{"id":94,"date":"2023-07-06T02:49:13","date_gmt":"2023-07-06T02:49:13","guid":{"rendered":"https:\/\/blog.gwadej.org\/micro-musings\/?p=94"},"modified":"2023-07-06T02:49:13","modified_gmt":"2023-07-06T02:49:13","slug":"mistaking-performance-for-competence","status":"publish","type":"post","link":"https:\/\/blog.gwadej.org\/micro-musings\/2023\/07\/mistaking-performance-for-competence\/","title":{"rendered":"Mistaking Performance for Competence"},"content":{"rendered":"\n<p><a rel=\"noreferrer noopener\" href=\"https:\/\/spectrum.ieee.org\/gpt-4-calm-down\" target=\"_blank\">Just Calm Down About GPT-4 Already<\/a> from Rodney Brooks talks about his long-term study of AI and how he is not really excited about this new technology.<\/p>\n\n\n\n<p>The title of this post is his observation about humans mistaking AI.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote\">\n<p>Rapid and pivotal advances in technology have a way of unsettling people, because they can reverberate mercilessly, sometimes, through business, employment, and cultural spheres. And so it is with the current shock and awe over large language models, such as <a href=\"https:\/\/openai.com\/product\/gpt-4\" rel=\"noreferrer noopener\" target=\"_blank\">GPT-4<\/a> from <a href=\"https:\/\/openai.com\/\" rel=\"noreferrer noopener\" target=\"_blank\">OpenAI<\/a>.<\/p>\n\n\n\n<p>It\u2019s a textbook example of the mixture of amazement and, especially, anxiety that often accompanies a tech triumph. And we\u2019ve been here many times, says <a href=\"https:\/\/www.csail.mit.edu\/person\/rodney-brooks\" rel=\"noreferrer noopener\" target=\"_blank\">Rodney Brooks<\/a>. Best known as a <a href=\"https:\/\/spectrum.ieee.org\/topic\/robotics\/\">robotics<\/a> researcher, academic, and entrepreneur, Brooks is also an authority on AI: he directed the <a href=\"https:\/\/www.csail.mit.edu\/\" rel=\"noreferrer noopener\" target=\"_blank\">Computer Science and Artificial Intelligence Laboratory<\/a> at MIT until 2007, and held faculty positions at Carnegie Mellon and Stanford before that. Brooks, who is now working on his third robotics startup, <a href=\"https:\/\/www.robust.ai\/\" rel=\"noreferrer noopener\" target=\"_blank\">Robust.AI<\/a>, has written <a href=\"http:\/\/people.csail.mit.edu\/brooks\/publications.html\" rel=\"noreferrer noopener\" target=\"_blank\">hundreds of articles and half a dozen books<\/a> and was featured in the motion picture <a href=\"https:\/\/vimeo.com\/454131115\" rel=\"noreferrer noopener\" target=\"_blank\"><em>Fast, Cheap &amp; Out of Control<\/em><\/a>. He is a rare technical leader who has had a stellar career in business and in academia and has still found time to engage with the popular culture through books, <a href=\"https:\/\/rodneybrooks.com\/my-ieee-spectrum-columns-and-articles\/\" rel=\"noreferrer noopener\" target=\"_blank\">popular articles<\/a>, <a href=\"https:\/\/www.ted.com\/speakers\/rodney_brooks\" rel=\"noreferrer noopener\" target=\"_blank\">TED Talks<\/a>, and other venues.<\/p>\n\n\n\n<p>\u201cIt gives an answer with complete confidence, and I sort of believe it. And half the time, it\u2019s completely wrong.\u201d<br>\u2014Rodney Brooks, Robust.AI<\/p>\n\n\n\n<p><em><a href=\"https:\/\/spectrum.ieee.org\/\">IEEE Spectrum<\/a><\/em> caught up with Brooks at the recent <a href=\"https:\/\/corporate-awards.ieee.org\/event\/2023-vic-summit-honors-ceremony-gala\/\" rel=\"noreferrer noopener\" target=\"_blank\">Vision, Innovation, and Challenges Summit<\/a>, where he was being honored with the 2023 <a href=\"https:\/\/corporate-awards.ieee.org\/award\/ieee-founders-medal\/\" rel=\"noreferrer noopener\" target=\"_blank\">IEEE Founders Medal<\/a>. He spoke about this moment in AI, which he doesn\u2019t regard with as much apprehension as some of his peers, and about his latest startup, which is working on robots for medium-size warehouses.<\/p>\n\n\n\n<p><strong>Rodney Brooks on\u2026<\/strong><\/p>\n\n\n\n<ul>\n<li><a href=\"https:\/\/spectrum.ieee.org\/gpt-4-calm-down#agi\">Will GPT-4 and other large language models lead to an artificial general intelligence in the foreseeable future?<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/spectrum.ieee.org\/gpt-4-calm-down#valuations\">Will companies marketing large language models ever justify the enormous valuations some of these companies are now enjoying?<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/spectrum.ieee.org\/gpt-4-calm-down#level5\">When are we going to have full (level-5) self-driving cars?<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/spectrum.ieee.org\/gpt-4-calm-down#warehouse\">What are the most attractive opportunities now in warehouse robotics?<\/a><\/li>\n<\/ul>\n\n\n\n<p id=\"agi\"><strong>You wrote a famous article in 2017, \u201c<a href=\"https:\/\/rodneybrooks.com\/the-seven-deadly-sins-of-predicting-the-future-of-ai\/\" rel=\"noreferrer noopener\" target=\"_blank\">The Seven Deadly Sins of AI Prediction.<\/a>\u201c You said then that you wanted an artificial general intelligence to exist\u2014in fact, you said it had always been your personal motivation for working in robotics and AI. But you also said that AGI research wasn\u2019t doing very well at that time at solving the basic problems that had remained intractable for 50 years. My impression now is that you do not think the emergence of <a href=\"https:\/\/spectrum.ieee.org\/tag\/gpt-4\">GPT-4<\/a> and other large language models means that an AGI will be possible within a decade or so.<\/strong><\/p>\n\n\n\n<p><strong>Rodney Brooks:<\/strong> You\u2019re exactly right. And by the way, GPT-3.5 guessed right\u2014I asked it about me, and it said I was a skeptic about it. But that doesn\u2019t make it an AGI.<\/p>\n\n\n\n<p>The large language models are a little surprising. I\u2019ll give you that. And I think what they say, interestingly, is how much of our language is very much rote, R-O-T-E, rather than generated directly, because it can be collapsed down to this set of parameters. But in that \u201cSeven Deadly Sins\u201d article, I said that one of the deadly sins was how we humans mistake performance for competence.<\/p>\n\n\n\n<p>If I can just expand on that a little. When we see a person with some level performance at some intellectual thing, like describing what\u2019s in a picture, for instance, from that performance, we can generalize about their competence in the area they\u2019re talking about. And we\u2019re really good at that. Evolutionarily, it\u2019s something that we ought to be able to do. We see a person do something, and we know what else they can do, and we can make a judgement quickly. But our models for generalizing from a performance to a competence don\u2019t apply to AI systems.<\/p>\n\n\n\n<p>The example I used at the time was, I think it was a <a href=\"https:\/\/www.nytimes.com\/2014\/11\/18\/science\/researchers-announce-breakthrough-in-content-recognition-software.html\" target=\"_blank\" rel=\"noreferrer noopener\">Google program labeling an image of people playing Frisbee<\/a> in the park. And if a person says, \u201cOh, that\u2019s a person playing Frisbee in the park,\u201d you would assume you could ask him a question, like, \u201cCan you eat a Frisbee?\u201d And they would know, of course not; it\u2019s made of plastic. You\u2019d just expect they\u2019d have that competence. That they would know the answer to the question, \u201cCan you play Frisbee in a snowstorm? Or, how far can a person throw a Frisbee? Can they throw it 10 miles? Can they only throw it 10 centimeters?\u201d You\u2019d expect all that competence from that one piece of performance: a person saying, \u201cThat\u2019s a picture of people playing Frisbee in the park.\u201d<\/p>\n\n\n\n<p>What the large language models are good at is saying what an answer should <em>sound like<\/em>, which is different from what an answer should <em>be<\/em>.<\/p>\n<cite>Rodney Brooks, Robust.AI<\/cite><\/blockquote>\n","protected":false},"excerpt":{"rendered":"<p>Just Calm Down About GPT-4 Already from Rodney Brooks talks about his long-term study of AI and how he is not really excited about this new technology. The title of this post is his observation about humans mistaking AI.<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[33],"tags":[11,65,5,66],"_links":{"self":[{"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/posts\/94"}],"collection":[{"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/comments?post=94"}],"version-history":[{"count":1,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/posts\/94\/revisions"}],"predecessor-version":[{"id":95,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/posts\/94\/revisions\/95"}],"wp:attachment":[{"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/media?parent=94"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/categories?post=94"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.gwadej.org\/micro-musings\/wp-json\/wp\/v2\/tags?post=94"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}