI-based works.
A strongly-worded report from Communications and Digital Committee (CDC) issued a wake-up call to the government, urging it to raise its game in educating future generations of tech-savvy professionals, and tackling key regulatory challenges. These included reviewing reforms to intellectual property law, strengthening the rights of performers and artists, and taking action to support the creative sector in adapting to the disturbances caused by swift and stormy technological change.
While developing Al is important, it should not be pursued at all costs, the CDC stressed. It deplored the failure of the Department for Digital, Culture, and Media to offer a defence against proposed changes to intellectual property law that would give copyright exemption (版权豁免) to any work, anywhere in the world, involving AI text and data mining.
The challenges of AI are both philosophical, as Cave suggested, and practical. They will unfold over the short and long term. State-of -the-art creative industries have a key role to play in shaping and exploring the philosophical ones, but they must have the practical help they require to survive and be successful. They need it now.
1. Why does the author mention the four picture comparisons in Paragraph 2?A.To stress the similarities between AI art and human art. |
B.To argue that human art will be replaced by AI art. |
C.To prove AI is stretching the boundaries of art. |
D.To imply AI art cannot be underestimated. |
A.Clearly analyzed. | B.Bravely suffered. |
C.Strongly criticized. | D.Accurately perceived. |
A.Creative industries are responsible for causing the AI problem. |
B.Tech professionals need more training to better understand AI art. |
C.Some artists see AI as a tool even though it is a threat. |
D.The quality of AI art dismisses concerns about intellectual property. |
A.The Creative Thief: AI Makes Perfect Art |
B.AI in Art: A Battle That Must Be Fought |
C.A Great Opportunity: The Importance of AI on Art |
D.The Rise of AI Art: An attempt to imitate songs |
【推荐3】When you're a teen you start being more aware of what other people think. There seems to be a “right” thing to wear, or say, or do. There also seem to be things that you shouldn’t do-things that could be embarrassing, or lose your points with friends. This can lead to social anxiety.
Some kids feel so anxious that they develop something called social anxiety disorder (障碍), which is diagnosed (诊断) when you worry so much about how you appear to others that you stop doing things you need to and want to do for fear of embarrassing yourself.
Kids with social anxiety disorder aren't just nervous when they’re at parties or giving a speech in class. Even small interactions (互动), like answering a question in class or eating with friends in the cafeteria can feel extremely scary to kids with social anxiety disorder. That's because they fear they might accidentally do something embarrassing or offensive, and it will make others judge or even reject them.
And while kids who are just shy will gradually warm up to new people and situations over time, kids with social anxiety don't. Shyness might hold you back to some degree from doing things, but it won't deeply influence your ability to do your job as a teenager, which is to function in school, function in your family, and to have friends and be a part of your peer-related community. But social anxiety will.
Many teens experience anxiety disorders. Being brave and telling someone how you feel might seem scary, but if you can get over that obstacle, someone will want to listen. Asking for help can be hard, but it really is important.
1. What is the main cause of social anxiety?A.Lack of friendship. | B.One's appearance. |
C.School performance. | D.Other’s judgment. |
A.Alice, who skips school to avoid answering questions in class. |
B.David, who always thinks he can't do as well as his classmates. |
C.Chris, who feels nervous every time he gives a speech in public. |
D.Jenny, who has been in low spirits since she failed the last exam. |
A.it influences people around you |
B.it only makes your grades suffer |
C.it greatly affects your normal life |
D.it is related to your study or work |
A.What Is Social Anxiety Disorder |
B.How Social Fear Ruins Relationships |
C.Why Teens Suffer from Social Anxiety? |
D.When Anxiety Disorders Go Unnoticed |
【推荐1】Using the power of artificial intelligence (AI), scientists have revealed new insights into the creation and destruction of mass extinction. Contrary to conventional knowledge, their study suggests that larger extinctions are not always a form of “creative destruction” that allows new organisms (生物体) to radiate and evolve. Instead, it suggests that mass extinction is rarely associated with new species of radiation.
Dr. Hoyal Cuthill, the lead study author from the University of Essex in the UK and the Tokyo Institute of Technology, said in a statement, “Some of the most challenging things to understand the history of life are the vast timelines involved and the number of species. New machine learning applications can help us understand this information in human-readable form. This means that we can, so to speak, hold the evolution of half a billion years in the palm of our hand and gain new insights from what we see.”
They concluded that mass extinction and later radiation were not connected as previously thought. Within 5 percent of the most significant periods of disruption (中断), AI detected “big five” mass extinctions, seven more mass extinctions, two mass extinction-radiation events, and 15 mass radiations. Most importantly, it discovers that massive radiation and extinction rarely occurred with each other, changing the view that greater extinction leads to a kind of deep cycle-like species radiation of nature. It appears that larger extinctions are certainly not the engine of evolutionary radiation. Take the Cambrian eruption for example and it was about 41 million years ago when a large group of animals first appeared on the record of the first fossil record and the dawn of a high mobile animal equipped with modern physical features.
This new study found that a handful of other notable explosions of biodiversity, including the Cambrian eruption, usually occurred at a time when they were largely isolated (隔离) from extinction. Dr. Nicholas Guttenberg, a study co-author from the Tokyo Institute of Technology explained, “Ecosystems are dynamic and you don’t need anything to exist to allow something new to appear.”
1. What does the first paragraph serve as?A.An explanation of artificial intelligence. | B.A background of researchers’ study. |
C.The reasons for creative destruction. | D.The result of researchers’ new study. |
A.AI contributes a lot to the study of evolution. | B.Understanding the history of life is very difficult. |
C.New AI machines learn applications better. | D.Biological evolution can be controlled easily. |
A.Mass extinction is unlikely to cause evolutionary radiation. |
B.The first animal with modern features occurred during eruption. |
C.The volcanic eruption led to a mass extinction and radiation. |
D.The idea of deep cycle-like species radiation has turned out true. |
A.New processes of biological evolution. | B.New view of radiation from natural species. |
C.New understanding of mass extinction. | D.New outbreaks of biological populations. |
【推荐2】No one can say whether human-like robots will have a sweet dream, but they will almost certainly need periods of rest that offer benefits like what sleep provides for living brains, according to new research from Los Alamos National Laboratory (LANL).
“We study spiking neural networks (尖峰神经网络), which are systems that learn much as living brains do,” said Yijing Watkins, a computer scientist from LANL. “We trained a neuromorphic processor in a way how humans and other biological systems learn from their environment during childhood development.” Watkins and her research team found that the network simulations (模拟) became unstable after continuous periods of unattended learning. When they exposed the networks to states that are analogous to what living brains experience during sleep, stability was recovered. “It was as though we were giving the neural networks a good night’s rest,” said Watkins.
The discovery came about as the research team worked to develop neural networks that are as close as how humans and other biological systems learn to see. The group initially struggled with stabilizing simulated neural networks undergoing unattended dictionary training, which involves classifying objects without providing examples to compare them to. The researchers exposed the networks to an artificial simulation of sleep as nearly a final effort to stabilize them. They experimented with various types of noise. The best results came when they used waves of so-called Gaussian noise. These waves can make sure that the neural networks keep stable.
The group’s next goal is to apply their algorithm (算法) to Intel’s Loihi, a product which uses spiking neural networks to work. They hope allowing Loihi to sleep from time to time will enable it to stably process information from a camera in real time. If it can confirm the need for sleep in artificial brains, we can probably expect the same to be true of human-like robots and other intelligent machines that may come about in the future.
1. What did Watkins and her research team find?A.Neural networks are far from stable. |
B.Artificial brains may need sleep as well. |
C.Neural networks are very sensitive to noise. |
D.Human-like robots need equal rest to humans. |
A.Contrary. | B.Beneficial. | C.Similar. | D.Related. |
A.To better update Loihi. |
B.To make sure Loihi get enough sleep. |
C.To enable Loihi to process information faster. |
D.To further confirm their discovery. |
A.Electronics & Physics. | B.Mind & Brain. |
C.Business & Industry. | D.Computers & Science. |
【推荐3】Over a decade ago, the science fiction series Black Mirror showed a story about using artificial intelligence to bring back loved ones. Thanks to technological advancement, the latest AI technology brings hope of recreating loved ones through virtual form.
Recently, it was revealed that renowned musician Tino Bao had created a digital version of his daughter, who in 2021 passed away due to a rare disease at the age of 22. With the help of Xiaoice, Microsoft Asia’s AI branch based in China, Bao was able to recreate his daughter Bao Rong as an AI-powered virtual figure. Tino Bao, his family and the team went to great lengths to piece together pictures, audio-visual recordings, and narrated memories of Rong. After thousands of trials and errors, they presented the life-like digital human who speaks in both Mandarin and English. Bao has finally launched a company to bring the technological miracle to other families experiencing such loss.
However, this technological achievement has also been accompanied by many controversies. The idea of bringing back a loved one through AI may seem comforting. Still, researchers caution that constantly interacting with an AI version of a deceased loved one could potentially affect the natural grieving process, leaving individuals stuck in a state of denial. This could lead to prolonged grief and other mental health issues. Furthermore, there’s a risk that people may become overly dependent on the technology, prioritizing their interactions with the AI version over forming new, real-life relationships.
With the case of Bao Rong out in the open, AI resurrection (复活) has become a business. From simple “talking pictures” that sound nothing like the loved ones to deep fake videos of actors speaking with the faces of those who passed away, the lower end of the market is more “artificial” than “intelligent”. Therefore, it is crucial to establish proper regulations and ethical standards to ensure that the use of AI-resurrection technology is safe and responsible.
1. Why does the author mention Black Mirror?A.To advertise Microsoft’s virtual products. |
B.To lead to the discussion about AI resurrection. |
C.To prove the predictive power of science fiction. |
D.To highlight the advancement of artificial intelligence. |
A.The popularity of AI-resurrection technology. |
B.Reasons for making a digital version of Bao Rong. |
C.Challenges of developing a life-like digital human. |
D.The process of creating an AI-powered virtual figure. |
A.It still has some technical limitations. |
B.It helps people build new real-life relationships. |
C.It can be a person’s obstacle to recovering from sadness. |
D.It enables individuals to preserve good memories of loved ones. |
A.Strengthen the management of the technology. |
B.Apply the technology to making pictures and videos. |
C.Put the technology into the market as soon as possible. |
D.Educate the public about the potential risks of technology. |