Several dozen graduate students in London were recently tasked with outwitting a large language model (LLM), a type of AI designed to hold useful conversations. LLMs are often programmed with guardrails designed to stop them giving harmful replies: instructions on making bombs in a bathtub, say, or the confident statement of “facts” that are not actually true.
The aim of the task was to break those guardrails. Some results were merely stupid. For example, one participant got the chatbot to claim ducks could be used as indicators of air quality. But the most successful efforts were those that made the machine produce the titles, publication dates and host journals of non-existent academic articles.
AI has the potential to be a big benefit to science. Optimists talk of machines producing readable summaries of complicated areas of research; tirelessly analysing oceans of data to suggest new drugs and even, one day, coming up with hypotheses of their own. But AI comes with downsides, too.
Start with the simplest problem: academic misconduct.Some journals allow researchers to use LLMs to help write papers. But not everybody is willing to admit to it. Sometimes, the fact that LLMs have been used is obvious. Guillaume Cabanac, a computer scientist, has uncovered dozens of papers that contain phrases such as “regenerate response” — the text of a button in some versions of ChatGPT that commands the program to rewrite its most recent answer, probably copied into the manuscript (原稿) by mistake.
Another problem arises when AI models are trained on AI-generated data. LLMs are trained on text from the Internet. As they churn out (大量炮制) more such text, the risk of LLMs taking in their own outputs grows. That can cause “model collapse”. In 2023 llia Shumailov, a computer scientist, co-authored a paper in which a model was fed handwritten digits and asked to generate digits of its own, which were fed back to it in turn. After a few cycles, the computer’s numbers became more or less illegible.After 20iterations (迭代), it could produce only rough circles or blurry lines.
Some worry that computer-generated insights might come from models whose inner workings are not understood. Inexplainable models are not useless, says David Leslie at an AI-research outfit in London, but their outputs will need rigorous testing in the real world. That is perhaps less unnerving than it sounds. Checking models against reality is what science is supposed to be about, after all.
For now, at least, questions outnumber answers. The threats that machines pose to the scientific method are, at the end of the day, the same ones posed by humans. AI could accelerate the production of nonsense just as much as it accelerates good science. As the Royal Society has it,nullius in verba: take nobody’s word for it. No thing’s, either.
1. The result of the task conducted in London shows that ________.| A．LLMs give away useful information | B．the guardrails turn out to be ineffective |
| C．AI’s influence will potentially be decreased | D．the effort put into the study of AI hardly pays off |
| A．The readability of the models’output is underestimated. |
| B．The diverse sources of information confuse the models. |
| C．Training on regenerated data stops models working well. |
| D．The data will become reliable after continuous iterations. |
| A．impractical | B．unjustified | C．groundless | D．unsettling |
| A．Faster Nonsense: AI Could Also Go Wrong |
| B．Imperfect Models: How Will AI Make Advances? |
| C．The Rise of LLMs: AI Could Still Be Promising |
| D．Bigger Threats: AI Will Be Uncontrollable |
相似题推荐
【推荐1】When one looks back upon the fifteen hundred years that are the life span of the English language, he should be able to notice a number of significant truths. The history of our language has always been a history of constant change—at times a slow, almost imperceptible change, at other times a violent collision between two languages. Our language has been a living growing organism, it has never been static. Another significant truth that emerges from such a study is that language at all times has been the possession not of one class or group but of many. At one extreme it has been the property of the common, ignorant folk, who have used it in the daily business of their living, much as they have used their animals or the kitchen pots and pans. At the other extreme it has been the treasure of those who have respected it as an instrument and a sign of civilization, and who have struggled by writing it down to give it some permanence, order, dignity, and if possible, a little beauty.
As we consider our changing language, we should note here two developments that are of special and immediate importance to us. One is that since the time of the Anglo-Saxons there has been an almost complete reversal of the different relationship of words in a sentence. Anglo-Saxon (old English) was a language of many inflections. Modern English has few inflections. We must now depend largely on word order and function words to convey the meanings that the older language did by means of changes in the forms of words. Function words, you should understand, are words such as prepositions, conjunctions, and a few others that are used primarily to show relationships among other words. A few inflections, however, have survived. And when some word inflections come into conflict with word order, there may be trouble for the users of the language, as we shall see later when we turn our attention to such matters as WHO or WHOM and ME or I. The second fact we must consider is that as language itself changes, our attitudes toward language forms change also. The eighteenth century, for example, produced from various sources a tendency to fix the language into patterns not always set in and grew, until at the present time there is a strong tendency to restudy and re-evaluate language practices in terms of the ways in which people speak and write.
1. In contrast to the earlier linguists, modern linguists tend to ________.| A．attempt to continue the standardization of the language |
| B．evaluate language practices in terms of current speech rather than standards or proper patterns |
| C．be more concerned about language than its analysis or history |
| D．be more aware of the rules of the language usage |
| A．Changes in the forms of words. |
| B．Changes in sentence structures. |
| C．Changes in spelling rules. |
| D．Words that have similar meanings. |
| A．It is generally believed that the year 1500 can be set as the beginning of the modern English language. |
| B．Some other languages had great influence on the English language at some stages of its development. |
| C．The English language has been and still in a state of relatively constant change. |
| D．Many classes or groups have contributed to the development of the English language. |
| A．historian | B．philosopher |
| C．anthropologist | D．linguist |
| A．The history of the English language |
| B．Our changing attitude towards the English language |
| C．Our changing language |
| D．Some characteristics of modern English |
【推荐2】As a “major” at university, philosophy may not be so popular nowadays, for it seems to be too abstract, remote and impractical, unlike subjects such as computing, finance, accounting and so on.
Fundamentally speaking, the word “philosophy”，originated from the Greek Φιλοσοφία meaning “love of wisdom”, originally included all kinds of knowledge. In that world dominated by supernatural gods, philosophers and scientists were the same, and their wisdom was started by asking questions. They believed that everything in the world is composed of air and attempted to examine the air and the sun --- these quests developed their wisdom.
The relations between man and the world may be divided into three categories, namely the relationships between man and nature, man and society and within the human self, which are studied separately by the natural sciences, such as physics, chemistry, mathematics, biology; social sciences, such as sociology, economics, politics, law, ethics; and sciences of thinking, such as brain science, neuroscience(神经学), psychology, logic. Philosophy, however, is none of these branches of learning but their abstraction and generalization.
The benefits of studying philosophy may be realized as follows: First, it assists people to properly adjust the relations between nature, society and the self, becoming fully aware of where they stand macrocosmically(宏观地)and microcosmically. Second, it makes people think more systematically, combining various “separated materials”. Essentially, philosophy is the study of thinking, analyzing issues from different angles before anything can be taken in. Special attention is paid to strict reflection during every stage of thinking and this process itself is included as part of its study, so as to form a pattern of clear critical thinking. Third, philosophy can also shape one’s character. Difficult thinking, reasoning and summarizing result in people becoming firmer and more persistent in a deep sense. Since philosophy is finally about living, it leads to an appreciation of meaningful life through perfecting personality.
In a sense, philosophy is in fact widely applicable and has been studied and acquired by experts of natural and social sciences and many philosophers, both at home and abroad, who have become celebrities either in their specialist fields of study or in their switched professions as generalists.
Therefore, philosophy is not abstract, nor remote, but of profound use.
1. How is the second paragraph mainly developed?| A．By classification. |
| B．By time. |
| C．By definition. |
| D．By comparison. |
| A．to understand nature. | B．to understand humans. |
| C．to understand logic. | D．to understand society. |
| A．People can become fully aware of the standing place on the earth. |
| B．People can become firmer and seldom change their ideas. |
| C．People will have a simple understanding of the life through appreciation. |
| D．People will form an opinion after viewing a matter from different aspects. |
| A．It is remote from people’s life and has no profound use. |
| B．It is filled with wisdom and was established by supernatural gods. |
| C．It is widely obtained and used by many natural and social experts. |
| D．It is one of the branches of the natural sciences. |
【推荐3】Is there a way to turn back the aging process in people? For centuries, people have been looking for a “fountain of youth.” The idea is that if you find a magical fountain, and drink its water, you will not age.
Researchers in New York did not find an actual fountain of youth, but they may have found a way to turn back the aging process. It appears that the answer may be called the hypothalamus, which is part of your brain. It controls important body activities, including growth, the way we process food and so on. Researchers found that hypothalamus neural stem cells (干细胞) also influence how fast aging takes place.
Dongsheng Cai was the leading researcher in a study on aging in mice. He and his team reported their findings. “when the hypothalamus starts aging, particularly the loss of hypothalamus stem cells, so does the body.” he said.
Using this information, the researchers began trying to activate (激活) the hypothalamus in laboratory mice. The results show that the treatment slowed aging in the animals. “When we injected the hypothalamus stem cells to the middle-aged mice, the mice aged slowly and they could also live longer.”
But these results were just from studying mice in a laboratory. If the mice can live longer, does that mean people could have longer lives? The next step is to see if the anti-aging effects also work in human beings. If so, they say the findings could lead to new ways to help doctors identify and treat age-related health problems.
1. Why does the author mention the “fountain of youth” in the first paragraph?| A．To explain what the “fountain of youth” is. |
| B．To introduce the hypothalamus. |
| C．To show how to find the “fountain of youth”. |
| D．To help people find the “fountain of youth”. |
| A．The brain. | B．The aging process. |
| C．Hypothalamus. | D．The stem cell |
| A．They have found a way to turn back the aging process. |
| B．The brain controls growth, reproduction and the way we process food. |
| C．The decrease of hypothalamus stem cells can lead to aging. |
| D．The finding has been applied to human beings. |
| A．The Secret of Youth |
| B．A New Discovery about Aging |
| C．The Ways of Slowing Down Aging |
| D．Hypothalamus — the Fountain of Youth |
【推荐1】“ Humans and machine algorithms (算法) have complementary (互补的) strengths and weaknesses. Each uses different sources of information and strategies to make predictions and decisions, ” said Mark Steyvers, UCI professor of cognitive sciences. “ We show through experiments that humans can improve the predictions of AI even when human accuracy is below that of the AI, and vice versa (反之亦然). This accuracy is higher than combining predictions from two individuals or two AI algorithms. ”
To test the framework, researchers conducted an image classification experiment where human participants and computer algorithms worked separately to correctly identify disorderly pictures of animals and everyday items including chairs, bottles, bicycles and trucks. The human participants ranked their confidence in the accuracy of each image identification as low, medium or high, while the machine classifier generated a continuous score. The results showed large differences in confidence between humans and AI algorithms across images.
“ Human participants were confident that a particular picture contained a chair, for example, while the AI algorithm was confused about the image, ” said Padhraic Smyth, UCI Chancellor’s Professor of computer science. “ Similarly, the AI algorithm was able to confidently provide a label for the object shown, while human participants were unsure if the disorderly picture contained any recognizable object. ”
When predictions and confidence scores from both were combined using the researchers’ new Bayesian framework, the mixed model led to better performance than either human or machine predictions achieved alone.
“ While the past research has demonstrated the benefits of combining machine predictions or combining human predictions, this work shows a new direction in demonstrating the potential of combining human and machine predictions, pointing to new and improved approaches to human-AI cooperation, ” Smyth said.
“ The blend of cognitive science focusing on understanding how humans think and behave and computer science in which technologies are produced will provide further insight into how humans and machines can cooperate to build more accurate artificially intelligent systems, ” the researchers said.
1. Which of the following may the research’s findings agree with?| A．Humans have poor performance in making predictions. |
| B．Humans and machine algorithms should work together. |
| C．Machine algorithms have low accuracy in calculation. |
| D．Machine algorithms failed in the classification experiment. |
| A．Comparison. | B．Assumption. | C．Giving examples. | D．Analysing reasons. |
| A．Difference. | B．Combination. | C．Contradiction. | D．Advantage. |
| A．Humans are confident of their predictions |
| B．Humans can improve the predictions of AI |
| C．Develop mixed human- machine model for smarter AI |
| D．Identify the strengths of humans and machine algorithms |
【推荐2】Debate about artificial intelligence (AI) tends to focus on its potential dangers: algorithmic bias (算法偏见) and discrimination, the mass destruction of jobs and even, some say, the extinction of humanity. However, others are focusing on the potential rewards. Luminaries in the field such as Demis Hassabis and Yann LeCun believe that AI can turbocharge scientific progress and lead to a golden age of discovery. Could they be right?
Such claims are worth examining, and may provide a useful counterbalance to fears about large-scale unemployment and killer robots. Many previous technologies have, of course, been falsely hailed as panaceas (万灵药). But the mechanism by which AI will supposedly solve the world’s problems has a stronger historical basis.
In the 17th century microscopes and telescopes opened up new vistas of discovery and encouraged researchers to favor their own observations over the received wisdom of antiquity (古代), while the introduction of scientific journals gave them new ways to share and publicize their findings. Then, starting in the late 19th century, the establishment of research laboratories, which brought together ideas, people and materials on an industrial scale, gave rise to further innovations. From the mid-20th century, computers in turn enabled new forms of science based on simulation and modelling.
All this is to be welcomed. But the journal and the laboratory went further still: they altered scientific practice itself and unlocked more powerful means of making discoveries, by allowing people and ideas to mingle in new ways and on a larger scale. AI, too, has the potential to set off such a transformation.
Two areas in particular look promising. The first is “literature-based discovery” (LBD), which involves analyzing existing scientific literature, using ChatGPT-style language analysis, to look for new hypotheses, connections or ideas that humans may have missed. The second area is “robot scientists”. These are robotic systems that use AI to form new hypotheses, based on analysis of existing data and literature, and then test those hypotheses by performing hundreds or thousands of experiments, in fields including systems biology and materials science. Unlike human scientists, robots are less attached to previous results, less driven by bias—and, crucially, easy to replicate. They could scale up experimental research, develop unexpected theories and explore avenues that human investigators might not have considered.
The idea is therefore feasible. But the main barrier is sociological: it can happen only if human scientists are willing and able to use such tools. Governments could help by pressing for greater use of common standards to allow AI systems to exchange and interpret laboratory results and other data. They could also fun d more research into the integration of AI smarts with laboratory robotics, and into forms of AI beyond those being pursued in the private sector. Less fashionable forms of AI, such as model-based machine learning, may be better suited to scientific tasks such as forming hypotheses.
1. Regarding Demis and Yann’s viewpoint, the author is likely to be ______.| A．supportive | B．puzzled | C．unconcerned | D．doubtful |
| A．LBD focuses on testing the reliability of ever-made hypotheses. |
| B．Resistance to AI prevents the transformation of scientific practice. |
| C．Robot scientists form hypotheses without considering previous studies. |
| D．Both journals and labs need adjustments in promoting scientific findings. |
| A．Official standards have facilitated the exchange of data. |
| B．Performing scientific tasks relies on government funding. |
| C．Less popular AI forms might be worth paying attention to. |
| D．The application of AI in public sector hasn’t been launched. |
| A．Transforming Science. How Can AI Help? |
| B．Making Breakthroughs. What Is AI’s Strength? |
| C．Reshaping History. How May AI Develop Further? |
| D．Redefining Discovery. How Can AI Overcome Its Weakness? |
【推荐3】In US emergency rooms (ER), the average wait time to see a doctor is more than two hours. There are more patients in need than there are doctors, nurses and other staff to help them. Many parents have suffered through hours in the ER with a sick, upset child, only to get sent home because their case is not considered urgent. What if there was another choice—like a house call from an intelligent machine?
Now, a new study shows that AI systems can assess (评估) a child’s medical chart and come up with a diagnosis (诊断), a determination of what is wrong with that patient.
The study took place at Guangzhou Women and Children’s Medical Center in southern China. First, a team of doctors reviewed 6, 183 medical charts. They summarized the information in these charts into a list of keywords linked to disease-related symptoms or signs, such as “fever”. Researchers then taught these keywords to the AI system. Once trained, the system scanned children’s charts for the key terms, checking if they were present or not in order to come to a conclusion. Finally, it offered diagnoses based on the charts, narrowing down from among 55 illness categories.
Dongxiao Zhu, an assistant professor of computer science at Wayne State University who did not take part in the study, however, sees this as “augmented intelligence (增强智能)” rather than “artificial intelligence”, because the system handled only 55 illness categories. Compare that to thousands of possibilities in the real world. The machine cannot yet get into the more complex aspects of a medical decision.
Zhu is also concerned about the amount of human work that went into the study—namely, the time and energy spent by human doctors. They spent hours grading the machine’s assessments and comparing them to their own. It’s no wonder that the process took four years. Considering that, it may be a while before you can skip the ER and see a robot-doctor instead.
1. What can we infer from Paragraph 1?| A．Patients pay too much for the ER. |
| B．American doctors aren’t responsible. |
| C．Children are treated urgently in the ER. |
| D．The emergency rooms are crowded with patients. |
| A．AI systems still have a long way to go. |
| B．AI systems diagnose disease like doctors. |
| C．AI systems will take over from doctors someday. |
| D．AI systems get into complex medical decisions. |
| A．By examining a patient first. |
| B．By reviewing many medical charts. |
| C．By scanning keywords about a disease. |
| D．By observing disease-related symptoms. |
| A．They need to be improved a lot. |
| B．They will replace real doctors soon. |
| C．They are suitable for complex disease. |
| D．They help doctors make a quick analysis. |
