Is AI really as advanced as we think it is? Are we overestimating its capabilities?<\/p>\n\n\n\n
In their book Rebooting AI<\/em>, Gary Marcus and Ernest Davis explore the gap between AI hype and reality. They discuss how narrow AI<\/a> differs from the strong AI we often imagine. The authors also explain why we tend to overestimate AI’s abilities.<\/p>\n\n\n\n Keep reading to discover the truth behind AI’s current capabilities and the potential risks of relying too heavily on narrow AI systems.<\/p>\n\n\n\n\n\n\n\n In their book, Marcus and Davis address how public perception of artificial intelligence is skewed. The computer industry, science fiction, and the media have contributed to the AI hype, priming<\/a> the public to imagine \u201cstrong AI\u201d\u2014computer systems that actually think and can do so much faster and more powerfully than humans. Instead, what AI developers have delivered is \u201cnarrow AI\u201d\u2014systems trained to do one specific task while having no more awareness of the larger world than a doorknob can understand what a door is for. The authors explain how AI\u2019s capabilities are currently being oversold and why software engineers and the public at large are susceptible to overestimating narrow AI\u2019s capabilities.\u00a0<\/p>\n\n\n\n (Shortform note: We cover two levels of AI, though some software engineers now divide them into three: narrow, general, and strong<\/a>. Narrow, or \u201cweak,\u201d AI is trained to perform specific tasks<\/a>, such as chatbots that mimic human conversation<\/a> or self-driving systems in cars<\/a>. General AI will be able to mimic the human mind itself<\/a> in terms of learning, comprehension, and perhaps even consciousness. Strong, or \u201csuper,\u201d AI will be the level of artificial intelligence that exceeds the human mind\u2019s capabilities<\/a> and can think in ways we can\u2019t even imagine. Some computer scientists, including Marcus and Davis, treat general and strong AI as the same thing.)<\/p>\n\n\n\n The AI systems making an impact<\/a> today are far from the benevolent androids or evil computer overlords depicted in science fiction. Instead, Davis and Marcus describe modern AIs as hyper-focused morons who don\u2019t understand that a world exists beyond the tasks they\u2019re trained on. Yet the public\u2019s confusion is understandable\u2014tech companies like to garner attention by billing every incremental step forward as a giant leap for AI,<\/strong> while the press eagerly overhypes AI progress in search of more readers and social media clicks. <\/p>\n\n\n\n (Shortform note: One can generally expect tech companies to paint their progress in a generous light, but contrary to Davis and Marcus\u2019s claims, the news media doesn\u2019t always follow suit. In 2023, when the electric car manufacturer Tesla released a video demonstrating its self-driving AI system<\/a>, some outlets criticized Tesla\u2019s driving test<\/a> for its ideal\u2014rather than realistic\u2014road conditions. Earlier that year, The Washington Post<\/em> reported that Tesla\u2019s AI driver-assist system resulted in more accidents than human drivers<\/a> without AI support. Thanks to media scrutiny and legal pressure, Tesla recalled more than 2 million cars<\/a> due to faulty self-driving AI.)<\/p>\n\n\n\n To be fair, technology has made great advances in using big data to drive machine learning. Marcus and Davis take care to point out that we haven\u2019t built machines that <\/strong>understand <\/em><\/strong>the data they use and adapt to real-world settings,<\/strong> and the authors list some reasons why this is sometimes hard to see. The first is our tendency to project human qualities on objects that don\u2019t have them\u2014such as claiming that your car gets \u201ccranky\u201d in cold weather, or imagining that your house\u2019s bad plumbing has a vendetta against you. When a computer does things that were once uniquely human, such as answering a question or giving you driving directions, it\u2019s easy to mistake digital abilities for actual thought.<\/p>\n\n\n\n (Shortform note: Our tendency to anthropomorphize objects doesn\u2019t end with computers, or even machines. We do this because our brains are geared to interpret social cues very quickly<\/a>\u2014therefore, if your mind projects feelings onto an object, you\u2019re better able to easily fit that object into your mental model of the world. This tendency may explain the broad appeal of Marie Kondo\u2019s (The Life-Changing Magic of Tidying Up<\/em><\/a>) system for organizing your belongings, which involves anthropomorphizing everyday objects<\/a>. A study on the irrational humanization of computers that Marcus and Davis caution against shows that people do it unconsciously<\/a>, leading to potentially faulty judgments about whether or not a computer\u2019s outputs should be trusted.)<\/p>\n\n\n\n The second reason that Davis and Marcus say that modern AI\u2019s abilities are overstated is that developers tend to mistakenly believe that progress on small computing challenges equates to progress toward larger goals.<\/strong> Much of the work on narrow AI has been done in environments with clear rules and boundaries, such as teaching a computer to play chess or a robot to climb a set of stairs. Unfortunately, these situations don\u2019t map onto the real world, where the number of variables and shifting conditions are essentially infinite. No matter how many obstacles you teach a stair-climbing robot to avoid under laboratory conditions, that can never match the infinitude of problems it might stumble upon in an everyday house.<\/p>\n\n\n\n (Shortform note: When AI developers overestimate their progress in the way that Davis and Marcus say they do, this translates into unrealistic expectations<\/a> from consumers using AI in the real world. For instance, despite warnings about the fallibility of language-generating AI<\/a>, two New York attorneys used a legal brief written by ChatGPT<\/a> in an actual federal court. The AI-generated brief referenced several nonexistent court cases, resulting in sanctions against the attorneys<\/a>, who claimed to have thought that the generative AI was a valid search engine<\/a> whose responses were accurate. Despite this dramatic failure, some tech advocates are still hyping language-generating AI<\/a> as a useful tool in legal writing.)<\/p>\n\n\n\n So far, according to Marcus and Davis, AI developers have largely ignored the issue of how to train AI to cope in situations it\u2019s not programmed for, such as setting the stair-climbing robot\u2019s stairs on fire. This is because, for most of its history, artificial intelligence has been used in situations where the cost of failure is usually low, such as recommending books based on your reading history or calculating the best route to work based on current traffic conditions. The authors warn that if we start to use narrow AI in situations where the cost of failure is high, it will put people\u2019s health and safety at risk<\/strong> in ways that no one will be able to predict.<\/p>\n\n\n\n (Shortform note: Marcus and Davis\u2019s predictions for how AI will endanger lives are already coming true, even in a field as seemingly innocuous as using AI to generate text. For instance, the New York Mycological Society has issued a warning to consumers<\/a> about AI-generated guides to edible mushrooms, some of which may\u2014by way of computer error and lack of human oversight\u2014recommend varieties of mushrooms that are deadly<\/a>. Amazon\u2019s self-publishing platform, Kindle Direct Publishing, has seen a flood of AI-generated books,<\/a> many of which are fraudulent attempts to mimic works by established authors and experts<\/a>, including the aforementioned mushroom foraging<\/a> guides.)<\/p>\n","protected":false},"excerpt":{"rendered":" Is AI really as advanced as we think it is? Are we overestimating its capabilities? In their book Rebooting AI, Gary Marcus and Ernest Davis explore the gap between AI hype and reality. They discuss how narrow AI differs from the strong AI we often imagine. The authors also explain why we tend to overestimate AI’s abilities. Keep reading to discover the truth behind AI’s current capabilities and the potential risks of relying too heavily on narrow AI systems.<\/p>\n","protected":false},"author":9,"featured_media":128539,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[160,24],"tags":[1570],"class_list":["post-128531","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-science","category-society","tag-rebooting-ai","","tg-column-two"],"yoast_head":"\nThe Hype About AI<\/h2>\n\n\n\n
AI Fantasy Versus Reality<\/h3>\n\n\n\n
Shortsightedness in AI Development<\/h3>\n\n\n\n