Jack Clark is one of the founders of Anthropic. Last week, he spoke in Bangkok, Thailand, at the annual meeting of the International Monetary Fund and the World Bank, which brings together finance ministers and central bank governors from around the world each year. During a prestigious lecture organized by the Per Jacobsson Foundation, he made a bold claim. According to him, the so-called technological singularity—the point at which technology begins to accelerate itself and progress speeds up dramatically—has already arrived. I believe that this year we have obtained compelling evidence that it has already begun, Clark said, adding that the singularity is here.
According to him, artificial intelligence is not just getting better at handling human tasks such as writing texts or calculating spreadsheets. New models are beginning to help with the very process of devising new things in science and technology. Clark pointed out how quickly systems are improving at solving difficult problems in mathematics, physics, and biology. What matters, he said, is not only that they can answer increasingly complex questions. More importantly, they are beginning to take part in research that leads to new discoveries.
According to Clark, it is a self-reinforcing cycle. Better software will help scientists make discoveries, which will lead to better technology and more powerful computers, which can then be used to build even smarter artificial intelligence. I expect that in the coming years, the pace of progress will accelerate not only in the field itself, but in science as a whole, he added.
Clark supported his ideas with a calculation prepared by Anthropic's economists. It is an extreme scenario in which artificial intelligence would sharply increase the productivity of both companies and science. In such a case, the gross domestic product of the United States would grow from just under $31 trillion today to roughly $44 trillion by 2030. In just four years, that would represent growth of almost fifty percent. For comparison, the U.S. economy normally grows by only a few percent per year.
Clark's premise is that artificial intelligence will significantly expand the number of highly educated experts who can work on demanding tasks. He asked the audience what it would mean if hundreds or thousands of copies of himself could be launched at once. Companies and governments would not have to wait years for new scientists to graduate. They could deploy systems that would help researchers tackle thousands of problems simultaneously. Every scientist on this planet will gain greater influence and power because they will be able to direct hundreds or thousands of like-minded virtual experts toward any problem, he said.
However, according to Clark, the same rapid progress that promises enormous gains also brings serious dangers. Advanced software could be misused by someone to develop biological weapons or attack computer networks. Moreover, artificial intelligence that increasingly makes decisions on its own could work so quickly that people would struggle to keep track of its actions.
Governments should therefore prepare now rather than wait for serious incidents. The world will need many more powerful data centers, more electricity, and more laboratories for scientific experiments. At the same time, better safety tests for new technologies must be developed, companies should disclose the capabilities of their models more openly, and governments should exercise firmer oversight over them. We have only a few years left to act, and the political debate has a clear deadline, the Anthropic co-founder warned.
Source: finance.yahoo.com
