Good morning. Here’s the latest:
More news is below. But first, we answer your questions on A.I.
A.I. questions
Artificial intelligence is no longer a gee-whiz technology. It’s already reshaping the workplace, the academy, the culture. The Times employs many reporters who cover those changes. We recently invited readers of The Morning to submit questions about artificial intelligence. You wanted to know who’s winning and losing, how the tech works and how governments are coping. To answer, we enlisted beat experts from across the newsroom. (Got a question for us? Submit it here.) The techHow does A.I. use energy and how much does it use? Stephanie Christie, San Clemente, Calif. Karen Weise, a tech reporter who has covered data centers, writes: A.I. needs a lot of power. There’s not enough electricity to meet that demand, so U.S. energy consumption will rise. Some could come from growing renewable sources and a potential revival of nuclear power, but for now much is coming from natural gas, which contributes to climate change. In Louisiana, for example, three new gas plants will be built to power a massive Meta data center. In the long term, new breakthroughs could reduce the climate burden: Maybe A.I. will become more efficient, and maybe it will make other industries more efficient. Tech companies hope new power technologies pan out, including smaller nuclear reactors, improved batteries, geothermal sources and nuclear fusion. I’m retired — and not exposed to A.I. at work or in school. But I don’t want to be a Luddite, like elders of yore who said they didn’t need to know how to use a computer or search the web. What should I do to keep up? Lindy Washburn, Fair Lawn, N.J. Brian X. Chen, a tech columnist, writes: OpenAI’s ChatGPT chatbot is what started this whole conversation about A.I., so, at minimum, I’d suggest using a web browser to visit chatgpt.com and start typing away. Some ideas: Ask the chatbot to draft a letter by typing “write a testy letter to my electricity provider asking why my bill was so high last month.” Ask the chatbot a question that you would normally type into Google.com, such as, “What’s the difference between the iPhone 17 and iPhone 12?” Ask the chatbot to produce images by typing something like “generate an illustration of a cat on a window sill.” From there, just use your imagination. Tech companies are trying to stop hallucinations, where A.I.s give fictitious answers. Why does this phenomenon happen and what work is being done to rectify this problem? Julie Lynn Moore, Marion, Ind. Cade Metz, who reports on artificial intelligence, writes: Companies have been working on this problem for years. And in some cases, they have managed to reduce the number of hallucinations and other mistakes made by these A.I. systems. But the truth is that there is no known way of ensuring that they produce accurate information. When you type questions into a chatbot, it uses mathematical probabilities to choose each response. This means that a certain number of responses are going to be wrong. The trick is to always be aware that these systems make mistakes. Be skeptical of everything they say — and always double-check important information. It’s getting harder to tell A.I.-generated content from human-made content. What efforts exist to address this? Joel Dixon, Round Rock, Texas Stuart Thompson, who covers the spread of misinformation, writes: There are some companies that analyze images, videos and text for signs that they are made by A.I. Our tests from a few years ago showed that some were pretty good, while others really struggled. The companies say it is an arms race: as they improve their A.I. detectors, the A.I. tools get better, too. For text, we have sometimes used GPTZero. For images, we have used AI Or Not. These tools are merely a starting point. If you see something dubious, just pause and think. If you have any doubts, don’t share. (To see how hard it is to recognize A.I. videos now, try my quiz.)
Business and the economyWhat job industries are less likely to be taken over by A.I. and more likely to require humans? Audrey Toda, Woodside, Calif. Lydia DePillis, who reports on the American economy, writes: Hi Audrey — this is a simple question with a surprisingly complex answer. So far, the most endangered occupations are those that process a lot of formulaic information: Think financial consultants, medical billing analysts, language translators. In some professions, such as computer programming, A.I. might augment a worker’s capabilities rather than replace them, making the worker more productive. If there’s enough demand for those services, the industry could actually add jobs. Right now, there’s little evidence that A.I. is wreaking havoc on employment, but it’s early days. Ultimately, the industries safest from A.I. are those that require skilled physical work, whether it’s carpentry or classical dance. Which countries stand to gain the most in terms of access to minerals, computer chips, data centers and the skill sets that will be most valuable in managing A.I. for the world? Lindsey Deperi-Franz, Boulder, Colorado Adam Satariano, a tech correspondent based in London, writes: The U.S. is the biggest winner, but other countries stand to gain. China has promising A.I. companies and controls the supply of many key minerals needed for chips, data centers and other tech. The Netherlands is home to the complex lithography machines needed to make leading A.I. chips. Taiwan is the global hub for chip manufacturing. The U.A.E., Malaysia and Sweden are building data centers. Australia, the Democratic Republic of Congo and Indonesia could also benefit through the production of minerals like lithium, cobalt and nickel. Large language models hoover up writing and data from around the web. How do copyright and intellectual property rights factor in? Charlotte Keene, Austin, Texas Cade Metz, who reports on artificial intelligence, writes: That question is currently being decided by more than 40 court cases across the country. This includes a suit The New York Times brought against OpenAI and its partner, Microsoft, claiming that the companies trained their A.I. using our stories without permission or payment. (OpenAI and Microsoft have denied those claims.) This month, the A.I. start-up Anthropic agreed to pay $1.5 billion to a group of authors and publishers after a judge ruled that the company had illegally downloaded and stored millions of copyrighted books. It was the largest payout in the history of U.S. copyright cases, and it could pave the way for more companies to pay large sums to rights holders, either through court settlements or licensing fees. But the many court battles over A.I. and copyright are only just getting started — and there are many legal questions yet to be decided. Government and oversightWhat government oversight is in place to prevent the use of A.I. for propaganda and disinformation? Lori Brown, Santa Cruz, Calif. Cecilia Kang, a D.C.-based reporter who covers tech policy, writes: The short answer is none. There is no federal law against A.I.-generated propaganda or disinformation. And if there were, tech companies would surely challenge it in court on free speech grounds. But there are efforts in states to curb the use of deepfakes and other A.I.-generated content in elections, with required disclosures when the technology is being used. If A.I. displaces millions of jobs — making it harder for a huge share of the population to pay taxes and buy goods or services — will federal or state governments have a plan to address such a historic disruption? Todd Hill, Fredericktown, Ohio Benjamin Casselman, the Times’s chief economics correspondent, writes: Everyone in Silicon Valley is talking about the ways A.I. could change the economy. But that discussion has barely begun to shape policy debates. And given the uncertainty around A.I., it’s hard to know what those policies should even be. Will it displace a subset of workers or lead to a more fundamental change to the nature of work? Will it create new categories of jobs, and, if so, what skills will workers need? The government’s history of addressing major economic shifts isn’t encouraging — think about the failure to help workers displaced by the industrial revolution or globalization. That is partly because the effects of technological changes can be hard to predict. In the 1990s, many people thought the internet would change the economy, but few foresaw how. That means government policies almost always end up responding to problems once they develop, rather than anticipating them before they occur.
Charlie Kirk
Politics
Sudan
More International News
Other Big Stories
New York City’s surveillance is so extensive that it has begun to erode basic civil rights, Elizabeth Daniel Vasquez writes. When fewer companies go public, the rich get richer while innovation declines. Deregulating the stock market could incentivize them to join it, Bryce Tingle writes. Here’s a column by David French on political violence. New: The Times family subscription is here. One rate. Four individual logins. Savings for all. Now you and three others can enjoy unlimited access to The Times, while personalizing your own experience. Learn more.
Delayed for decades: A postcard from the U.N. building was returned to the sender — 72 years late. Mock the rich: It’s a counterintuitive way to sell upscale Florida real estate. But it’s working. Metropolitan Diary: Sheet cake and a swig of milk on the 6 train. Distilling the past: The archaeologist Patrick McGovern, who died at 80, studied the history of alcohol — and then recreated beverages from traces in ancient drinking vessels.
|
