IELTS Academic Reading — Practice Test 19

Fresh water has always been unevenly distributed, but the pressure on it is now acute. Rivers that once ran to the sea are drained before they reach it; aquifers built up over thousands of years are being pumped dry within decades; and a warming climate is making rainfall less predictable in many of the regions that can least afford the uncertainty. The shortfall is not only a matter of quantity but of timing and place: water may be plentiful in one season and gone the next, or abundant in a region far from where it is actually needed. Set against this scarcity is an apparently limitless resource. More than nine-tenths of the planet's water is in its oceans, and the idea of drawing on that vast reserve — of turning seawater into something people can drink and grow crops with — has an obvious appeal. The difficulty has always been the cost of removing the salt.

For most of the twentieth century, desalination meant distillation: seawater was heated until it evaporated, leaving the salt behind, and the vapour was then condensed into fresh water. The method worked but devoured energy, and it made sense only where fuel was cheap. The technology that changed the economics is reverse osmosis. Here seawater is forced under great pressure through membranes so fine that water molecules pass while dissolved salts are held back. No boiling is required. Reverse osmosis now accounts for the majority of the world's desalinated water, and continuing improvements to the membranes have steadily lowered the pressure, and therefore the power, that the process demands.

The technology has taken hold wherever arid land meets the sea and money is available. The Gulf states were among the earliest large-scale users; cities such as Dubai and Riyadh depend on desalinated water for much of their supply. Israel, facing chronic shortages, built a string of large plants along its coast and now draws a substantial share of its drinking water from the Mediterranean. Australia, jolted by the long drought of the early 2000s, constructed several major facilities in a hurry, some of which sat idle once the rains returned. Thousands of plants now operate worldwide, and the global capacity for turning seawater into fresh water roughly doubles every decade.

None of this comes cheaply. Even the most efficient plants use far more energy per litre than conventional treatment of river or ground water, and when that energy is drawn from fossil fuels, desalination adds to the greenhouse gases that are worsening the very shortages it sets out to relieve. This is the paradox at the heart of the technology: a defence against a drier climate that can, if carelessly powered, help to bring that climate about. The financial cost, too, remains high enough that desalinated water is usually the option of last resort, adopted only when cheaper sources have been exhausted. Building a large plant also ties up capital for decades, a commitment that governments in drier, poorer countries can rarely afford to make.

A second cost attracts far less attention than it should. For every litre of fresh water it produces, a desalination plant leaves behind a volume of concentrated salt water, known as brine, that is denser and warmer than the sea and often laced with the chemicals used to clean the membranes. Most plants simply pump this brine back into the ocean, where it can sink and spread across the seabed, starving marine life of oxygen in the immediate area around the outfall. Diluting the brine before it is released, or spreading it through diffusers to disperse it more widely, can reduce the damage but pushes up the cost again. The scale of the problem has been consistently underestimated, and disposing of brine safely adds further to the expense.

Defenders of the technology point out how much has changed. The energy needed to desalinate a litre of seawater has fallen dramatically, helped by devices that recover pressure from the outflowing brine and reuse it, and a growing number of plants are being coupled with solar or wind power to shrink their carbon footprint. Critics counter that the cheapest litre of water is the one never used, and that fixing leaking pipes, pricing water sensibly and curbing waste often deliver more for less than any new plant. The argument, in essence, is between increasing supply and reducing demand — and the two need not be in opposition.

The sensible conclusion is that desalination is a tool, not a cure. For a wealthy, coastal, sun-baked city with no rivers to spare, it may be the only realistic way to keep the taps running. For a poor community far inland, it offers nothing: the water would have to be pumped uphill over great distances at a cost no one could bear. Treated as one option among several, and paired with genuine efforts at conservation, desalination has a part to play. Treated as a substitute for using less, it promises more than it can deliver.

Artificial intelligence can now do things that not long ago seemed the exclusive province of the human mind. It translates between languages, holds fluent conversations, writes passable essays and spots tumours in medical scans that human eyes have missed. For most of its history the field advanced by narrow triumphs — a program that could play chess, another that could sort mail by its postcode — each impressive in its own lane and helpless the moment it stepped outside it. Yet a question hangs over these newer accomplishments, and it grows sharper as the systems improve. When a machine produces the right words, does it understand what those words mean, or is it merely shuffling symbols according to rules, with no more comprehension than a calculator has of arithmetic? The distinction between doing and understanding lies at the centre of the debate over artificial general intelligence — the prospect of a machine that could grasp and reason across any domain, as a person can, rather than excelling at a single narrow task.

One long-standing answer sidesteps the question of inner comprehension altogether. The mathematician Alan Turing proposed that if a machine could converse so convincingly that a human judge could not tell it apart from another person, then to insist it was not 'really' thinking would be an empty complaint. What we cannot observe, on this view, we have no business asserting; behaviour is all the evidence we have of thought in other people, so behaviour should suffice for machines too. The test he described judges only what a system does, not what, if anything, is going on within it. For decades this behavioural standard framed how success in the field was imagined.

The most famous challenge to that standard was posed by the philosopher John Searle. Imagine, he said, a person who speaks no Chinese locked in a room with a vast rulebook. Slips of paper bearing Chinese characters are passed in; by following the rulebook, which tells him only which symbols to send out in response to which symbols coming in, the person produces replies so apt that Chinese speakers outside are convinced they are corresponding with a fellow speaker. Yet the person understands nothing; he is manipulating shapes he cannot read. A computer running a program, Searle argued, is in exactly this position. Processing symbols by their form — their syntax — can never, by itself, generate an understanding of their meaning, their semantics.

Searle's critics were quick to respond, and their most influential objection is known as the systems reply. The person in the room, they concede, does not understand Chinese — but the person is only one component, like a single neuron in a brain. It is the entire system, the person together with the rulebook, the paper and the practised procedure, that understands, and there is no contradiction in a whole possessing a property that none of its parts possesses on its own. This answer flows from a broader position, functionalism, which holds that what makes a state a mental state is the role it plays — its pattern of causes and effects — and not the physical material that happens to carry it. If that is right, then understanding could in principle be realised in silicon as readily as in flesh.

A different worry cuts across this dispute. Words, some argue, mean something only because they are anchored to the world beyond language — to things seen, touched and acted upon. A system that has learned solely from text has, in a sense, only ever been told how words relate to other words; its dictionary is closed, each term defined endlessly in terms of others, never connected to an actual object or experience. This is the grounding problem, and it leads some researchers to insist that genuine understanding will require a body and senses, a way of bumping up against reality, rather than ever more text processed in isolation.

It may be, too, that the whole argument rests on a false assumption: that understanding is a single thing a system either has or lacks. Perhaps it comes in degrees, so that a machine might grasp some things partially and others not at all, just as a young child or an animal does. There is a further, uncomfortable possibility. Our confidence that other people understand rests largely on how much they resemble us, and we may be applying the same instinct, unexamined, when we deny understanding to a machine simply because it is unlike us. If our judgements are shaped by this anthropocentric bias, they are a poor guide to what is actually there.

This is why the question refuses to be settled by any experiment we can presently devise. To test whether a machine understands, we would first need to agree on what understanding is, and about that even the study of our own minds has reached no firm conclusion. Nor is the difficulty merely academic: how we answer will shape the rights we are prepared to extend to the systems we build and the trust we are willing to place in their judgements. The debate over artificial understanding turns out to be, in large part, a debate about ourselves — about what we are doing when we grasp a meaning, and whether that act is something a mechanism could share. It is possible that the honest answer, for now, is that we do not know how we would tell; and that admission, rather than any confident verdict, may be the most accurate description of where the enquiry stands.