Last year, more than 36 million children under the age of 5 suffered from acute malnutrition worldwide. No child should suffer from lack of sustenance, but natural disasters, poverty and war continue to put the youngest and most vulnerable at risk. Child malnutrition in Sudan is at emergency levels, according to the World Health Organization. And in northern Gaza, up to a quarter of children are malnourished. Children in Afghanistan, Haiti, Nigeria, Yemen and other countries are also afflicted.
Helping children recover from malnutrition isn’t a matter of just giving them food. They are far more vulnerable to illness and death even after receiving treatment, and if they survive, they face an increased risk of health challenges throughout their lives. Scientists are trying to figure out why malnutrition hits the body so hard and how to help children recover. Recent research finds that prolonged malnutrition weakens the immune system and causes tissue and organ damage, making it difficult to absorb nutrients. Therapeutic food that boosts the gut microbiome may help, as might medications that repair the gut lining. But solutions feel much too far away with so many lives at stake.
Solving an intractable problem is also at the heart of another feature in this issue. Conservation biologist Ximena Velez-Liendo has devoted her life to saving the Andean bears of Bolivia. The charismatic creatures were at risk of extinction due to habitat loss and conflicts with humans. She worked with local communities to develop new sources of income, notably beekeeping, to slow the need to clear forests for farming and to help people coexist with bears. Both bears and people are thriving. “Conservation is changing,” she says, “from the hands of biologists to the hands of the people.”
And in the third feature, we delve into the challenges of assessing the cognitive skills of machines. Over the last few years, generative AI chatbots like ChatGPT have dazzled people with the ability to write computer code, buff up resumes, help with homework and more. Along with these feats have come extravagant claims that the algorithms can think and reason. Finding out if they actually do, it turns out, is complicated.
Attempts to test machine intelligence date back to 1950, when famed British mathematician Alan Turing designed the imitation game, in which an interrogator asks questions of a computer and a person. If the interrogator could not tell which answers came from a human, “one will be able to speak of machines thinking without being contradicted,” Turing said. He predicted that a computer would win the game by the end of the 20th century; that didn’t happen. But interpreting the results of the imitation game and other tests isn’t clear-cut. Today’s bots can appear to best humans at tests of mathematics, language comprehension and more. Instead of demonstrating superior intellect, however, AI may simply be regurgitating what it learned during training. (I was gratified to read that bots that appeared to have grasped the concept of multiplication had just memorized the answer.) Researchers are now devising new tests that produce less ambiguous results.
This is our summer double issue; we’ve included more articles to tide you over until the August 10 issue lands in your mailbox.