IELTS Academic Reading Practice Test 186

For most of the nineteenth century, keeping a carpet clean was heavy work. In a prosperous household the carpet was regularly taken outside, hung over a line and beaten with a cane, raising clouds of dust. Indoors, servants swept with brooms and hand brushes, and first scattered damp tea leaves over the surface so that fine dust would cling to them rather than rise into the air. Soot from coal fires settled on everything, and a carpet acted as a reservoir for it. What was needed was a machine that could remove dust without simply moving it about.

Patents for carpet sweepers worked by bellows, which blew or sucked air through the fabric, were filed in the United States in the 1850s and 1860s, but none was practical. The first machine to be manufactured in any numbers was designed by Ives W. McGaffey, a Chicago instrument maker, and patented in 1869 under the name Whirlwind. A fan inside a metal case was driven by a hand crank, and the resulting suction drew dust into a cloth bag. One hand was needed to turn the crank and the other to push the machine, so the operator had to stoop over it. The Whirlwind was expensive and tiring, and few households bought one.

Hubert Cecil Booth, a British engineer whose firm built bridges and fitted out ships, took the idea of suction seriously. In 1901 he watched a demonstration of an American device that blew compressed air into a cushion to drive out the dust inside it, with the result that the dust settled on everything in the room. Booth is said to have concluded that air should be drawn through cloth rather than blown, and he tested the notion by placing a handkerchief over his mouth, pressing his lips to the upholstery of a chair and inhaling.

Booth's machine was powerful and enormous. An internal-combustion engine drove a pump, and the whole apparatus was mounted on a horse-drawn van, far too heavy to bring indoors. A flexible hose was passed through a window and the operator worked from inside the room. Booth did not sell his machines; he sold the service of cleaning, and his uniformed operators visited houses, shops and theatres on request. In 1902 his equipment was used at Westminster Abbey before the coronation of Edward VII, an event that brought the invention considerable publicity.

Portable electric cleaners appeared a few years later. The decisive figure was James Murray Spangler, a janitor in Canton, Ohio, who suffered from asthma and found the dust raised by sweeping unbearable. In 1907 he assembled a working machine from a fan motor, a soap box and a pillowcase, and patented it the following year. Unable to manufacture it at a profit himself, he sold the patent to William Henry Hoover, a leather-goods manufacturer whose wife had already bought one of the cleaners.

Hoover's company transformed the product. Its first electric cleaners, sold from 1908, were upright machines with a cloth bag hanging from the handle, and later models added a rotating brush that loosened dirt as they passed over the carpet. The upright design suited the fitted carpets becoming common in American homes, and the cleaner turned into a symbol of the modern household. In Europe the preference was for cylinder machines, pulled behind the operator on runners.

Between the wars the cleaner spread from wealthy households to modest ones as electricity supplies reached new districts and prices fell. After 1945, wall-to-wall carpeting, which cannot be taken outside and beaten, made the machine almost indispensable, and manufacturers competed on suction and on filtration.

Filtration turned out to matter as much as suction. The dust a cleaner collects is held in a bag or container, but the air that passes through it carries the finest particles back into the room, and those are the particles most likely to be inhaled. Filters capable of trapping them were developed for industrial and military use in the middle of the twentieth century and were later adapted for domestic machines. Modern cleaners are rated by the efficiency with which they retain particles as well as by the power of the motor.

The next upheaval came from a British designer, James Dyson, who noticed in 1978 that the bag of his household cleaner clogged with fine dust, so that suction fell sharply within minutes of switching on. Manufacturers, he argued, had no incentive to solve the problem, since replacement bags were a steady source of profit. Dyson borrowed a principle from sawmills, where tall conical chambers spin dusty air so that particles are thrown to the wall. He scaled the cyclone down, and after thousands of prototypes launched a bagless cleaner in Britain in 1993.

Two further changes followed. In 1979 a power-tool company introduced a rechargeable handheld cleaner, and better batteries later made possible the cordless stick model. In 2002 the first mass-market robot cleaner went on sale, moving across the floor on its own and changing direction when it met an obstacle; later versions built maps of rooms with cameras and laser rangefinders and returned to a docking station to recharge.

A cleaner does not pull dirt; it creates a region of low pressure, and the surrounding air, rushing to fill it, carries the dirt with it. Suction is quoted in air watts, a figure that combines the rate of airflow with the pressure difference that produces it.

The retina is a thin layer of tissue at the back of the eye, containing some 120 million rods and six million cones, the photoreceptors that convert light into electrical signals, together with several layers of nerve cells that process those signals before passing them along the optic nerve. In retinitis pigmentosa, an inherited condition, the photoreceptors gradually die, starting at the edges of the visual field, and many patients eventually lose most of their sight. In age-related macular degeneration, the leading cause of sight loss among older people, the damage begins at the centre. In both diseases, however, one fact has encouraged engineers: the nerve cells that carry signals onward to the brain often survive long after the rods and cones have gone.

An implant exploits that survival. A tiny array of electrodes is placed against the retina so that it can deliver small pulses of current to the surviving cells, which respond much as they would to light: the patient sees spots of brightness, known as phosphenes, arranged in a grid. Because the array replaces the photoreceptors rather than the whole visual pathway, a device of this kind can only help someone whose retina is damaged but whose optic nerve and visual cortex are intact. It also works best in patients who have seen normally at some point in their lives, since the brain must already have learned to interpret signals from the eye.

Most devices tested in patients so far are epiretinal: the array is fixed to the inner surface of the retina, facing the cells that will be stimulated. The patient wears spectacles carrying a small video camera. A processor, worn on a belt or clipped to the glasses, converts the image into a pattern of stimulation, and a coil behind the ear transmits power and data through the skin to a receiver attached to the eye. The best-known example, Argus II, carried sixty electrodes in a grid and received approval for use in Europe in 2011 and in the United States in 2013. The company that developed it later turned its attention to stimulating the brain directly instead.

What patients actually experience is limited but not negligible. With sixty electrodes, the image resembles a grid of lights, roughly the effect of looking at a large outdoor scoreboard from a distance. Users report that they can locate doorways, follow the edge of a pavement, distinguish a plate from a cup on a table and, in some cases, read very large letters. They cannot recognise faces or appreciate colour, and several months of training are needed before the patterns become useful. In trials, scores on tests of everyday visual tasks improved modestly but consistently, and some participants reported greater confidence when moving about independently.

A different design places the array beneath the retina, where the photoreceptors used to be. Because the electrodes then sit in the position the dead cells once occupied, the optical system of the eye itself focuses the image, so no camera is needed and the wearer can look around naturally. A German group developed a subretinal device containing about 1,500 photodiodes, each connected to its own amplifier and electrode and powered through a cable led out of the eye to a unit implanted behind the ear. Because every electrode works independently, the resolution is potentially finer than that of an epiretinal grid, but the surgery is more delicate and the implant more exposed to damage.

Engineering, rather than biology, sets the present limits. A healthy eye has more than a hundred million photoreceptors, while the best implants offer a few thousand electrodes at most, and current spreads through the tissue so that neighbouring cells are stimulated together, blurring the image further. The array must also be sealed against the salty fluid inside the eye for many years, must not grow hot enough to cook the tissue around it, and must fit into a space only a fraction of a millimetre deep. Each of these requirements pulls against the others, and progress has been slow.

Candidates are selected with care. Implants are offered to adults who have lost useful vision but who retain a functioning optic nerve, and who are willing to undergo surgery and months of rehabilitation. People blind from birth are generally excluded, because the parts of the brain that handle vision require visual experience in early childhood to develop; in someone who has never seen, those areas respond to sound and touch instead. Counselling before surgery is important, since patients who expect the devices to restore normal sight are likely to be disappointed. The devices are also expensive, and only a small number of centres have the surgical teams and follow-up services to support them.

Research now runs along several lines. One is optogenetics, in which a gene carrying the instructions for a light-sensitive protein is delivered to the surviving retinal cells, making them respond directly to light; because the light must be strong and of the right wavelength, patients wear goggles that project the image in a suitable colour. Another approach uses photovoltaic pixels that generate their own current when illuminated, removing the need for a power supply. Biologists, meanwhile, are trying to transplant photoreceptors grown from stem cells, which could in principle restore the natural pathway rather than bypass it, and early trials have begun. A fourth route abandons the eye altogether and stimulates the visual cortex directly, which would in theory help people whose optic nerve has been destroyed.

At a growing number of airports the officer in the glass booth has been replaced by a camera on a pillar. The traveller steps up, the machine photographs the face, compares it with the image stored in the chip of the passport and, if the two match closely enough, opens a gate. The whole encounter may last five seconds. For the airport it means fewer staff and faster processing. The question that concerns me is not whether the technology works, since it plainly does much of the time, but what kinds of decision we are willing to hand to it.

We should begin by distinguishing two operations that are often confused. Verification asks a narrow question: is this person the holder of this document? One face is compared with one stored template, and the traveller has usually consented by choosing to use the machine. Identification asks a far broader question: who is this person? A face is compared with every face in a database or a watchlist, and the person may know nothing about it. The first is a convenience; the second is a search. Almost all the reassuring statistics about accuracy are produced in the first setting, while the risks that worry critics belong mainly to the second.

The engineering behind these systems is impressive and has improved rapidly. An ordinary photograph is a poor basis for matching, since it changes with lighting, angle and expression, so most gates illuminate the face with near-infrared light and record its three-dimensional shape as well as its surface pattern. The result is not a picture stored for later inspection but a mathematical template, a set of measurements of features such as the distance between the eyes. Independent testing organised by the United States National Institute of Standards and Technology has shown that the accuracy of the best algorithms improved dramatically during the 2010s.

Yet accuracy is not uniform. The same American testing programme reported in 2019 that many algorithms made more false matches for some demographic groups than for others, with the highest rates typically found among certain Asian and African faces and among women. The cause is not mysterious: systems trained on photographs that over-represent one section of the population will perform less well on the rest. A false match at a gate may mean that a traveller is pulled aside and questioned, while a false rejection means that everyone whose face the system misreads pays in delay. Neither error is catastrophic in itself; both fall most heavily on particular groups.

Every system must also choose where to set its threshold, and the choice cannot be avoided by better engineering. Set the bar high and the machine rejects genuine travellers, clogging the queue with people who must be examined by hand. Set it low and it waves through those it should have stopped. Because the two errors cost different things, and because the cost of a missed threat is felt far more acutely than the cost of an unnecessary delay, operators have a standing temptation to tighten the system and let the queue absorb the consequences.

Privacy is usually presented as the opposing consideration, but the argument is more specific than a clash of abstractions. A password can be changed; a face cannot. Biometric data are therefore unlike any other personal information, and the rules governing them matter. In Europe, data protection law treats biometrics as a special category requiring explicit justification, and several countries have restricted the retention of images collected at borders. The greater danger, in my view, is not the photograph taken at the gate but the database it may be checked against, and the ease with which a system built for one purpose can be turned to another.

Nor should we assume that the technology is beyond challenge. Faces can be copied. Masks, printed photographs and video played on a phone have all been used to fool recognition systems, and the industry has responded with liveness checks that look for subtle movements, skin texture and depth. Each countermeasure invites a countermeasure of its own, and the arrival of convincing synthetic video raises the stakes. A gate that can be opened with a good imitation is worse than no gate at all, because it creates an appearance of control that is not really there.

Finally, there is the human being who is supposed to be in charge. When a machine produces a match, the officer who reviews it tends to accept it; psychologists call this automation bias, and it is well documented in aviation and medicine as well as in border control. A system that generates many alerts, most of them false, will eventually be ignored, while one that generates few will be trusted too readily. Accountability is therefore not a matter of adding a person to the process; it requires that the person has the time, the information and the authority to disagree.

My own view is that the distinction drawn earlier should govern policy. Verification against a passport, offered as an option and designed to be abandoned when it fails, is a sensible use of the technology: it is quick, it is consented to and its errors are borne by the system rather than by the traveller. Identification from a watchlist, conducted without the knowledge of the person concerned, deserves far more scepticism, independent auditing and a strict limit on how long the data are kept. Framed as a contest between security and privacy, the debate goes nowhere. Framed as a question about which decisions may be automated, and on whose evidence, it becomes answerable.