IELTS Academic Reading Practice Test 131

On a bench in a laboratory in Rochester, New York, in December 1975, a young engineer named Steven Sasson aimed a strange device at a colleague. It weighed about 3.6 kilograms and contained a lens taken from a movie camera, several circuit boards and a cassette tape recorder. When Sasson pressed the button, the machine took twenty-three seconds to record a single black-and-white image, which it stored as a long string of numbers on the tape. The first digital camera had produced its first photograph.

The sensor at the heart of the machine had been invented six years earlier, in 1969, by two physicists at Bell Laboratories in New Jersey, Willard Boyle and George Smith. They called it a charge-coupled device, or CCD, and their original purpose was to store data, not to capture light. A CCD is a thin slice of silicon divided into a grid of tiny elements. When light falls on an element it releases electrons in proportion to the brightness, and the resulting packets of charge can be shifted from one element to the next and read out at the edge of the chip. Boyle and Smith shared the Nobel Prize in Physics for the work in 2009.

Sasson's prototype was a demonstration of principle rather than a product. Its images measured a hundred pixels by a hundred, which is to say ten thousand pixels in total, or a hundredth of a megapixel; a single frame from a modern phone holds more than a thousand times as much detail. The tape allowed only thirty images to be recorded, and viewing them required a playback unit connected to a television set. Kodak's senior managers were impressed but cautious. The company earned most of its money from film, photographic paper and the chemicals that developing required, and executives were uncertain how a camera without film could be sold at a profit.

Other companies were experimenting too. In 1981 Sony announced the Mavica, a camera that looked much like a conventional single-lens reflex model but recorded still images as analogue video signals on a small magnetic disk instead of on film. The pictures could be played back on a television, and the absence of any processing made them immediately useful to newspapers, which were beginning to transmit photographs electronically. The Mavica was nonetheless an analogue device of modest image quality, and most similar still-video cameras sold poorly.

True digital cameras emerged more slowly. In 1991 Kodak produced the DCS-100, a professional model built around a Nikon F3 film body with a digital back containing a 1.3-megapixel sensor. It was priced far beyond the reach of ordinary photographers and was bought mainly by newspapers and studios. The pictures it produced were far more detailed than anything that had gone before, and they could be sent down a telephone line, a decisive advantage for news agencies.

The first models intended for ordinary households appeared in the mid-1990s. Apple introduced the QuickTake 100 in 1994; it was made for the company by Kodak, recorded 640 by 480 pixels and cost less than a thousand dollars. A year later the Japanese firm Casio added a colour liquid-crystal screen to the back of a similar camera, so that photographers could see each picture as soon as it was taken and decide whether to keep it. The screen changed how people thought about photographs: they became something to be examined and deleted rather than developed.

Meanwhile the number of pixels a sensor could hold, and so the detail in the image, rose steadily. Nikon's D1 of 1999, with 2.7 megapixels, was the first digital single-lens reflex camera to attract a wide following among press photographers. In 2003 Canon went further, selling an interchangeable-lens model with a resolution of six megapixels for less than a thousand dollars.

The other great change concerned the sensor itself. CCDs were delicate and expensive to produce, because each one had to be made on a specialised manufacturing line. A different design, the complementary metal-oxide-semiconductor sensor, or CMOS, had been known since the 1960s but was long thought unsuitable for photography. In a CMOS sensor every pixel contains its own amplifier and can be read individually, so the chip can be made with the same equipment as ordinary computer chips; it is therefore cheaper to manufacture and uses less power. Early CMOS sensors produced noisier images than CCDs, and several years of engineering were needed before the gap was closed. By the middle of the first decade of this century most new cameras contained CMOS sensors.

The final stage was the disappearance of the camera into another device. In 2000 Sharp began selling a mobile phone in Japan with a small camera built in; the pictures were poor, but the idea spread with extraordinary speed. Within a decade most phones sold in wealthy countries contained a camera, and the number of photographs taken around the world each year had risen enormously.

The consequences for the industry were severe. Kodak, which had dominated photography for a century, saw sales of film and paper fall year after year, and in January 2012 the company sought protection from its creditors in American courts. Its own engineers had built the first digital camera, yet it proved unable to turn that early lead into a business. Sasson continued to work for Kodak for many years, and the camera he built in 1975 is now kept in a museum collection.

Farmers have been changing the genetic makeup of crops for thousands of years, though for most of that time they did so without knowing what genes were. Since the middle of the twentieth century, plant breeders have also used mutation breeding, exposing seeds to radiation or to harsh chemicals in order to create random changes in their DNA and then searching among the damaged plants for useful ones. The process is, in effect, accelerated and undirected evolution — worth remembering when gene editing is described as unnatural.

The tools that make editing possible come from bacteria. Many bacteria store fragments of the DNA of viruses that have attacked them and use these fragments as a kind of memory, known as CRISPR, to recognise the same viruses if they return. When a match is found, the bacterium produces an enzyme, Cas9, which cuts the invading DNA and destroys it. In 2012 a team led by Emmanuelle Charpentier and Jennifer Doudna showed that this system could be reprogrammed to cut almost any sequence of DNA chosen by the researcher. The two scientists shared the Nobel Prize in Chemistry in 2020.

In practice the researcher supplies a short piece of RNA whose sequence matches the target gene, and this guide molecule leads the Cas9 enzyme to the right place in the genome. The enzyme then cuts both strands of the DNA. The cell repairs the break clumsily, usually leaving the gene unable to function, which is what happens when the aim is to switch a gene off. To give a plant a new characteristic rather than take one away, a second piece of DNA carrying the desired sequence can be supplied, and the cell may use it as a template when repairing the cut.

Editing is often confused with the genetic modification that caused such controversy in the 1990s. The two are not the same. Older techniques inserted genes from unrelated organisms — a gene from a bacterium, for instance, to make a crop tolerate a particular weedkiller — so the resulting plant contained DNA that no conventional breeding could have produced. Editing can change a plant's own genes, and the alteration is often one that could have arisen by accident in a field. That distinction is the basis of the argument that edited crops deserve lighter regulation, though critics reply that the two methods serve the same commercial purposes.

Regulators have struggled to decide how edited plants should be treated. In the United States, the Department of Agriculture announced in 2016 that it would not regulate a gene-edited mushroom designed not to turn brown, on the grounds that it contained no foreign DNA. Other countries took the opposite view. In 2018 the Court of Justice of the European Union ruled that crops produced by newer editing methods fell within the Union's legislation on genetically modified organisms, and must therefore be labelled and tested like any other GM crop.

Several edited products have already reached the market or the field. A soybean whose oil is unusually high in oleic acid, and therefore more stable when heated, has been grown and sold in the United States. Resistance to disease is the area where the gains could be greatest, because it offers farmers an alternative to spraying. Researchers in China have disabled three copies of a single wheat gene and produced plants that resist powdery mildew, and similar work is being done on bananas threatened by a soil fungus.

The editing machinery sometimes alters DNA at places other than the intended target, and such off-target edits have to be checked for. More importantly, the technique depends on the ability to grow a complete plant from a single modified cell, and this works well in only a few species, such as tomato, rice and maize, while many other crops remain difficult. The traits that matter most to farmers and to consumers — tolerance of drought, resistance to heat, higher yield — are controlled by dozens or hundreds of genes, each with a small effect, and these are not promising targets for a tool that works best one gene at a time.

The tools themselves continue to change. The first generation of editors cut both strands of DNA and left the cell to mend the damage: good at disabling a gene, poor at correcting one. Base editors, developed at Harvard University in 2016, avoid the double cut. A modified Cas9 nicks a single strand, and an enzyme attached to it chemically converts one DNA letter into another — a C into a T, for example — without breaking the helix. Prime editing, described in 2019, goes further: a special guide molecule both finds the target and carries the sequence that is to be written, so that short stretches of DNA can be replaced rather than merely disabled. Researchers are also trying to avoid the slow step of growing plants from single cells. One approach is to fire DNA-coated particles into pollen; another is to use modified viruses.

Gene editing, then, is a powerful addition to the breeder's toolkit rather than a replacement for it. It makes some improvements quicker and more precise, but it does not make plant biology any simpler, and the characteristics that matter most to farmers remain as stubborn as ever. The question that will decide how widely the technology is used is not whether it works, but whether the public and its regulators can be persuaded that it is safe.

When large language models became freely available at the end of 2022, teachers discovered within weeks that essays could be produced on demand. Some schools banned the technology outright. Both responses missed the point. Students will use these tools, just as they use calculators, search engines and translation software, and pretending otherwise wastes time that could be spent deciding what we actually want them to learn. My concern is not that machines will do students' work for them. It is that we will accept the appearance of learning in place of the substance.

The strongest argument for these machines is that they might supply something schools have never been able to afford: individual tutoring. In 1984 the educational psychologist Benjamin Bloom reported that students taught one-to-one by a tutor, and then tested, outperformed students taught in a conventional class by about two standard deviations — a difference so large that the average tutored student would out-perform 98 per cent of a conventionally taught class. Bloom called this the two sigma problem. A patient machine that never loses its temper and adapts to each learner looks, on the face of it, like an answer.

We should be sceptical. Programmes that attempt to teach without a human being have existed since the 1960s. Reviews of this work generally find real but modest gains, largest in tightly defined subjects such as algebra and smallest where the task demands extended writing or the weighing of evidence. Gains also tend to shrink when the programmes are used at scale, outside the classrooms of enthusiastic developers. That is not an argument for abandoning the attempt, but it is an argument for not announcing a revolution before the evidence has arrived.

The panic about cheating is the least interesting part of the story. If an essay question can be answered by a machine in thirty seconds, the question was probably testing recall and fluency rather than understanding, and it deserved to be exposed. Schools that care about learning might ask students to explain their reasoning aloud, to revise a piece of work in front of a teacher, or to solve a problem they have not seen before.

The second promise is duller and more convincing. Teachers spend a great deal of time on tasks that are necessary but not educational: writing comments on work, drafting reports, planning lessons, composing letters to parents. Software can already do much of this, and where it does, teachers may gain hours that can be returned to teaching. The risk is that the hours are not returned but filled: that a teacher ends up checking and correcting the output of a machine, which can take longer than doing the job in the first place. Whether the promise is kept depends less on the software than on how schools choose to use the time it saves.

There are also dangers that the enthusiasts rarely mention. These systems produce confident statements that are simply wrong, and a tutor that is wrong some of the time is worse than a textbook, because a textbook can be checked. They collect detailed information about children — what each one finds easy, where each one gives up, how long each one hesitates — and that information is extremely valuable to companies. Schools that adopt such systems are not simply buying a product; they are handing over a record of their pupils' difficulties to a business whose interests may not be the same as theirs.

Then there is fairness. Schools in wealthy areas have always bought advantages for their pupils, and they will buy these ones. What worries me more is the opposite case: schools serving poor communities may be offered the technology cheaply, or free of charge, precisely because there is no money for extra teachers. A child in a well-funded school will be taught by a human being, supported by a machine; a child in an under-funded school may be taught by a machine alone. The rhetoric of personalisation can too easily become a justification for doing without people.

Finally, there is the matter of relationship. Learning is not only the transfer of information from a text to a mind; it happens between people. A teacher notices that a child is quiet, or hungry, or discouraged, and responds in a way that no system can plan for. Personalised too often means alone: a pupil working through exercises at a screen, with no one to argue with, no one to be embarrassed in front of, no one whose approval matters.

None of this is a case for keeping the machines out of schools. A tool that gives every pupil a patient tutor, and gives every teacher more time, would be a genuine advance, and the evidence may yet show that these systems can do both. What I am arguing for is scepticism, patience and a clear sense of priority. Schools should adopt these tools where they have been shown to help, insist on knowing what data they collect and who profits from it, and refuse to let the appearance of efficiency stand in for the education of children. The question is not whether machines can teach. It is what we still want teachers to do.