{"id":38117,"date":"2024-04-29T11:06:55","date_gmt":"2024-04-29T11:06:55","guid":{"rendered":"https:\/\/news.talkwithrattan.com\/index.php\/2024\/04\/29\/heres-why-ai-like-chatgpt-probably-wont-reach-humanlike-understanding\/"},"modified":"2024-04-29T11:06:55","modified_gmt":"2024-04-29T11:06:55","slug":"heres-why-ai-like-chatgpt-probably-wont-reach-humanlike-understanding","status":"publish","type":"post","link":"https:\/\/news.talkwithrattan.com\/index.php\/2024\/04\/29\/heres-why-ai-like-chatgpt-probably-wont-reach-humanlike-understanding\/","title":{"rendered":"Here\u2019s why AI like ChatGPT probably won\u2019t reach humanlike understanding"},"content":{"rendered":"<div style=\"text-align:center\"><img loading=\"lazy\" decoding=\"async\" width=\"1440\" height=\"1920\" src=\"https:\/\/i1.wp.com\/www.snexplores.org\/wp-content\/uploads\/2024\/04\/1440_deep_blue_inline.jpg?resize=1440,1920&amp;ssl=1\" class=\"attachment-post-thumbnail size-post-thumbnail wp-post-image\" alt=\"Here\u2019s why AI like ChatGPT probably won\u2019t reach humanlike understanding\" title=\"Here\u2019s why AI like ChatGPT probably won\u2019t reach humanlike understanding\" \/><\/div><p> <br \/>\n<\/p>\n<p>If you ask ChatGPT whether it thinks like a human, this chatbot will tell you that it doesn\u2019t. \u201cI can process and understand language to a certain extent,\u201d ChatGPT writes. But \u201cmy understanding is based on patterns in data, [not] humanlike comprehension.\u201d<\/p>\n<p>Still, talking to this <a href=\"https:\/\/www.snexplores.org\/article\/scientists-say-artificial-intelligence-definition-pronunciation\">artificial intelligence<\/a>, or AI, system can sometimes <em>feel<\/em> like talking to a human. A pretty smart, talented person at that. ChatGPT can answer questions about math or history on demand \u2014 and in a lot of different languages. It can crank out stories and computer code. And other similarly \u201cgenerative\u201d AI models can produce artwork and videos from scratch.<\/p>\n<aside class=\"wp-block-sciencenews-inline-related-post alignleft\">\n<h4><a href=\"https:\/\/www.snexplores.org\/article\/lets-learn-about-artificial-intelligence\">Let\u2019s learn about artificial intelligence<\/a><\/h4>\n<\/aside>\n<p>\u201cThese things seem really smart,\u201d said Melanie Mitchell. She\u2019s a computer scientist at the Santa Fe Institute in New Mexico. <a href=\"https:\/\/www.aaas.org\/news\/scientific-pioneers-break-down-walls-address-pressing-societal-issues-aaas-annual-meeting\" rel=\"noopener\">She spoke at the annual meeting of the American Association for the Advancement of Science<\/a>. It was held in Denver, Colo., in February.<\/p>\n<p>AI\u2019s increasing \u201csmarts\u201d have a lot of people worried. They fear generative AI could take people\u2019s jobs \u2014 or take over the world. But Mitchell and other experts think those fears are overblown. At least, for now.<\/p>\n<p>The problem, those experts argue, is just what ChatGPT says. Today\u2019s most impressive AI still doesn\u2019t truly understand what it is saying or doing the way a human would. And that puts some hard limits on its abilities.<\/p>\n<h4 class=\"wp-block-heading\">Concerns about AI are not new<\/h4>\n<p>People have worried for decades that machines are getting too smart. This fear dates back to at least 1997. That\u2019s when the computer Deep Blue defeated world chess champion Garry Kasparov.<\/p>\n<p>At that time, though, it was still easy to show that AI failed miserably at many things we do well. Sure, a computer could play a mean game of chess. But could it diagnose disease? Transcribe speech? Not very well. In many key areas, humans remained supreme.<\/p>\n<div class=\"wp-block-image  has-alignleft\">\n<figure class=\"alignleft size-full\"><figcaption class=\"wp-element-caption\"><span class=\"caption wp-caption-3139194\">A Deep Blue computer similar to this one defeated world chess champion Garry Kasparov in 1997.<\/span><span class=\"credit wp-credit-3139194\">Christina Xu\/Wikimedia Commons (<a href=\"https:\/\/creativecommons.org\/licenses\/by\/2.0\/deed.en\" rel=\"noopener\">CC BY 2.0 DEED<\/a>)<\/span><\/figcaption><\/figure>\n<\/div>\n<p>About a decade ago, that began to change.<\/p>\n<p>Computer brains \u2014 known as neural networks \u2014 got a huge boost from a new technique called&nbsp;<a href=\"https:\/\/www.ibm.com\/topics\/deep-learning\" target=\"_blank\" rel=\"noreferrer noopener\">deep learning<\/a>. This is a powerful type of machine learning. In <a href=\"https:\/\/snexplores.org\/article\/scientists-say-machine-learning\" rel=\"noopener\">machine learning<\/a>, computers master skills through practice or looking at examples.<\/p>\n<p>Suddenly, thanks to deep learning, computers rivaled humans at many tasks. Machines could identify images, read signs and enhance photographs. They could even reliably convert speech to text.<\/p>\n<p>Yet those abilities had their limits. For one thing, deep-learning neural networks could be easy to trick. A few stickers placed on a stop sign, for example, made an AI think the sign said \u201cSpeed Limit 80.\u201d Such \u201csmart\u201d computers also needed extensive training. To pick up each new skill, they had to view tons of examples of what to do.<\/p>\n<p>So deep learning produced AI models that were excellent at very specific jobs. But those systems couldn\u2019t adapt that expertise very well to new tasks. You couldn\u2019t, for instance, use an English-to-Spanish AI translator for help on your French homework.<\/p>\n<p>But now, things are changing again.<\/p>\n<p>\u201cWe\u2019re in a new era of AI,\u201d Mitchell says. \u201cWe\u2019re beyond the deep-learning revolution of the 2010s. And we\u2019re now in the era of generative AI of the 2020s.\u201d<\/p>\n<div class=\"wp-block-image  has-aligncenter\">\n<figure class=\"aligncenter size-full\"><img decoding=\"async\" loading=\"lazy\" width=\"1440\" height=\"869\" src=\"https:\/\/www.snexplores.org\/wp-content\/uploads\/2024\/04\/1440_AI_chatbot_inline1.jpg\" alt=\"a person holds their smartphone in their hands, and above the phone a graphic shows the conversation they're having with a chatbot; the chatbot says &quot;can I help you?&quot; and the person says &quot;yes, I need help&quot;\" class=\"wp-image-3139195\" \/><figcaption class=\"wp-element-caption\"><span class=\"caption wp-caption-3139195\">The AI behind tools like ChatGPT can do a lot of things \u2014 from summarizing text to spell-checking essays. That can make it tempting to use these tools for help on homework or other tasks. But proceed with caution: Because an AI system like ChatGPT doesn\u2019t truly understand what it\u2019s doing the way a human would, it\u2019s also been known to make up information.  <\/span><span class=\"credit wp-credit-3139195\">Krongkaew\/Moment\/Getty Images<\/span><\/figcaption><\/figure>\n<\/div>\n<h4 class=\"wp-block-heading\">This gen-AI era<\/h4>\n<p>Generative AI systems are those that can produce text, images or other content on demand. This type of AI can generate many things that long seemed to require human creativity. That includes everything from brainstorming ideas to writing poems.<\/p>\n<p>Many of these abilities stem from large language models \u2014 LLMs, for short. ChatGPT is one example of tech based on LLMs. Such language models are said to be \u201clarge\u201d because they are trained on huge amounts of data. Essentially, they study everything on the internet, which includes scanned copies of countless print books.<\/p>\n<p>\u201cLarge\u201d can also refer to the number of different types of things LLMs can \u201clearn\u201d in their reading. These models don\u2019t just learn words. They also pick up phrases, symbols and math equations.<\/p>\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\">\n<div class=\"wp-block-embed__wrapper\">\n<\/div><figcaption class=\"wp-element-caption\">Here\u2019s a quick rundown of what large language models are and what kinds of things they can do.<\/figcaption><\/figure>\n<p>By learning patterns in how the building blocks of language are combined, LLMs can predict in what order words should go. That helps the models write sentences and answer questions. Basically, an LLM calculates the odds that one word should follow another in a given context.<\/p>\n<p>This has allowed LLMs to do things such as writing in the style of any author and solving riddles.<\/p>\n<p>Some researchers have suggested that when LLMs accomplish these feats, they understand what they\u2019re doing. Those researchers think LLMs can reason like people or could even become conscious in some sense.<\/p>\n<p>But Mitchell and others insist that LLMs do not (yet) really understand the world. At least, not as humans do.<\/p>\n<h4 class=\"wp-block-heading\">Narrow-minded AI<\/h4>\n<p>Mitchell and her colleague Martha Lewis recently exposed one major limit of LLMs. Lewis studies language and concepts at the University of Bristol in England. The pair shared their work at arXiv.org. (Studies posted there have usually not yet been vetted by other scientists.)<\/p>\n<p>LLMs still do not match humans\u2019 ability to adapt a skill to a new situation, their new paper shows. Consider this letter-string problem. You start with one string of letters: ABCD. Then, you get a second string of letters: ABCE.<\/p>\n<p>Most humans can see the difference between the two strings. The final letter in the first string is replaced with the next letter of the alphabet in the second string. So when humans are shown a different string of letters, such as IJKL, they can guess what the second string should be: IJKM.<\/p>\n<p>Most LLMs can solve this problem, too. That\u2019s to be expected. The models have, after all, been well trained on the English alphabet.<\/p>\n<p>.cheat-sheet-cta {<br \/>\n  border: 1px solid #ffffff;<br \/>\n  margin-top: 20px;<br \/>\n  background-image: url(&#8220;https:\/\/www.snexplores.org\/wp-content\/uploads\/2022\/12\/cta-module@2x-2048&#215;239-1.png&#8221;);<br \/>\n  padding: 10px;<br \/>\n  clear: both;<br \/>\n}<\/p>\n<div class=\"wp-block-group cheat-sheet-cta is-layout-flow\">\n<div class=\"wp-block-group__inner-container\">\n<h2 class=\"wp-block-heading has-text-align-center\">Do you have a science question? We can help!<\/h2>\n<p class=\"has-text-align-center\"><a href=\"https:\/\/forms.gle\/YbhPosFTMqjbSNnV7\" target=\"_blank\" rel=\"noreferrer noopener\">Submit your question here<\/a>, and we might answer it an upcoming issue of&nbsp;<em>Science News Explores<\/em><\/p>\n<\/div>\n<\/div>\n<p>But say you pose the problem with a different alphabet. Perhaps you jumble up the letters in our alphabet to be in a different order. Or you use symbols instead of letters.&nbsp;Humans are still very good at solving letter-string problems. But LLMs usually fail. They are not able to take the concepts they learned with one alphabet and apply them to another. All the GPT models tested by Mitchell and Lewis struggled with these kinds of problems.<\/p>\n<p>Other similar tasks also show that LLMs don\u2019t do well in situations they weren\u2019t trained for. For that reason, Mitchell doesn\u2019t believe they show what humans would call \u201cunderstanding\u201d of the world.<\/p>\n<h4 class=\"wp-block-heading\">The importance of understanding<\/h4>\n<p>\u201cBeing reliable and doing the right thing in a new situation is, in my mind, the core of what understanding actually means,\u201d Mitchell said at the AAAS meeting.<\/p>\n<p>Human understanding, she says, is based on \u201cconcepts.\u201d These are mental <a href=\"https:\/\/www.snexplores.org\/article\/scientists-say-model-definition-pronunciation\">models<\/a> of things like categories, situations and events. Concepts allow people to determine cause and effect. They also help people predict the likely results of different actions. And people can do this even in situations they haven\u2019t seen before.<\/p>\n<div class=\"wp-block-image  has-alignright\">\n<figure class=\"alignright size-full\"><img decoding=\"async\" loading=\"lazy\" width=\"1440\" height=\"1337\" src=\"https:\/\/www.snexplores.org\/wp-content\/uploads\/2024\/04\/1440_limits_AI_understanding_inline2.jpg\" alt=\"glowing blue lines representing circuitry form the shape of a human brain against a gray background\" class=\"wp-image-3139196\" \/><figcaption class=\"wp-element-caption\"><span class=\"caption wp-caption-3139196\">AI models might someday achieve a truly intelligent understanding of the world. But machine understanding may not look anything like human understanding. <\/span><span class=\"credit wp-credit-3139196\">Boris SV\/Moment\/Getty Images<\/span><\/figcaption><\/figure>\n<\/div>\n<p>\u201cWhat\u2019s really remarkable about people \u2026 is that we can abstract our concepts to new situations,\u201d Mitchell said.<\/p>\n<p>She does not deny that AI might someday reach a level of intelligent understanding similar to humans. But machine understanding may turn out to be different from human understanding, she adds. Nobody knows what sort of technology might achieve that understanding. But if it\u2019s anything like human understanding, it will probably not be based on LLMs.<\/p>\n<p>After all, LLMs learn in a way opposite to humans. These models start out learning language. Then, they try to use that knowledge to grasp abstract concepts. Human babies, meanwhile, learn concepts first and then the language to describe them.<\/p>\n<p>So talking to ChatGPT may sometimes feel like talking to a friend, teammate or tutor. But the computerized number-crunching behind it is still nothing like a human mind.<\/p>\n<p><br \/>\n<br \/><a href=\"https:\/\/www.snexplores.org\/article\/ai-limits-chatbot-artificial-intelligence\">Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>If you ask ChatGPT whether it thinks like a human, this chatbot will tell you that it doesn\u2019t. \u201cI can process and understand language to a certain extent,\u201d ChatGPT writes. But \u201cmy understanding is based on patterns in data, [not] humanlike comprehension.\u201d Still, talking to this artificial intelligence, or AI, system can sometimes feel like [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":38118,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"tdm_status":"","tdm_grid_status":"","fifu_image_url":"https:\/\/www.snexplores.org\/wp-content\/uploads\/2024\/04\/1440_deep_blue_inline.jpg","fifu_image_alt":"","footnotes":""},"categories":[606],"tags":[10240,671,25042,4591,4478,2829],"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/posts\/38117"}],"collection":[{"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/comments?post=38117"}],"version-history":[{"count":1,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/posts\/38117\/revisions"}],"predecessor-version":[{"id":38119,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/posts\/38117\/revisions\/38119"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/media\/38118"}],"wp:attachment":[{"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/media?parent=38117"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/categories?post=38117"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/news.talkwithrattan.com\/index.php\/wp-json\/wp\/v2\/tags?post=38117"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}