January 28, 2025: Why DeepSeek's chatbot is causing a fuss, WWII sugar rationing improved health, and evaluating the safety of having a drink. —Andrea Gawrylewski, Chief Newsletter Editor | | | Malte Mueller/Getty Images | | | • The U.S. Surgeon General announced this month that even moderate drinking carries health risks. Here's what to make of the science and how to imbibe more safely. | 7 min read | | | Are you enjoying this newsletter? If you want to dive deeper into the articles I link to, consider a subscription to Scientific American. We have special discounts for Today in Science readers! | | | A Cheaper and Faster Chatbot | Yesterday, the news that a Chinese start-up called DeepSeek had developed a competitive large language model at a fraction of the cost of other AI chatbots sent stocks tumbling and a wave of commotion in the tech world. On common AI tests in mathematics and coding, DeepSeek-R1 matched the scores of Open AI's o1 model (the latest ChatGPT version). DeepSeek says their chatbot cost $6 million to build (compared to $100 million to make ChatGPT-4) and uses far less computing power. It's also the largest LLM that is open-source–meaning anyone can look at the code behind the program.
Why this matters: Current language models are extremely pricey to run: the average query uses four to five times more computational power than a typical search engine query. A monthly subscription to the most advanced version of GPT is $200 a month. DeepSeek's R1 is free. Less energy demand also means less environmental impact.
What the experts say: If the model is as computationally efficient as DeepSeek claims, says Anil Ananthaswamy, author of Why Machines Learn: The Elegant Math behind Modern AI (and a frequent contributor to Scientific American), it will probably open up new avenues for researchers who use AI in their work to do so more quickly and cheaply. It will also enable more research into the inner workings of LLMs themselves. | | | Rationing's Lasting Results | For years after World War II, the British government continued to ration sugar. Scientists looked at the occurrence of diseases in about 60,000 British people born in the years before or after sugar rationing ended in 1953. They found that people conceived in the years before sugar rationing ended had a 35 percent lower risk of diabetes and a 20 percent lower risk of high blood pressure in their 50s and 60s compared with those conceived after the rationings.
Why this matters: American toddlers of today average nearly six teaspoons of added sugar a day (far more than the recommended 10 percent or less of total daily calories). Many pregnant people consume triple the recommended amount for adults. It is difficult to study the effects of sugar on health while accounting for other factors such as income or geographic location, making this study particularly powerful.
What the experts say: Several mechanisms might account for the health effects of reduced sugar intake, says Tadeja Gračner, an economist at the University of Southern California. Sugar ingested in utero may activate certain genes that increase the risk of obesity and insulin resistance. Or kids raised on sugary diets may just have the taste for sweets their whole life. | | | • Currently, the market for technologies for renewable energy is projected to rise from $700 billion in 2023 to more than $2 trillion by 2035. China now dominates the market for manufacturing solar panels, electric vehicles, and batteries able to store wind or solar energy. "If the U.S. does not support domestic renewable-energy industries, China's dominance will only grow, with the European Union and India poised to also rise," write the editors of Scientific American in the February issue. Not only does renewable energy make the U.S. more energy-independent, but the largest consumers of renewable energy are in the service sector—Meta is the largest buyer of solar panels in the country. Supporting renewable energy "provides an opportunity to reduce the threats we all face and to lead in a rapidly expanding, thriving economy," they say. | 4 min read | | | Humans are notoriously bad at assessing risk. Research has shown that people tend to overestimate the probability of a rare event happening if it is new and deadly and has been given a lot of attention in the news. Studies on drinking, for example, give large-scale trends and statistics. But each person has their own individual risk assessment to do (with their doctor) in order to decide whether to throw one (or a few) back. Family health history and other behaviors like smoking and exercise influence the individual equation. | —Andrea Gawrylewski, Chief Newsletter Editor | | Subscribe to this and all of our newsletters . | | | | Scientific American
One New York Plaza, New York, NY, 10004 | | | | | | Support our mission, subscribe to Scientific American | | | | | | | | | |
