Questões de Concurso Comentadas sobre inglês

Questões Discursivas

Foram encontradas 19.439 questões

Q3781917 Inglês
Leia o texto a seguir para responder a questão:

        Creativity is a quality which manifests itself in many different ways, and this is one of the reasons it has proved so difficult to define. As Amabile (1996) points out, ‘a clear and sufficiently detailed articulation of the creative process is not yet possible.’ Yet we readily recognise creativity when we meet it, even if we cannot define it precisely. For all practical purposes this is enough, and we do not need to spend too much time agonising over a definition.

        There are of course some features which are almost always present in a creative act. The core idea of ‘making something new’ is at the heart of creativity. But novelty is not alone sufficient for something to be recognised as creative. We could, for example, wear a clown’s red nose to class. This would certainly be doing something new and unusual but it would only count as creative if we then did something with it, like creating a new persona. It is also necessary for creative acts to be recognised and accepted within the domain in which they occur. They need to be relevant and practicable – not just novel. Sometimes creative ideas are ahead of their time and have to wait for technology to catch up. Leonardo da Vinci designed an aeroplane in the 15th century, but before aeroplanes could become a reality, materials and fuels had to be available.

        […]

        […] Boden (1990) takes an AI (artificial intelligence) approach to investigating creativity. She asks what a computer would need to do to replicate human thought processes. This leads to a consideration of the self-organising properties of complex, generative systems through processes such as parallel distributed processing. For her, creativity arises from the systematic exploration of a conceptual space or domain (mathematical, musical or linguistic). She draws attention to the importance of constraints in this process. ‘Far from being the antithesis of creativity, constraints on thinking are what make it possible’ (p. 82). Chaos theory (Gleick, 1987) tends to support her ideas. Boden’s approach is richly suggestive for language acquisition, materials writing and for teaching, in that all are rooted in complex, self-organising systems.

(Alan Maley, Nik Peachey. Creativity in the English language classroom.)
According to the second paragraph, in addition to being new, the necessary condition for an act or idea to be recognized and accepted as creative within its specific domain is
Alternativas
Q3781914 Inglês
Leia o texto a seguir para responder a questão:

        There is no single way of teaching English, no single way of learning it, no single motive for doing so, no single syllabus or textbook, no single way of assessing proficiency and, indeed, no single variety of English which provides the target of learning. It is tempting, but unhelpful, to say there are as many combinations of these as there are learners and teachers. The proliferation of acronyms in English Language Teaching reflects this diversity of models. By a ‘model’ I do not mean a particular variety of English – such as US or British – though selection of a particular variety may play a role.

        By a ‘model’ of English I mean a complex framework, which includes issues of methodology and variety, but goes beyond these to include other dimensions of the context and practice of learning English.

        It is becoming clear that these issues are not easily separable. The appropriateness of content clearly depends on such things as the age of the learner and whether English is to be used primarily as a language of international communication or for survival communication with native speakers, perhaps whilst on holiday in the UK or some other English-speaking country. This is why I have identified broad models which can be thought of as configurations of the factors listed in the box.

        There are many stakeholders involved in the teaching and learning process, each of whom may have a different view. Learners, their families, teachers, governments, employers, textbook publishers, examination providers – all now possess an interest in the English language business.

(David Graddol. English Next. Adaptado)
According to the author, the defining characteristic of a ‘model’ of English in the context of language teaching is a
Alternativas
Q3781913 Inglês
Leia o texto a seguir para responder a questão:

        There is no single way of teaching English, no single way of learning it, no single motive for doing so, no single syllabus or textbook, no single way of assessing proficiency and, indeed, no single variety of English which provides the target of learning. It is tempting, but unhelpful, to say there are as many combinations of these as there are learners and teachers. The proliferation of acronyms in English Language Teaching reflects this diversity of models. By a ‘model’ I do not mean a particular variety of English – such as US or British – though selection of a particular variety may play a role.

        By a ‘model’ of English I mean a complex framework, which includes issues of methodology and variety, but goes beyond these to include other dimensions of the context and practice of learning English.

        It is becoming clear that these issues are not easily separable. The appropriateness of content clearly depends on such things as the age of the learner and whether English is to be used primarily as a language of international communication or for survival communication with native speakers, perhaps whilst on holiday in the UK or some other English-speaking country. This is why I have identified broad models which can be thought of as configurations of the factors listed in the box.

        There are many stakeholders involved in the teaching and learning process, each of whom may have a different view. Learners, their families, teachers, governments, employers, textbook publishers, examination providers – all now possess an interest in the English language business.

(David Graddol. English Next. Adaptado)
What core idea regarding the current state of English Language Teaching (ELT) does the author aim to convey?
Alternativas
Q3781911 Inglês
Leia o texto a seguir para responder as questões:

Shifting paradigms in language teaching

        Foreign language teaching has long relied on written texts as a source of language input. Until relatively recently, however, the sentence has been the privileged unit of meaning and analysis. The grammar-translation method of the nineteenth and twentieth centuries, for example, illustrated grammatical principles via exemplary sentences. The pedagogical goal was to recode sentences written in the foreign language into one’s mother tongue, with heavy emphasis placed on accuracy and completeness. During the audiolingual era, from the 1940s to the 1960s, the emphasis shifted to spoken language and dialogues were used as language models, but the individual sentence remained the focus of repetition and drills. Again, formal accuracy remained paramount. In the 1960s, with the advent of ‘cognitive-code learning’ theory (following Chomsky’s rejection of behavioristic models of language learning in the late 1950s), teachers’ goals gradually shifted from instilling accurate language habits, to fostering learners’ mental construction of a second language system. Rule learning was reintroduced, but still only at the level of the individual sentence. Indeed, even today, many introductory level foreign language courses are organized around a planned sequence of grammatical structures that are exemplified in sample sentences for intensive practice.

(Richard Kern. Literacy and language teaching)
Based on the historical overview, the sustained pattern that can be observed regarding the unit of linguistic focus across the Grammar-Translation, Audiolingual, and early Cognitive-Code learning periods is
Alternativas
Q3781910 Inglês
Leia o texto a seguir para responder a questão:

        The learning principles that good games incorporate are by no means unknown to researchers in the learning sciences. In fact current research on learning supports the sorts of learning principles that good games use, though these principles are often exemplified in games in particularly striking ways (for a survey and citations of the literature, see Gee 2003). However, many of these principles are much better reflected in good games than they are in today’s schools, where we also ask young people to learn complex and challenging things. With the current return in our schools to skill-and-drill and curricula driven by standardized tests, good learning principles have, more and more, been left on the cognitive scientist’s laboratory bench and, I will argue, inside good computer and video games.

        Game design involves modeling human interactions with and within complex virtual worlds, including learning processes as part and parcel of these interactions. This is, in fact, not unlike design research in educational psychology where researchers model new forms of interaction connected to learning in classrooms (complex worlds, indeed), study such interactions to better understand how and why they lead to deep learning, and then ultimately disseminate them across a great many classrooms (see, for example, the chapters in Kelly 2003).

(James Paulo Gee. Situated Language and Learning: a critique of traditional schooling)
Based on the author’s comparison of game design to design research in education, a pedagogical practice an English as a Foreign Language teacher should prioritize to move beyond the criticized approach is
Alternativas
Q3781909 Inglês
Leia o texto a seguir para responder a questão:

        The learning principles that good games incorporate are by no means unknown to researchers in the learning sciences. In fact current research on learning supports the sorts of learning principles that good games use, though these principles are often exemplified in games in particularly striking ways (for a survey and citations of the literature, see Gee 2003). However, many of these principles are much better reflected in good games than they are in today’s schools, where we also ask young people to learn complex and challenging things. With the current return in our schools to skill-and-drill and curricula driven by standardized tests, good learning principles have, more and more, been left on the cognitive scientist’s laboratory bench and, I will argue, inside good computer and video games.

        Game design involves modeling human interactions with and within complex virtual worlds, including learning processes as part and parcel of these interactions. This is, in fact, not unlike design research in educational psychology where researchers model new forms of interaction connected to learning in classrooms (complex worlds, indeed), study such interactions to better understand how and why they lead to deep learning, and then ultimately disseminate them across a great many classrooms (see, for example, the chapters in Kelly 2003).

(James Paulo Gee. Situated Language and Learning: a critique of traditional schooling)
According to the author, “good learning principles” are neglected and relegated because of
Alternativas
Q3781908 Inglês
Leia o texto a seguir para responder a questão:

Language monitor

        A new topic area will quickly generate the need to acquire new language in the form of vocabulary, structures, and pronunciation. It is a good idea to have ready a way of coping with this demand.

        If students can feel that they have the time and opportunity to master the use of language that either you or they have identified as being necessary for a certain stage in a project, this will go a long way to increasing their confidence and language competence.

        One way to do this is to produce a language monitor which focuses on vocabulary and structures that have been identified as being useful.

        This allows other students to read it and absorb the word or phrase, the meaning, pronunciation, associated words or collocations, and how to use it in a sentence. They can also add their own cards. The vocabulary monitor remains on the noticeboard throughout the project, constantly available for reinforcement and consolidation. It can also be used as a source of vocabulary games.

        In addition to this or as an alternative, if you have suitable computer facilities available, electronic lists could be created. Students can add to the lists in the same way as the noticeboard. The updated list can be printed out at regular intervals and put on the noticeboard and handouts given to the students.

(Diana L. Fried-Booth. Project Work. Adaptado)
Considering the concept of the “Language Monitor,” an English as a Foreign Language teacher best integrates this tool to foster learner autonomy and confidence in a project-based learning environment by
Alternativas
Q3781907 Inglês
Leia o texto a seguir para responder a questão:

Language monitor

        A new topic area will quickly generate the need to acquire new language in the form of vocabulary, structures, and pronunciation. It is a good idea to have ready a way of coping with this demand.

        If students can feel that they have the time and opportunity to master the use of language that either you or they have identified as being necessary for a certain stage in a project, this will go a long way to increasing their confidence and language competence.

        One way to do this is to produce a language monitor which focuses on vocabulary and structures that have been identified as being useful.

        This allows other students to read it and absorb the word or phrase, the meaning, pronunciation, associated words or collocations, and how to use it in a sentence. They can also add their own cards. The vocabulary monitor remains on the noticeboard throughout the project, constantly available for reinforcement and consolidation. It can also be used as a source of vocabulary games.

        In addition to this or as an alternative, if you have suitable computer facilities available, electronic lists could be created. Students can add to the lists in the same way as the noticeboard. The updated list can be printed out at regular intervals and put on the noticeboard and handouts given to the students.

(Diana L. Fried-Booth. Project Work. Adaptado)
According to the text, the main purpose of the “Language Monitor” is to
Alternativas
Q3781906 Inglês
Leia o texto a seguir para responder a questão:

        […] The action research cycle results show that task design should follow a certain sequence: First, tasks should focus on gaining an understanding of the e-literacy skills required when working with tools such as forums, wikis, and social bookmarking sites for language learning and teaching purposes. Ideally, this understanding should enable teachers to provide a rationale for using bespoke tools. Next, tasks should raise their awareness of a tool’s specific affordances, i.e. the constraints and possibilities of the modes available for meaning making and communication (Hampel & Hauck, 2006). This will allow the teachers to move to the next level of Hampel and Stickler’s (2005) skills pyramid by fostering their multimodal communicative competence and thus their professional literacy (Willis, 2001). These steps are a prerequisite for the subsequent phase in which teachers themselves design tasks with the goal of fostering, in turn, their learners’ multimodal competence and autonomy since merely equipping learners with creative and democratic representational online resources will not necessarily result in higher student control over the learning process or the development of autonomy (Hampel & Hauck, 2006).

(Carolin Fuchs, Andreas Müller-Hartmann, Mirjam Hauck. Promoting learner autonomy through multiliteracy skills development in cross-institutional exchanges. Adaptado)
Based on the final phase described in the text, where teachers design tasks to foster learner autonomy and multimodal competence, a teacher planning a task using a discussion forum should ensure the task
Alternativas
Q3781905 Inglês
Leia o texto a seguir para responder a questão:

        […] The action research cycle results show that task design should follow a certain sequence: First, tasks should focus on gaining an understanding of the e-literacy skills required when working with tools such as forums, wikis, and social bookmarking sites for language learning and teaching purposes. Ideally, this understanding should enable teachers to provide a rationale for using bespoke tools. Next, tasks should raise their awareness of a tool’s specific affordances, i.e. the constraints and possibilities of the modes available for meaning making and communication (Hampel & Hauck, 2006). This will allow the teachers to move to the next level of Hampel and Stickler’s (2005) skills pyramid by fostering their multimodal communicative competence and thus their professional literacy (Willis, 2001). These steps are a prerequisite for the subsequent phase in which teachers themselves design tasks with the goal of fostering, in turn, their learners’ multimodal competence and autonomy since merely equipping learners with creative and democratic representational online resources will not necessarily result in higher student control over the learning process or the development of autonomy (Hampel & Hauck, 2006).

(Carolin Fuchs, Andreas Müller-Hartmann, Mirjam Hauck. Promoting learner autonomy through multiliteracy skills development in cross-institutional exchanges. Adaptado)
The text mentions that understanding a tool’s specific affordances involves recognizing “the constraints and possibilities of the modes available for meaning making and communication”.
For a language teacher, a significant implication of focusing on a tool’s constraints in task design is
Alternativas
Q3781900 Inglês
Leia o texto a seguir para responder a questão:

        Our analysis of the Time to Share series revealed that it follows the principles of recent theories in the development of the learning activities although there is an excessive use of the Portuguese language. Students are somewhat well encouraged to actively participate but more could be done in this sense. Further responsibility could be put upon the students in terms of interactions and research on the web. They are digital natives and their familiarity with this new world can make a difference in their involvement with learning English online for today and for the future. Their success depends on them, and they must be encouraged to learn by themselves.

(Reinildes Dias & Ana Emília Fajardo Turbin. The two “multis” and the multiliteracies pedagogy: “shaking hands” in the Brazilian English public education for teens.)
Choose the statement that most accurately conveys the authors’ critique and suggestions for improving the Time to Share series as presented in the text.
Alternativas
Q3781899 Inglês
Leia o texto a seguir para responder a questão:

        Our analysis of the Time to Share series revealed that it follows the principles of recent theories in the development of the learning activities although there is an excessive use of the Portuguese language. Students are somewhat well encouraged to actively participate but more could be done in this sense. Further responsibility could be put upon the students in terms of interactions and research on the web. They are digital natives and their familiarity with this new world can make a difference in their involvement with learning English online for today and for the future. Their success depends on them, and they must be encouraged to learn by themselves.

(Reinildes Dias & Ana Emília Fajardo Turbin. The two “multis” and the multiliteracies pedagogy: “shaking hands” in the Brazilian English public education for teens.)
Based on the authors’ analysis, what is the most emphatically suggested pedagogical shift needed to enhance student involvement and leverage their status as “digital natives” in the learning process?
Alternativas
Q3781898 Inglês
Leia o texto a seguir para responder a questão:

        The more traditional methods and approaches to teaching culture, such as movies and video, can be enhanced through the integration of digital media. Feature films have become readily available and have been included in numerous textbooks and designed to actively involve the learner (Aparisi, Blanco, & Rinka, 2007; Blanco & Tocaimaza-Hatch, 2007). Foreign language instructors are beginning to incorporate more movies in the foreign language classroom as “an accessible window” (Bueno, 2009, p. 319) to the target culture through “combined effects of images, sounds, camera, plots and dialogue” (Stephens, 2001, p. 2). According to Bueno (2009), media literacy promotes cross-cultural competence and comprehension focused on meaning rather than on form, as well as repeated exposure to L2 cultural products, practices, and perspectives, and the target language itself.

(Oxana Dema & Aleidine Kramer Moeller. Teaching culture in the 21st century language classroom. Adaptado.)
Considering the author’s point of view on the use of digital media for cultural and intercultural purposes, which of the following activities best exemplifies an enhanced use of films and digital media in the foreign language classroom?
Alternativas
Q3781897 Inglês
Leia o texto a seguir para responder a questão:

        The more traditional methods and approaches to teaching culture, such as movies and video, can be enhanced through the integration of digital media. Feature films have become readily available and have been included in numerous textbooks and designed to actively involve the learner (Aparisi, Blanco, & Rinka, 2007; Blanco & Tocaimaza-Hatch, 2007). Foreign language instructors are beginning to incorporate more movies in the foreign language classroom as “an accessible window” (Bueno, 2009, p. 319) to the target culture through “combined effects of images, sounds, camera, plots and dialogue” (Stephens, 2001, p. 2). According to Bueno (2009), media literacy promotes cross-cultural competence and comprehension focused on meaning rather than on form, as well as repeated exposure to L2 cultural products, practices, and perspectives, and the target language itself.

(Oxana Dema & Aleidine Kramer Moeller. Teaching culture in the 21st century language classroom. Adaptado.)
According to the author, the pedagogical value of incorporating films and digital media into foreign language instruction is
Alternativas
Q3780407 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
The text concludes that the Victorian engineers’ decision 
Alternativas
Q3780406 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
The metaphor used in avoid handing it the keys to the kingdom (7th paragraph) means avoid giving the system
Alternativas
Q3780405 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
Introducing in by introducing safety margins (6th paragraph) is similar in meaning to 
Alternativas
Q3780404 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
The phrase shoddy work costs lives (3rd paragraph) refers to work that is 
Alternativas
Q3780403 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
The word tricked (1st paragraph) means that LLMs can be
Alternativas
Q3780402 Inglês
Read the following text and answer the questions.


Artificial Intelligence: The “lethal trifecta”

    LARGE LANGUAGE MODELS (LLMs), a trendy way of building artificial intelligence, have an inherent security problem: they cannot separate code from data. As a result, they are at risk of a type of attack called a prompt injection, in which they are tricked into following commands they should not. Sometimes the result is merely embarrassing, as when a customer-help agent is persuaded to talk like a pirate. On other occasions, it is far more damaging.

    The worst effects of this flaw are reserved for those who create what is known as the “lethal trifecta”. If a company, eager to offer a powerful AI assistant to its employees, gives an LLM access to untrusted data, the ability to read valuable secrets and the ability to communicate with the outside world at the same time, then trouble is sure to follow. And avoiding this is not just a matter for AI engineers. Ordinary users, too, need to learn how to use AI safely, because installing the wrong combination of apps can generate the trifecta accidentally. 

   Better AI engineering is, though, the first line of defence. And that means AI engineers need to start thinking like engineers, who build things like bridges and therefore know that shoddy work costs lives.

  The great works of Victorian England were erected by engineers who could not be sure of the properties of the materials they were using. In particular, whether by incompetence or malfeasance, the iron of the period was often not up to snuff. As a consequence, engineers erred on the side of caution, overbuilding to incorporate redundancy into their creations. The result was a series of centuries-spanning masterpieces.

   AI-security providers do not think like this. Conventional coding is a deterministic practice. Security vulnerabilities are seen as errors to be fixed, and when fixed, they go away. AI engineers, inculcated in this way of thinking from their schooldays, therefore often act as if problems can be solved just with more training data and more astute system prompts.

   These do, indeed, reduce risk. The cleverest frontier models are better at spotting and refusing malicious requests than their older or smaller cousins. But they cannot eliminate risk altogether. Unlike most software, LLMs are probabilistic. Their output is driven by random selection from likely responses. A deterministic approach to safety is thus inadequate. A better way forward is to copy engineers in the physical world and learn to work with, rather than against, capricious systems that can never be guaranteed to function as they should. That means becoming happier dealing with unpredictability by introducing safety margins, risk tolerance and error rates.

   Overbuilding in the AI age might, for instance, mean using a more powerful model than is needed for the task at hand, to reduce the risk it will be tricked into doing something inappropriate. It might mean imposing limits on the number of queries LLMs can take from external sources, calibrated to the risk of damage from a malicious query. And mechanical engineering emphasises failing safely. If an AI system must have access to secrets, then avoid handing it the keys to the kingdom.

   In the physical world, bridges have weight limits – even if they are not always stated clearly to drivers. And, importantly, these are well within the actual tolerances that calculations suggest a bridge will bear. The time has now come for the virtual world of AI systems to be similarly equipped.

Adapted from The Economist, September 27th, 2025, p. 10
By referring to LLMs as a trendy way of building artificial intelligence (1st paragraph), the author implies they are 
Alternativas
Respostas
2261: B
2262: D
2263: B
2264: A
2265: B
2266: D
2267: E
2268: C
2269: C
2270: E
2271: D
2272: A
2273: B
2274: C
2275: B
2276: A
2277: C
2278: E
2279: D
2280: E