Đề thi thử số 3 — Reading 40 câu (có bài đọc)

Public transportation systems are essential to the functioning of modern cities, providing millions of people with affordable and efficient means of getting to work, school, and other destinations. From buses and trams to underground railways and ferries, public transit networks reduce traffic congestion, lower carbon emissions, and improve the overall quality of urban life. Despite these benefits, many cities around the world struggle to maintain and expand their public transportation infrastructure to meet the needs of growing populations.

One of the most important advantages of public transportation is its ability to reduce traffic congestion. In cities where a large percentage of the population relies on private cars, roads become clogged during peak hours, leading to long commute times, increased fuel consumption, and higher levels of air pollution. By contrast, a single bus can carry the same number of passengers as 40 or more cars, while a metro train can transport hundreds of people at once. Cities that have invested in efficient public transit systems, such as Tokyo, Singapore, and Vienna, consistently rank among the most livable cities in the world, partly because their residents spend less time stuck in traffic.

Public transportation also plays a crucial role in promoting social equity. For many low-income individuals and families, owning and maintaining a private vehicle is prohibitively expensive. Public transit provides these communities with essential access to employment, healthcare, education, and social services. Without reliable bus or train services, residents of underserved neighborhoods may find themselves isolated from economic opportunities, perpetuating cycles of poverty. Recognizing this, some cities have introduced reduced-fare programs for students, seniors, and low-income residents to ensure that public transportation remains accessible to all members of society.

The environmental benefits of public transportation are substantial and well-documented. According to the American Public Transportation Association, public transit produces 95 percent less carbon monoxide and nearly 50 percent less carbon dioxide per passenger mile compared to private automobiles. Electric buses and trains, which are becoming increasingly common in cities worldwide, further reduce the environmental footprint of public transit by eliminating tailpipe emissions entirely. Additionally, cities with robust public transit networks tend to have more compact urban development patterns, which preserve green spaces and reduce the environmental impact of suburban sprawl.

However, public transportation systems face numerous challenges that can undermine their effectiveness. Insufficient funding is perhaps the most common obstacle, as building and maintaining transit infrastructure requires significant capital investment. Many cities struggle to balance the need for affordable fares with the costs of operating and expanding their networks. Overcrowding during peak hours, aging infrastructure, and safety concerns can also discourage potential riders from using public transit. To address these challenges, transportation planners are increasingly turning to technology solutions such as real-time tracking apps, contactless payment systems, and data-driven route optimization to improve the passenger experience and attract more riders.

The rapid expansion of online education has transformed the landscape of learning at every level, from primary school to postgraduate studies. What was once considered a niche alternative to traditional classroom instruction has become a mainstream educational delivery method, accelerated dramatically by the global disruptions caused by the COVID-19 pandemic. Today, millions of students worldwide access courses, degree programs, and professional certifications through digital platforms, raising important questions about the quality, accessibility, and future direction of education.

Massive Open Online Courses, commonly known as MOOCs, have been at the forefront of the online education revolution. Platforms such as Coursera, edX, and Udemy offer thousands of courses from prestigious universities and industry experts, often at a fraction of the cost of traditional enrollment or even for free. A student in rural Vietnam can now access lectures from professors at MIT, Harvard, or Oxford without leaving home, a possibility that would have been unimaginable just two decades ago. This democratization of knowledge has the potential to reduce educational inequality and provide learning opportunities to populations that have historically been excluded from elite academic institutions.

Online education offers several practical advantages that traditional classrooms cannot easily replicate. Flexibility is perhaps the most frequently cited benefit, as students can access course materials, watch recorded lectures, and complete assignments according to their own schedules. This flexibility is particularly valuable for working professionals who wish to advance their careers through continuing education without sacrificing their employment. Additionally, online platforms can leverage technology to personalize the learning experience, using algorithms to identify areas where individual students struggle and recommending supplementary materials or exercises to address specific weaknesses.

Despite these advantages, online education faces legitimate criticisms that must be addressed if it is to fulfill its transformative potential. The absence of face-to-face interaction between students and instructors can diminish the quality of learning, particularly for subjects that benefit from hands-on practice, laboratory work, or collaborative discussion. Completion rates for online courses remain significantly lower than those of traditional programs, with some MOOCs reporting completion rates below ten percent. This attrition is often attributed to the lack of structured accountability and social engagement that physical classrooms naturally provide. Furthermore, the digital divide means that students without reliable internet access or suitable devices are effectively excluded from online learning opportunities, potentially widening rather than narrowing educational inequalities.

The future of education will likely involve a hybrid model that combines the best elements of online and traditional instruction. Blended learning approaches, in which students attend some sessions in person while completing other components online, have shown promising results in research studies. Universities are increasingly investing in sophisticated learning management systems that facilitate interaction between students and instructors through video conferencing, discussion forums, and collaborative digital workspaces. As technology continues to evolve, innovations such as virtual reality classrooms, AI-powered tutoring systems, and blockchain-based credentialing may further blur the boundaries between online and offline education, creating a more flexible and inclusive learning ecosystem for students worldwide.

Plastic pollution has emerged as one of the most visible and pervasive environmental crises of the twenty-first century, with the world's oceans bearing a disproportionate share of the burden. An estimated eight million metric tons of plastic waste enter the oceans every year, accumulating in vast garbage patches, washing up on remote beaches, and infiltrating the deepest ocean trenches. This relentless influx of synthetic materials threatens marine ecosystems, compromises human health through the food chain, and poses fundamental questions about the sustainability of current production and consumption patterns.

The scale of oceanic plastic pollution is staggering. The Great Pacific Garbage Patch, the largest accumulation of ocean plastic in the world, covers an area approximately three times the size of France and contains an estimated 80,000 metric tons of floating debris. Contrary to popular imagination, however, this patch is not a solid island of trash but rather a diffuse concentration of microplastics — fragments smaller than five millimeters — suspended in the water column. These microplastics are virtually impossible to remove with current technology and persist in the marine environment for hundreds of years, slowly breaking down into ever smaller particles but never fully biodegrading.

The ecological consequences of plastic pollution are devastating and far-reaching. More than 800 marine species are known to be affected by plastic debris through ingestion, entanglement, or habitat disruption. Sea turtles frequently mistake plastic bags for jellyfish, their primary food source, leading to intestinal blockages and starvation. Seabirds ingest small plastic fragments, which accumulate in their stomachs and can cause malnutrition, internal injuries, and death. Perhaps most concerning is the phenomenon of bioaccumulation, in which toxic chemicals associated with plastics — including bisphenol A, phthalates, and persistent organic pollutants — concentrate as they move up the food chain, ultimately reaching human consumers through contaminated seafood.

Addressing oceanic plastic pollution requires intervention at multiple points along the plastic lifecycle. Reducing plastic production, particularly of single-use items such as bags, bottles, straws, and packaging, represents the most effective upstream approach. More than 60 countries have implemented bans or levies on single-use plastic bags, with measurable reductions in plastic litter observed in many cases. However, production-side interventions alone are insufficient without corresponding improvements in waste management infrastructure, especially in developing nations where inadequate collection and recycling systems allow significant quantities of plastic to leak into waterways. International agreements such as the United Nations Global Plastics Treaty, currently under negotiation, aim to establish binding commitments for reducing plastic pollution across the entire value chain.

Innovative technologies offer promising but partial solutions to the crisis. Ocean cleanup systems, such as the floating barriers developed by The Ocean Cleanup organization, have demonstrated the feasibility of capturing surface-level plastic debris in certain conditions. Biodegradable alternatives to conventional plastics, including materials derived from seaweed, cornstarch, and bacterial cellulose, are gaining commercial traction but remain more expensive and less versatile than their petroleum-based counterparts. Advanced recycling technologies, including chemical recycling processes that can break polymers down into their original monomers for reprocessing, represent a significant improvement over traditional mechanical recycling but have yet to achieve the scale necessary to meaningfully impact global plastic waste volumes. Ultimately, scientists and policymakers agree that no single technological solution can substitute for fundamental changes in how societies produce, consume, and dispose of plastic materials.

The rapid advancement of artificial intelligence technologies has precipitated an urgent and multifaceted discourse on the ethical dimensions of developing and deploying systems capable of autonomous decision-making. As AI applications proliferate across domains as diverse as healthcare diagnostics, criminal justice, autonomous vehicles, financial services, and military operations, the decisions made by algorithmic systems increasingly carry profound consequences for individual lives and societal structures. The absence of comprehensive regulatory frameworks governing AI development has intensified calls from researchers, ethicists, and policymakers for the establishment of robust principles to guide the responsible creation and deployment of intelligent systems.

Algorithmic bias represents one of the most extensively documented ethical challenges in contemporary AI development. Machine learning systems are trained on historical datasets that frequently reflect and encode existing social inequalities related to race, gender, socioeconomic status, and other demographic characteristics. When these biased datasets are used to train predictive models, the resulting systems can perpetuate and even amplify discriminatory patterns in consequential decision-making contexts. A widely publicized analysis of a major healthcare algorithm used by hospitals across the United States revealed that the system systematically underestimated the healthcare needs of Black patients compared to equally ill White patients, a bias traceable to the algorithm's use of healthcare spending rather than actual illness as a proxy for medical need.

The opacity of many advanced AI systems poses a distinct but related ethical challenge. Deep learning models, which power some of the most capable AI applications, operate through complex neural network architectures involving millions or even billions of parameters whose individual contributions to outputs are exceedingly difficult to interpret. This opacity — often characterized as the "black box" problem — undermines the principle of accountability, as it becomes unclear who bears responsibility when an opaque algorithm produces a harmful or discriminatory outcome. The tension between model performance and interpretability presents a genuine dilemma for practitioners, as the most accurate models are frequently the least transparent, while more interpretable approaches may sacrifice predictive power.

Privacy erosion constitutes another critical ethical concern as AI systems demand ever larger quantities of personal data to function effectively. Facial recognition technology, which has achieved remarkable accuracy through training on vast databases of human faces, exemplifies this tension between capability and privacy. While proponents argue that facial recognition enhances public safety through improved surveillance and identification of criminal suspects, critics contend that the technology enables unprecedented mass surveillance, disproportionately misidentifies individuals from minority ethnic groups, and creates an architecture of control that fundamentally alters the relationship between citizens and the state. Several major cities, including San Francisco and Brussels, have enacted moratoriums or outright bans on government use of facial recognition in public spaces, reflecting growing public unease with the technology's implications for civil liberties.

The international governance landscape for AI ethics remains fragmented and largely voluntary. The European Union's AI Act, adopted in 2024, represents the most comprehensive legislative framework to date, establishing risk-based categories for AI applications and imposing stringent requirements on systems classified as high-risk. However, the regulation's effectiveness depends on consistent enforcement across member states and its influence beyond European borders. China has pursued a parallel but distinct regulatory approach, emphasizing state oversight and alignment with national strategic objectives. The United States has largely relied on sector-specific guidelines and voluntary industry commitments rather than comprehensive legislation. This regulatory heterogeneity creates challenges for multinational technology companies operating across jurisdictions and raises concerns about a potential race to the bottom, in which companies relocate AI development activities to regions with less stringent oversight to gain competitive advantages.