{"id":423957,"date":"2026-04-29T20:18:28","date_gmt":"2026-04-29T20:18:28","guid":{"rendered":"https:\/\/www.newsbeep.com\/ie\/423957\/"},"modified":"2026-04-29T20:18:28","modified_gmt":"2026-04-29T20:18:28","slug":"how-jagged-intelligence-can-reframe-the-ai-debate-the-irish-times","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ie\/423957\/","title":{"rendered":"How \u2018jagged intelligence\u2019 can reframe the AI debate \u2013 The Irish Times"},"content":{"rendered":"

Say what you will about whether artificial intelligence<\/a> (AI) will one day be as smart as a human. It has already become a star math student. Last summer, AI built by Google<\/a> and OpenAI<\/a> correctly answered five of six complex questions at the International Math Olympiad, an annual competition for the world\u2019s top high school students.<\/p>\n

AI\u2019s common sense, however, may still be a bit lacking. <\/p>\n

A few months later, Anuradha Weeraman, a software engineer in Sri Lanka<\/a>, noticed that leading AI systems struggled to answer what was essentially a trick question that most people would find laughably simple. When he told various chatbots that he needed to take his car to a repair shop that was only 50m (164ft) away and asked if he should walk or drive, the bots told him to walk.<\/p>\n

The strange way that AI looks like a genius at one moment and dense in another is what researchers, engineers and economists call \u201cjagged intelligence\u201d. They use this term to explain why AI is racing ahead in some areas \u2013 like math and computer programming \u2013 while still struggling to make headway in others.<\/p>\n

The term, which is widely used by the people building AI and analysing its effects, could help reframe the debate over whether these systems are becoming as smart as, or even smarter than, humans. Instead, researchers argue, AI is something completely different: far better than humans at some tasks and far worse at others.<\/p>\n

Understanding those strengths and weaknesses can also help economists get a better handle on what AI means for the future of employment. <\/p>\n

While entry-level programmers have reason to worry about their jobs, for example, it is not so clear \u2013 at least right now \u2013 how AI will affect other work. But watching where AI starts to make rapid improvements could help predict what kinds of jobs will be affected by the technology.<\/p>\n

\"
\n AI has always been compared to human intelligence, but that may not be the right way to think about it. Knowing what it does do well could help predict what jobs it may replace. (Pingnan Lu\/The New York Times) — FOR EDITORIAL USE ONLY WITH NYT STORY AI JAGGED INTELLIGENCE by METZ of APRIL 15, 2026. ALL OTHER USE PROHIBITED.<\/p>\n

\u201cThe performance of these systems varies, and it is not easy to tell when they will fail to do things a human can do,\u201d Weeraman said.<\/p>\n

The term \u201cjagged intelligence\u201d was coined by Andrej Karpathy, one of the founding researchers at OpenAI, a former head of self-driving technology at Tesla and, on social media, one of the most closely watched commentators on the rise of AI.<\/p>\n

\u201cSome things work extremely well [by human standards] while some things fail catastrophically [again by human standards],\u201d he wrote on social media in 2024, \u201cand it\u2019s not always obvious which is which.\u201d<\/p>\n

This, he wrote, is different from the human brain, \u201cwhere a lot of knowledge and problem solving capabilities are all highly correlated and improve linearly all together, from birth to adulthood\u201d.<\/p>\n

Since OpenAI started the AI boom in 2022, tech executives have seesawed between warning that their new creations could have a devastating effect on white-collar jobs and downplaying the long-term impact on employment.<\/p>\n

So far, outside of the technology industry, there is only anecdotal evidence that AI has become a job killer. But given how quickly the technology is improving, many tech experts argue that whether AI replaces other kinds of white-collar workers is not a question of if but when. Only a few years ago, these systems were just starting to show the most rudimentary programming skills.<\/p>\n

\u201cThese systems have been showing incredible improvements,\u201d said Alex Imas, an economist at the University of Chicago\u2019s Booth School of Business. \u201cEvery time there is a big new release, people are surprised by how much it can do.\u201d<\/p>\n

If a job involves a bunch of different tasks \u2013 and most jobs do \u2013 some tasks will be automated and some will not<\/p>\n

\u2014 \u00a0Alex Imas, economist<\/p>\n

But technology that adds to what workers can do without replacing them has plenty of precedent, and that\u2019s what some AI researchers and economists are arguing will happen. <\/p>\n

As far back as the 1960s, a pocket calculator could add, subtract and multiply much faster than a person. That did not mean a calculator could replace an accountant.<\/p>\n

[\u00a0Job losses and AI: Could technology bring Ireland back to the dark days of 2008?Opens in new window<\/a>\u00a0]<\/p>\n

Now, systems such as Anthropic\u2019s Claude and OpenAI\u2019s Codex can write computer code much faster, too. But they are not that good at understanding how each piece of code fits into a larger software application. They need human help with that.<\/p>\n

\u201cIf a job involves a bunch of different tasks \u2013 and most jobs do \u2013 some tasks will be automated and some will not,\u201d Imas said. \u201cAnd if that is the case, the worker may have more time to do bigger things.\u201d<\/p>\n

Last month, Fran\u00e7ois Chollet, a noted AI researcher, released a new digital benchmark test called ARC-AGI 3. It asks for solutions to hundreds of game-like puzzles without providing a single instruction for how to solve them. All of the puzzles can be solved by an average, untrained person, but the leading AI systems fail to master any of them, according to testing done by Chollet and the ARC Prize, the non-profit research lab that oversees the test.<\/p>\n

Once people realise that AI is a jagged intelligence, experts like Chollet said, they develop a better understanding of how AI is likely to evolve in the coming years \u2013 and what effect it might have on the labour market.<\/p>\n

AI does not have general intelligence. What it has is a lot of different skills<\/p>\n

\u2014 \u00a0Fran\u00e7ois Chollet, AI researcher<\/p>\n

\u201cThis will depend on what tasks it automates and how and when,\u201d Imas said.<\/p>\n

AI systems, such as Claude and OpenAI\u2019s ChatGPT, learn their skills by pinpointing patterns in digital data, including Wikipedia articles, news stories, computer programs and other text culled from across the internet. But that gets them only so far.<\/p>\n

The internet holds only a small fraction of human knowledge. It records what people do in the digital world, but contains comparatively little information about what happens in the physical world.<\/p>\n

That means these systems can write emails, answer questions, riff on almost any topic and generate computer code. But because AI systems reproduce the patterns they find in digital data, they are not good at planning ahead, generating new ideas or tackling tasks they have not seen before.<\/p>\n

\u201cAI does not have general intelligence,\u201d Chollet said. \u201cWhat it has is a lot of different skills.\u201d<\/p>\n

Now, companies like Anthropic and OpenAI are teaching these systems additional skills using a technique called reinforcement learning. By working through thousands of math problems, for example, they can learn which methods lead to the right answer and which do not.<\/p>\n

This works well in areas like math and computer programming, where AI companies can clearly define good and bad behaviour. The answer to a math problem is either right or wrong. Computer code either passes a performance test or it fails.<\/p>\n

The jaggedness of AI means that the problems can come from anywhere<\/p>\n

\u2014 \u00a0Joshua Gans, economist<\/p>\n

But reinforcement learning does not work as well in areas like creative writing or philosophy, or even some of the sciences, where the distinction between good and bad is harder to pin down.<\/p>\n

\u201cCoding \u2013 which everyone is enthusiastic about at the moment \u2013 is not representative of everything AI does,\u201d said Joshua Gans, an economist at the University of Toronto\u2019s Rotman School of Management. \u201cWith coding, it is much easier to use a feedback loop to figure out what is working and what isn\u2019t.\u201d<\/p>\n

For users, it is often hard to tell what AI does well and what it does not. And when people finally get a firm handle on the strengths and weaknesses of the systems, the technology changes.<\/p>\n

\u201cThe jaggedness of AI means that the problems can come from anywhere,\u201d Gans said. \u201cThere are gaps, and we don\u2019t always know where the gaps are.\u201d<\/p>\n

The wild card is that AI is quickly improving. Many of the weaknesses that Karpathy and others pointed out in 2024 and early 2025 are no longer there. Companies will find other shortcomings and fix them as well.<\/p>\n

\u201cThe valleys in the technology are closing,\u201d Imas said. \u2013 This article originally appeared in The New York Times<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"Say what you will about whether artificial intelligence (AI) will one day be as smart as a human.…\n","protected":false},"author":2,"featured_media":423958,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[220,3673,218,219,2670,61,60,1682,80],"class_list":["post-423957","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-anthropic","tag-artificial-intelligence","tag-artificialintelligence","tag-chatgpt","tag-ie","tag-ireland","tag-openai","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/423957","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=423957"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/423957\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/423958"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=423957"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=423957"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=423957"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}