IELTS Academic Reading Practice Test 124

In the decades before plastic cards existed, buying goods on credit was an informal and local matter. A customer who was well known to a shopkeeper might be allowed to pay at the end of the month, with the arrangement recorded in the trader's own ledger, but a stranger had little chance of such a favour. From the mid-nineteenth century, makers of expensive household goods such as sewing machines and pianos offered instalment plans, under which a buyer paid a deposit and then a fixed sum each week or month. Borrowing of this kind was tied to a particular product, however, and could not be transferred to another purchase or another shop.

By the 1920s, oil companies and hotel chains in the United States were issuing small metal plates stamped with the customer's name and account number. Laid on a machine, the plate printed those details onto a sales slip, which the customer then signed. Department stores followed with cards of their own. These were charge cards rather than credit cards: the whole balance fell due at the end of each month, and the card was accepted only by the business that had issued it.

The idea of a card honoured by many different businesses is usually credited to Frank McNamara, a New York businessman who, according to his company's own account, had once been unable to pay a restaurant bill after leaving his wallet at home. In 1950 he and his partner Ralph Schneider launched Diners Club, which for an annual fee of three dollars allowed members to sign for meals at a small group of New York restaurants and settle the account later. The first cards were made of cardboard.

Diners Club spread quickly beyond New York, and rival schemes appeared for travel and entertainment. Its card, however, still demanded full payment each month. The decisive step from a charge card to a true credit card — one that let the holder repay part of the balance over time while interest accumulated on the rest — was taken by a bank.

In September 1958 the Bank of America, then the largest bank in the United States, mailed tens of thousands of unsolicited BankAmericard cards to households in Fresno, California. The card offered revolving credit, so a customer who did not pay in full carried the debt forward and was charged interest, and the bank also took a fee from the shops and restaurants that accepted it. The Fresno launch is often treated as the beginning of the modern credit card industry. The choice of location mattered: the city was a self-contained market in which the scheme could be tested. Sending live cards to people who had not requested them was later made illegal.

American Express, a company founded in 1850 to move freight and money across the United States, entered the market in the same year with a charge card aimed at travellers. Both businesses soon met the weakness of a system built on paper. Cards could be stolen or forged, and a waiter or shop assistant had no quick way of checking whether an account was in good standing, so losses from fraud were heavy.

The remedy was to store account details in a form that a machine could read. Engineers at IBM developed a method of coating plastic tape with particles of iron oxide and recording data on it magnetically, and the notion of fixing such a stripe to the back of a card is credited to Forrest Parry, an engineer with the company, at the start of the 1960s. The stripe was standardised during the 1970s, and the routine of swiping a card through a reader and waiting for authorisation down a telephone line became part of everyday life.

Because no single bank could run a national or international scheme alone, banks began to combine. A group of California banks formed the Interbank Card Association in 1966; its card was later named Master Charge and, in 1979, MasterCard. The BankAmericard network was reorganised by a banker named Dee Hock into a cooperative owned by the banks that issued the cards, and in 1976 it was given a name that would work in any language: Visa. In Britain the first credit card, Barclaycard, appeared in 1966.

The next change came from France. Roland Moreno, a French inventor, patented a card containing a memory chip in 1974; in 1977 Michel Ugon, an engineer at the French computer firm Bull, added a microprocessor, and French banks began issuing chip cards in the following decade. The chip held a secret key and could calculate a code that changed with every transaction, which made a card far harder to copy than a magnetic stripe. An international standard, known as EMV after the three organisations that devised it — Europay, MasterCard and Visa — was agreed in the 1990s, and the replacement of stripes by chips cut fraud substantially.

The most recent stage has removed the card itself from the transaction. Contactless payment, which exchanges data over a short-range radio link, spread widely in the 2010s, and many people now pay with a phone or a watch. The metal plate of the 1920s would be unrecognisable to its users, yet the underlying idea — a promise to pay, recorded and trusted — has scarcely changed. The questions that remain are equally old: who bears the cost when the promise is broken, and who pays for the rewards that cards offer.

Walk into a modern pharmaceutical laboratory and you may see something that looks more like a small factory than a bench crowded with glassware. Mechanical arms move trays of plastic plates from stack to stack, needles dip into wells and withdraw fractions of a millilitre of liquid, and a computer logs every action. The reason is arithmetic. A chemist testing a new compound may need to run the same reaction hundreds of times with tiny variations, and a person cannot do that reliably at speed, nor should anyone handle some of the substances involved at all.

Automation entered chemistry gradually. The microtitre plate, a rectangle of plastic divided into a grid of small wells, usually 96 of them, was devised in the 1950s and allowed many separate reactions to be run side by side. From the 1990s, pharmaceutical companies combined such plates with robotic liquid handlers to screen large collections of molecules for biological activity, testing hundreds of thousands of compounds in a fraction of the time a manual approach would take. At first these machines merely carried out instructions. What has changed recently is that they can increasingly decide what to do next.

The core component of an automated liquid handler is a pipette that moves in three dimensions. A disposable plastic tip is pressed onto the end of a narrow tube, a pump draws up a measured volume, and the tip is lowered into a well to release its contents. Disposable tips prevent one sample from contaminating another; where the same tip is reused, the machine washes it thoroughly between transfers. Accuracy is one advantage. A skilled technician can pipette very precisely, but the machine can do so several thousand times without fatigue or variation.

A more radical departure from the traditional laboratory is flow chemistry. Rather than stirring reagents in a flask, chemists pump them through narrow tubes, where they mix and react as they travel. Because the tubing is thin, heat is exchanged rapidly and temperature can be controlled far more tightly than in a large vessel; because only a small volume is reacting at any moment, dangerous intermediates can be generated and consumed immediately instead of accumulating. Reactions that would be unthinkable in a flask, such as those involving explosive or highly reactive substances, become manageable. Production is increased by pumping for longer rather than by building bigger vessels.

Flow systems lend themselves to automation, and combining them with machine learning has produced what researchers call the self-driving laboratory. The idea is a loop with three stages: the robot makes a batch of candidate materials, a second instrument measures how well each one performs, and an algorithm uses the results to choose the next set of experiments. The algorithm needs no deep understanding of chemistry; it simply searches for conditions that improve the measured outcome, balancing experiments that exploit promising leads with others that explore unfamiliar territory.

A striking demonstration came in 2020 from a team at the University of Liverpool led by Andrew Cooper. They built a mobile robot that could move between stations, operate equipment and run its own experiments, and set it to search for materials that speed up the production of hydrogen from water using light. The robot worked for eight days with only limited human intervention and ran hundreds of experiments in a space of possible mixtures that would have taken a human team months to cover.

The weak point of such systems is often the data rather than the hardware. Machine-learning models that predict how a reaction will behave are trained on results reported in the scientific literature, and that literature is heavily skewed towards experiments that worked. A model taught mainly on successful reactions will tend to be over-optimistic about new ones, which is a serious flaw when the purpose of an experiment is to find out whether an idea fails. Several research groups are therefore building databases that record unsuccessful experiments as carefully as successful ones.

Other groups have concentrated on making the instructions portable. A team at the University of Glasgow led by Lee Cronin has developed a system, nicknamed the chemputer, in which a synthesis is described by a chemical programming language and then executed by a robotic platform. The attraction is reproducibility: a recipe that works in one laboratory can, in principle, be run identically in another. The group has shown that the approach can make a range of drug-like molecules, though not yet the full variety that a skilled chemist can produce.

That caveat points to the limits of laboratory robotics. Most automated systems are designed around liquid reagents and flat plates, whereas much of synthetic chemistry involves solids that must be weighed, together with grinding, distillation and the purification steps that follow a reaction. Robots are also expensive, and they work best with standardised containers and protocols; a laboratory that buys machines from several manufacturers may find that they do not easily communicate with one another. A robot, moreover, will only explore the possibilities it has been programmed to consider.

None of this means that chemists are becoming redundant. The tasks that machines do well — repetitive measurement, systematic variation, recording everything — are exactly the tasks that consume a researcher's time and patience. Freed from them, chemists may spend more time on the questions that require judgement: which problem is worth attacking, and what an unexpected result might mean.

A house that unlocks its own door, warms a room before its owner arrives and answers questions from the kitchen table no longer sounds futuristic. Connected thermostats, lights, locks, cameras and speakers have become ordinary consumer goods, sold in the same shops as kettles and toasters. The convenience they offer is real, and it is easy to see why millions of households have adopted them. Yet the debate about these devices is usually framed as a matter of personal choice: if you are worried about privacy, the argument runs, you need not buy one. That framing, I shall argue, is mistaken.

The first thing to understand is how much a smart home senses. A thermostat records temperature, humidity and, through its motion sensor, whether anybody is at home. A robot vacuum cleaner builds a map of the floor plan. Voice-controlled speakers sit in kitchens and bedrooms, waiting for a wake word and holding a short buffer of audio. Doorbell cameras capture the pavement outside as well as the hallway inside, and can identify a familiar face. Individually these measurements seem trivial. Combined, they produce an intimate picture of when a household is empty, who visits it, how its members sleep and what they say to one another.

The benefits deserve to be taken seriously. Heating and lighting can be adjusted to actual occupancy, which saves energy as well as money. Cameras allow a carer to check on an elderly relative, and sensors can raise an alarm when someone falls. For a person with limited mobility, a voice command that turns off a lamp is not a gimmick but a form of independence. Any serious criticism of the smart home has to begin by acknowledging this.

The difficulty is where the information goes. Much of it travels to the manufacturer's servers, where it may be retained indefinitely, and the terms that govern its use are long documents that hardly anyone reads. Consent obtained this way is not a meaningful choice. Few buyers could explain, a year later, what they agreed to. Nor is the purpose fixed: data gathered to improve a voice assistant can later be used for advertising, or combined with information from other sources to build a profile of the household. This steady widening of use, sometimes called function creep, is the rule in the industry rather than an aberration.

A second problem is that a smart home is a network, and a network is only as secure as its weakest point. A cheap plug or bulb may have been designed with very little attention to security, and once it is on the same network as a camera, it can become a way in. Manufacturers abandon support for older models, leaving devices in place for years after they stop receiving security updates. Nor is the buyer usually told, at the moment of purchase, how long that support will last. A house full of unpatched gadgets is not a private one.

The third problem is rarely discussed, because we tend to think of a household as a single unit. It is not. A camera in a shared living room, an account that one person controls and others merely use, a history of who came home at what time: these things redistribute power inside a family. Campaigners have documented cases in which connected locks, thermostats and cameras were used by one partner to monitor or control another. Children, whose movements and conversations are recorded from an early age, have no say at all.

The consequences also extend past the front door. In several countries, police investigating crimes have asked manufacturers for footage from doorbell cameras, and some companies have made it easy for officers to request footage from users directly. Insurers have shown interest in the driving data produced by connected cars. None of these developments is necessarily sinister in itself, but together they mean that a device bought for convenience can become part of a system of surveillance that the buyer never agreed to. Each request may look reasonable; accumulated, they turn private property into a source of evidence.

Defenders of the industry sometimes reply that only people with something to hide need worry. This is too simple. Privacy is not merely a shield for individual secrets; it is a condition of ordinary social life. People behave differently when they believe they are being watched, and a society in which every room is monitored would be a poorer one even if no recorded conversation was ever heard by a human being. Nor am I persuaded that clear rules would stifle innovation. Common standards for security, and a requirement to state plainly what a device records, would help honest manufacturers compete on trust rather than on price alone.

What would a better arrangement look like? Manufacturers, not buyers, should carry the burden of protecting the data their products gather. Devices should collect as little as possible and process it locally where that is feasible. Packaging should state plainly what a device is able to sense. Sending information to third parties should require explicit permission, never a clause buried in the terms. Security updates should be guaranteed for a stated number of years. Until then, the smart home will remain a bargain in which the household pays for convenience with information whose value it cannot see.