{"id":2772,"date":"2023-08-01T00:01:37","date_gmt":"2023-07-31T23:01:37","guid":{"rendered":"https:\/\/springerhealthcare.nl\/france\/?p=2772"},"modified":"2023-08-22T14:14:37","modified_gmt":"2023-08-22T13:14:37","slug":"how-ai-knows-things-no-one-told-it","status":"publish","type":"post","link":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/","title":{"rendered":"How AI Knows Things No One Told It"},"content":{"rendered":"

[vc_row full_width=”stretch_row” content_placement=”middle” css=”.vc_custom_1663187695627{padding-bottom: 37px !important;background-image: url(http:\/\/springerhealthcare.nl\/wp-content\/uploads\/sites\/24\/2022\/09\/science_connect_back.png?id=2486) !important;background-position: center !important;background-repeat: no-repeat !important;background-size: cover !important;}” el_class=”sciencetitle”][vc_column][vc_custom_heading text=”Science Connect” font_container=”tag:h2|text_align:left|color:%23ffffff” use_theme_fonts=”yes” link=”url:https%3A%2F%2Fspringerhealthcare.nl%2Ffrance%2Fscience-connect-august-2023%2F”][vc_column_text css=”.vc_custom_1663186747780{margin-bottom: 40px !important;}”]A Springer Healthcare Initiative For Pharma Professionals<\/span>[\/vc_column_text][vc_column_text]AUGUST 2023
\n<\/span>[\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1538651642953{padding-top: 75px !important;}”][vc_column width=”3\/4″ css=”.vc_custom_1538657061395{border-right-width: 1px !important;padding-top: 0px !important;border-right-color: #8faec1 !important;border-right-style: solid !important;}”][vc_custom_heading source=”post_title” font_container=”tag:h1|font_size:40px|text_align:left|line_height:60px” google_fonts=”font_family:Merriweather%3Aregular%2Citalic|font_style:700%20regular%3A700%3Anormal” el_class=”title” css=”.vc_custom_1664358136969{margin-bottom: 30px !important;}”][vc_column_text css=”.vc_custom_1686833173052{padding-bottom: 0px !important;}”]Researchers are still struggling to understand how AI models trained to parrot Internet text can perform advanced tasks such as running code, playing games and trying to break up a marriage<\/strong><\/span>[\/vc_column_text][vc_column_text el_class=”intro-box”]By George Musser<\/span>[\/vc_column_text]

\"how-ai-knows-things-no-one-told-it\" <\/div> <\/div>[vc_column_text css=”.vc_custom_1689357052532{margin-bottom: 30px !important;}” el_class=”sc-img-caption”] Credit: Chris Gash[\/vc_column_text][vc_column_text]

No one yet knows how\u00a0ChatGPT<\/a>\u00a0and its\u00a0artificial-intelligence cousins<\/a>\u00a0will transform the world, and one reason is that no one really knows what goes on inside them. Some of these systems’ abilities go far beyond what they were trained to do\u2014and even their inventors are baffled as to why. A growing number of tests suggest these AI systems develop internal models of the real world, much as our own brain does, although the machines’ technique is different.<\/p>\n

\u201cEverything we want to do with them in order to make them better or safer or anything like that seems to me like a ridiculous thing to ask ourselves to do if we don’t understand how they work,\u201d says Ellie Pavlick of Brown University, one of the researchers working to fill that explanatory void.<\/p>\n

At one level, she and her colleagues understand GPT (short for \u201cgenerative pre-trained transformer\u201d) and other large language models, or LLMs, perfectly well. The models rely on a machine-learning system called a neural network. Such networks have a structure modeled loosely after the connected neurons of the human brain. The code for these programs is relatively simple and fills just a few screens. It sets up an autocorrection algorithm, which chooses the most likely word to complete a passage based on laborious statistical analysis of hundreds of gigabytes of Internet text. Additional training ensures the system will present its results in the form of dialogue. In this sense, all it does is regurgitate what it learned\u2014it is a\u00a0\u201cstochastic parrot,\u201d<\/a>\u00a0in the words of Emily Bender, a linguist at the University of Washington. (Not to dishonor the late\u00a0Alex, an African Grey Parrot<\/a>\u00a0who understood concepts such as color, shape and \u201cbread\u201d and used corresponding words intentionally.) But LLMs have also managed to\u00a0ace the bar exam<\/a>, write a sonnet about the Higgs boson and make an attempt to\u00a0break up their users’ marriage<\/a>. Few had expected a fairly straightforward autocorrection algorithm to acquire such broad abilities.<\/p>\n

That GPT and other AI systems perform tasks they were not trained to do, giving them \u201cemergent abilities,\u201d has surprised even researchers who have been generally skeptical about the hype over LLMs. \u201cI don’t know how they’re doing it or if they could do it more generally the way humans do\u2014but they’ve challenged my views,\u201d says Melanie Mitchell, an AI researcher at the Santa Fe Institute.<\/p>\n

\u201cIt is certainly much more than a stochastic parrot, and it certainly builds some representation of the world\u2014although I do not think that it is quite like how humans build an internal world model,\u201d says Yoshua Bengio, an AI researcher at the University of Montreal.<\/p>\n

At a conference at New York University in March, philosopher Rapha\u00ebl Milli\u00e8re of Columbia University offered yet another jaw-dropping example of what LLMs can do. The models had already demonstrated the ability to write computer code, which is impressive but not too surprising because there is so much code out there on the Internet to mimic. Milli\u00e8re went a step further and showed that GPT can execute code, too, however. The philosopher typed in a program to calculate the 83rd number in the Fibonacci sequence. \u201cIt’s multistep reasoning of a very high degree,\u201d he says. And the bot nailed it. When Milli\u00e8re asked directly for the 83rd Fibonacci number, however, GPT got it wrong, which suggests the system wasn’t just parroting the Internet. Rather it was performing its own calculations to reach the correct answer.<\/p>\n

Although an LLM runs on a computer, it is not itself a computer. It lacks essential computational elements, such as working memory. In a tacit acknowledgment that GPT on its own should not be able to run code, its inventor, tech company OpenAI, has since introduced a specialized plug-in\u2014a tool ChatGPT can use when answering a query\u2014that allows it to do so. But that plug-in was not used in Milli\u00e8re’s demonstration. Instead he hypothesizes that the machine improvised a memory by harnessing its mechanisms for interpreting words according to their context\u2014a situation similar to how nature repurposes existing capacities for new functions.<\/p>\n

This impromptu ability demonstrates that LLMs develop an internal complexity that goes well beyond a shallow statistical analysis. Researchers are finding that these systems seem to achieve genuine understanding of what they have learned. In one study presented in May at the International Conference on Learning Representations, doctoral student Kenneth Li of Harvard University and his AI researcher colleagues\u2014Aspen K. Hopkins of the Massachusetts Institute of Technology; David Bau of Northeastern University; and Fernanda Vi\u00e9gas, Hanspeter Pfister and Martin Wattenberg, all at Harvard\u2014spun up their own smaller copy of the GPT neural network so they could study its inner workings. They trained it on millions of matches of the\u00a0board game Othello<\/a>\u00a0by feeding in long sequences of moves in text form. Their model became a nearly perfect player.<\/p>\n

To study how the neural network encoded information, they adopted a technique that Bengio and Guillaume Alain, also at the University of Montreal, devised in 2016. They created a miniature \u201cprobe\u201d network to analyze the main network layer by layer. Li compares this approach to neuroscience methods. \u201cThis is similar to when we put an electrical probe into the human brain,\u201d he says. In the case of the AI, the probe showed that its \u201cneural activity\u201d matched the representation of an Othello game board, albeit in a convoluted form. To confirm this, the researchers ran the probe in reverse to implant information into the network\u2014for instance, flipping one of the game’s black marker pieces to a white one. \u201cBasically we hack into the brain of these language models,\u201d Li says. The network adjusted its moves accordingly. The researchers concluded that it was playing Othello roughly like a human: by keeping a game board in its \u201cmind’s eye\u201d and using this model to evaluate moves. Li says he thinks the system learns this skill because it is the most parsimonious description of its training data. \u201cIf you are given a whole lot of game scripts, trying to figure out the rule behind it is the best way to compress,\u201d he adds.<\/p>\n

This ability to infer the structure of the outside world is not limited to simple game-playing moves; it also shows up in dialogue. Belinda Li (no relation to Kenneth Li), Maxwell Nye and Jacob Andreas, all at M.I.T., studied networks that played a text-based adventure game. They fed in sentences such as \u201cThe key is in the treasure chest,\u201d followed by \u201cYou take the key.\u201d Using a probe, they found that the networks encoded within themselves variables corresponding to \u201cchest\u201d and \u201cyou,\u201d each with the property of possessing a key or not, and updated these variables sentence by sentence. The system had no independent way of knowing what a box or key is, yet it picked up the concepts it needed for this task. \u201cThere is some representation of the state hidden inside of the model,\u201d Belinda Li says.<\/p>\n

Researchers marvel at how much LLMs are able to learn from text. For example, Pavlick and her then Ph.D. student Roma Patel found that these networks absorb color descriptions from Internet text and construct internal representations of color. When they see the word \u201cred,\u201d they process it not just as an abstract symbol but as a concept that has certain relations to maroon, crimson, fuchsia, rust, and so on. Demonstrating this was somewhat tricky. Instead of inserting a probe into a network, the researchers studied its response to a series of text prompts. To check whether it was merely echoing color relations from online references, they tried misdirecting the system by telling it that red is in fact green\u2014like the old philosophical thought experiment in which one person’s red is another person’s green. Rather than parroting back an incorrect answer, the system’s color evaluations changed appropriately to maintain the correct relations.<\/p>\n

Picking up on the idea that to perform its autocorrection function the system seeks the underlying logic of its training data, machine-learning researcher S\u00e9bastien Bubeck of Microsoft Research suggests that the wider the range of the data, the more general the rules the system will discover. \u201cMaybe we’re seeing such a huge jump because we have reached a diversity of data, which is large enough that the only underlying principle to all of it is that intelligent beings produced them,\u201d he says. \u201cAnd so the only way to explain all of the data is [for the model] to become intelligent.\u201d<\/p>\n

In addition to extracting the underlying meaning of language, LLMs can learn on the fly. In the AI field, the term \u201clearning\u201d is usually reserved for the computationally intensive process in which developers expose the neural network to gigabytes of data and tweak its internal connections. By the time you type a query into ChatGPT, the network should be fixed; unlike humans, it should not continue to learn. So it came as a surprise that LLMs do, in fact, learn from their users’ prompts\u2014an ability known as in-context learning. \u201cIt’s a different sort of learning that wasn’t really understood to exist before,\u201d says Ben Goertzel, founder of AI company SingularityNET.<\/p>\n

One example of how an LLM learns comes from the way humans interact with chatbots such as ChatGPT. You can give the system examples of how you want it to respond, and it will obey. Its outputs are determined by the last several thousand words it has seen. What it does, given those words, is prescribed by its fixed internal connections\u2014but the word sequence nonetheless offers some adaptability. Entire websites are devoted to \u201cjailbreak\u201d prompts that overcome the system’s \u201cguardrails\u201d\u2014restrictions that stop the system from telling users how to make a pipe bomb, for example\u2014typically by directing the model to pretend to be a system without guardrails. Some people use jailbreaking for sketchy purposes, yet others deploy it to elicit more creative answers. \u201cIt will answer scientific questions, I would say, better\u201d than if you just ask it directly, without the special jailbreak prompt, says William Hahn, co-director of the Machine Perception and Cognitive Robotics Laboratory at Florida Atlantic University. \u201cIt’s better at scholarship.\u201d<\/p>\n

Another type of in-context learning happens via \u201cchain of thought\u201d prompting, which means asking the network to spell out each step of its reasoning\u2014a tactic that makes it do better at logic or arithmetic problems requiring multiple steps. (But one thing that made Milli\u00e8re’s example so surprising is that the network found the Fibonacci number without any such coaching.)<\/p>\n

In 2022 a team at Google Research and the Swiss Federal Institute of Technology in Zurich\u2014Johannes von Oswald, Eyvind Niklasson, Ettore Randazzo, Jo\u00e3o Sacramento, Alexander Mordvintsev, Andrey Zhmoginov and Max Vladymyrov\u2014showed that in-context learning follows the same basic computational procedure as standard learning, known as gradient descent. This procedure was not programmed; the system discovered it without help. \u201cIt would need to be a learned skill,\u201d says Blaise Ag\u00fcera y Arcas, a vice president at Google Research. In fact, he thinks LLMs may have other latent abilities that no one has discovered yet. \u201cEvery time we test for a new ability that we can quantify, we find it,\u201d he says.<\/p>\n

Although LLMs have enough blind spots not to qualify as artificial general intelligence, or AGI\u2014the term for a machine that attains the resourcefulness of animal brains\u2014these emergent abilities suggest to some researchers that tech companies are closer to AGI than even optimists had guessed. \u201cThey’re indirect evidence that we are probably not that far off from AGI,\u201d Goertzel said in March at a conference on deep learning at Florida Atlantic University. OpenAI’s plug-ins have given ChatGPT a modular architecture a little like that of the human brain. \u201cCombining GPT-4 [the latest version of the LLM that powers ChatGPT] with various plug-ins might be a route toward a humanlike specialization of function,\u201d says M.I.T. researcher Anna Ivanova.<\/p>\n

At the same time, though, researchers worry the window may be closing on their ability to study these systems. OpenAI has not divulged the details of\u00a0how it designed and trained GPT-4<\/a>, in part because it is locked in competition with Google and other companies\u2014not to mention other countries. \u201cProbably there’s going to be less open research from industry, and things are going to be more siloed and organized around building products,\u201d says Dan Roberts, a theoretical physicist at M.I.T., who applies the techniques of his profession to understanding AI.<\/p>\n

And this lack of transparency does not just harm researchers, says Mitchell of the Santa Fe Institute. It also hinders efforts to understand the social impacts of the rush to adopt AI technology. \u201cTransparency about these models is the most important thing to ensure safety.\u201d<\/p>\n[\/vc_column_text][vc_column_text]\u00a9 2023 SCIENTIFIC AMERICAN, A DIVISION OF SPRINGER NATURE AMERICA, INC.[\/vc_column_text][vc_column_text]<\/p>\n

About the Author(s)<\/h3>\n

George Musser<\/a><\/strong>\u00a0is a contributing editor at\u00a0Scientific American<\/em>\u00a0and author of\u00a0Putting Ourselves Back in the Equation<\/em>\u00a0(Farrar, Straus and Giroux, 2023). Follow him on Mastodon\u00a0@gmusser@mastodon.social<\/a>, Bluesky\u00a0@gmusser.bsky.social<\/a>\u00a0and Threads\u00a0@georgemusserjr@threads.net<\/a>\u00a0<\/span><\/p>\n[\/vc_column_text][\/vc_column][vc_column width=”1\/4″ css=”.vc_custom_1538657003956{padding-top: 0px !important;padding-bottom: 0px !important;}”][vc_custom_heading text=”ISSUES” font_container=”tag:h3|font_size:12px|text_align:left|color:%23333333|line_height:18px” use_theme_fonts=”yes” el_class=”titlewidget”][vc_column_text el_class=”columnflex” css=”.vc_custom_1664404002021{margin-bottom: 40px !important;}”]

October 2023<\/a><\/p>\r\n

September 2023<\/a><\/p>\r\n

August 2023<\/a><\/p>[\/vc_column_text][vc_custom_heading text=”OTHER ARTICLES IN THIS ISSUE” font_container=”tag:h3|font_size:12px|text_align:left|color:%23333333|line_height:18px” use_theme_fonts=”yes” el_class=”titlewidget” css=”.vc_custom_1674576179369{margin-bottom: 15px !important;}”][vc_basic_grid post_type=”post” max_items=”-1″ style=”load-more” items_per_page=”4″ element_width=”12″ orderby=”rand” item=”538″ grid_id=”vc_gid:1692707962221-edbde508-e156-6″ taxonomies=”28″][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"Researchers are still struggling to understand how AI models trained to parrot Internet text can perform advanced tasks such as running code, playing games and trying to break up a marriage. ...more<\/a>","protected":false},"author":29,"featured_media":2773,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"inline_featured_image":false,"footnotes":""},"categories":[7],"tags":[28,30],"class_list":["post-2772","post","type-post","status-publish","format-standard","has-post-thumbnail","category-science-connect","tag-august-2023","tag-scientific-american"],"acf":[],"yoast_head":"\nHow AI Knows Things No One Told It - SH migration<\/title>\n<meta name=\"robots\" content=\"noindex, nofollow\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"How AI Knows Things No One Told It - SH migration\" \/>\n<meta property=\"og:description\" content=\"Researchers are still struggling to understand how AI models trained to parrot Internet text can perform advanced tasks such as running code, playing games and trying to break up a marriage.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/\" \/>\n<meta property=\"og:site_name\" content=\"SH migration\" \/>\n<meta property=\"article:published_time\" content=\"2023-07-31T23:01:37+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-08-22T13:14:37+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"590\" \/>\n\t<meta property=\"og:image:height\" content=\"393\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Lee Kettley\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Lee Kettley\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"1 minute\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/\"},\"author\":{\"name\":\"Lee Kettley\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/#\\\/schema\\\/person\\\/af9de08aa069306e77a209e7ffc24863\"},\"headline\":\"How AI Knows Things No One Told It\",\"datePublished\":\"2023-07-31T23:01:37+00:00\",\"dateModified\":\"2023-08-22T13:14:37+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/\"},\"wordCount\":389,\"image\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/wp-content\\\/uploads\\\/sites\\\/24\\\/2023\\\/08\\\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp\",\"keywords\":[\"August 2023\",\"Scientific American\"],\"articleSection\":[\"Science Connect\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/\",\"url\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/\",\"name\":\"How AI Knows Things No One Told It - SH migration\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/wp-content\\\/uploads\\\/sites\\\/24\\\/2023\\\/08\\\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp\",\"datePublished\":\"2023-07-31T23:01:37+00:00\",\"dateModified\":\"2023-08-22T13:14:37+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/#\\\/schema\\\/person\\\/af9de08aa069306e77a209e7ffc24863\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#primaryimage\",\"url\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/wp-content\\\/uploads\\\/sites\\\/24\\\/2023\\\/08\\\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp\",\"contentUrl\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/wp-content\\\/uploads\\\/sites\\\/24\\\/2023\\\/08\\\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp\",\"width\":590,\"height\":393},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/how-ai-knows-things-no-one-told-it\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"aa\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/#website\",\"url\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/\",\"name\":\"SH migration\",\"description\":\"Acclaimed Content, Global Reach, Diverse Delivery\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/#\\\/schema\\\/person\\\/af9de08aa069306e77a209e7ffc24863\",\"name\":\"Lee Kettley\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g\",\"caption\":\"Lee Kettley\"},\"url\":\"https:\\\/\\\/springerhealthplus.nl\\\/shmigrate\\\/author\\\/lee-kettleyspringer-com\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"How AI Knows Things No One Told It - SH migration","robots":{"index":"noindex","follow":"nofollow"},"og_locale":"en_US","og_type":"article","og_title":"How AI Knows Things No One Told It - SH migration","og_description":"Researchers are still struggling to understand how AI models trained to parrot Internet text can perform advanced tasks such as running code, playing games and trying to break up a marriage.","og_url":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/","og_site_name":"SH migration","article_published_time":"2023-07-31T23:01:37+00:00","article_modified_time":"2023-08-22T13:14:37+00:00","og_image":[{"width":590,"height":393,"url":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp","type":"image\/webp"}],"author":"Lee Kettley","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Lee Kettley","Est. reading time":"1 minute"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#article","isPartOf":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/"},"author":{"name":"Lee Kettley","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/#\/schema\/person\/af9de08aa069306e77a209e7ffc24863"},"headline":"How AI Knows Things No One Told It","datePublished":"2023-07-31T23:01:37+00:00","dateModified":"2023-08-22T13:14:37+00:00","mainEntityOfPage":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/"},"wordCount":389,"image":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#primaryimage"},"thumbnailUrl":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp","keywords":["August 2023","Scientific American"],"articleSection":["Science Connect"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/","url":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/","name":"How AI Knows Things No One Told It - SH migration","isPartOf":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/#website"},"primaryImageOfPage":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#primaryimage"},"image":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#primaryimage"},"thumbnailUrl":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp","datePublished":"2023-07-31T23:01:37+00:00","dateModified":"2023-08-22T13:14:37+00:00","author":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/#\/schema\/person\/af9de08aa069306e77a209e7ffc24863"},"breadcrumb":{"@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#primaryimage","url":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp","contentUrl":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-content\/uploads\/sites\/24\/2023\/08\/0AC797F3-92E4-44EC-950D4D40674DA9BD_source.webp","width":590,"height":393},{"@type":"BreadcrumbList","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/how-ai-knows-things-no-one-told-it\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/springerhealthplus.nl\/shmigrate\/"},{"@type":"ListItem","position":2,"name":"aa"}]},{"@type":"WebSite","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/#website","url":"https:\/\/springerhealthplus.nl\/shmigrate\/","name":"SH migration","description":"Acclaimed Content, Global Reach, Diverse Delivery","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/springerhealthplus.nl\/shmigrate\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/springerhealthplus.nl\/shmigrate\/#\/schema\/person\/af9de08aa069306e77a209e7ffc24863","name":"Lee Kettley","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/03dcff2dc955be0ae0163db297d8149d9df34f433a4f50e791447790cc161f0b?s=96&d=mm&r=g","caption":"Lee Kettley"},"url":"https:\/\/springerhealthplus.nl\/shmigrate\/author\/lee-kettleyspringer-com\/"}]}},"menu_order":0,"_links":{"self":[{"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/posts\/2772","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/users\/29"}],"replies":[{"embeddable":true,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/comments?post=2772"}],"version-history":[{"count":0,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/posts\/2772\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/media\/2773"}],"wp:attachment":[{"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/media?parent=2772"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/categories?post=2772"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/springerhealthplus.nl\/shmigrate\/wp-json\/wp\/v2\/tags?post=2772"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}