WHEN YOU USE A CALCULATOR TO DO A MATHEMATICAL OPERATION . , YOU TRUST THE FIGURE IT PRODUCES. CAN WE TRUST A.I.? ALWAYS? READ : Concerns about AI developing skills independently of its programmers’ wishes have long absorbed scientists, ethicists, and, of course, plenty of science fiction writers. But perhaps we should be more concerned about what it gets flagrantly wrong, sometimes with an alarming amount of confidence. The industry calls these screw-ups “hallucinations.” Scientists have discovred that a.i. has created a language that the machines creators don t know elon musk has even suggested stopping them for some time until we humans can know what they are doing .. this remidns me of a phrase i think MR CORTAZAR wrote some years ago : SINCE THE LAST ONE OF THEM FORGOT TO DISCONNECT US , WE HAVE GONE ON WORKING " SCARY, TO SAY THE LEAST PROF PAT GREY

A.I. ARE MACHINES FINALLY TAKING OVER FROM US ? CECILY

ZANNY BEDOES is editor in chief of the prestigious ECONOMIST MAGAZINE SHE SAID THIS IN HER LATEST ISSUE Read in browser APRIL 20TH 2023 The Economist this week Highlights from the latest issue ________________________________________ ________________________________________ Zanny Minton Beddoes Editor-in-chief I am fascinated and awed by the new generative AI models popping up everywhere. But I have struggled with how to think about them. Should we be excited or anxious about this fast-moving technology—or both? The abilities of “large language models”—of the sort that power ChatGPT, a chatbot made by OpenAI—seem to have surprised everyone, including their creators. One recently passed the American Uniform Bar Examination, a fearsome test of lawyers' skills, in the 90th percentile. Our aim with this week’s issue was to help make sense of it all. In a special, expanded Science and Technology section, our writers explain how large, creative AI models work, the ways they could go wrong and what capabilities they might gain next. They’ve done a superb job of describing complex concepts in a clear and accessible way, including with interactive diagrams in digital versions of the articles. In an accompanying essay we look to history—from the printing press to computer browsers—to understand how this technology might change the world. Our leader pages, meanwhile, weigh the risks and opportunities of AI. If you are a subscriber and enjoy this week’s deep-dive into artificial intelligence, you can sign up to our weekly Simply Science newsletter. The machines have not taken over entirely this week, however. I also recommend our deeply reported article on why conflicts are lasting longer. The escalating crisis in Sudan is shocking. Sadly, our analysis shows that it is part of a broader trend. Editor’s picks Must-reads this week Technology and society How to worry wisely about artificial intelligence Read in browser APRIL 20TH 2023 The Economist this week Highlights from the latest issue ________________________________________ ________________________________________ Zanny Minton Beddoes Editor-in-chief I am fascinated and awed by the new generative AI models popping up everywhere. But I have struggled with how to think about them. Should we be excited or anxious about this fast-moving technology—or both? The abilities of “large language models”—of the sort that power ChatGPT, a chatbot made by OpenAI—seem to have surprised everyone, including their creators. One recently passed the American Uniform Bar Examination, a fearsome test of lawyers' skills, in the 90th percentile. Our aim with this week’s issue was to help make sense of it all. In a special, expanded Science and Technology section, our writers explain how large, creative AI models work, the ways they could go wrong and what capabilities they might gain next. They’ve done a superb job of describing complex concepts in a clear and accessible way, including with interactive diagrams in digital versions of the articles. In an accompanying essay we look to history—from the printing press to computer browsers—to understand how this technology might change the world. Our leader pages, meanwhile, weigh the risks and opportunities of AI. If you are a subscriber and enjoy this week’s deep-dive into artificial intelligence, you can sign up to our weekly Simply Science newsletter. The machines have not taken over entirely this week, however. I also recommend our deeply reported article on why conflicts are lasting longer. The escalating crisis in Sudan is shocking. Sadly, our analysis shows that it is part of a broader trend. Editor’s picks Must-reads this week Technology and society How to worry wisely about artificial intelligence Rapid progress in AI is arousing fear as well as excitement. How worried should you be? ________________________________________ Related How AI could change computing, culture and the course of history Large language models’ ability to generate text also lets them plan and reason How generative AI models work How generative models could go wrong Forgotten conflicts The world’s deadliest war last year wasn’t in Ukraine Sudan is not a one-off. There’s a disturbing resurgence in civil wars ________________________________________ The art of interpretation How to explain the puzzle of the world economy Welcome to the Mona Lisa effect ________________________________________ The too-big four Why EY and its rivals may eventually break up, after all The commercial logic for splitting up the big four is only getting stronger ________________________________________ The car industry Everything about carmaking is changing at once The industry must reinvent itself to keep pace, says Simon Wright ________________________________________ Related Special report: The car industry Bagehot If English nationalism is on the rise, no one has told the English The rise of English identity is largely myth ________________________________________ Economist Films AI safety How to stop artificial intelligence going rogue Researchers are increasingly worried about existential risks ________________________________________ The A-Z of economics Our word of the week Fiscal drag A way in which inflation can boost tax revenues. In most tax systems, workers must earn a certain amount before they pay income tax, or pay it a higher rate. If those allowances are not uprated every year in line with inflation, workers end up paying more in tax in real terms when their wages rise. Read about how Britain’s tax take is getting bigger, but the country’s tax system is deeply flawed. • Find more definitions in our A-Z of economics We’d like to hear from you Share your feedback via the email address below. Email newsletters@economist.com Was this email forwarded to you? Sign up here. This email has been sent to: cei@centroingles.com.ar. If you'd like to update your details please click here. Replies to this email will not reach us. If you don't want to receive these updates anymore, please unsubscribe here. Keep updated Advertising Info Terms & Conditions Help Privacy Policy Copyright © The Economist Newspaper Limited 2023. All rights reserved. Registered in England and Wales. No.236383 Registered office: The Adelphi, 1–11 John Adam Street, London, WC2N 6HT Rapid progress in AI is arousing fear as well as excitement. How worried should you be? ________________________________________ Related How AI could change computing, culture and the course of history Large language models’ ability to generate text also lets them plan and reason How generative AI models work How generative models could go wrong Forgotten conflicts The world’s deadliest war last year wasn’t in Ukraine Sudan is not a one-off. There’s a disturbing resurgence in civil wars