{"id":1744,"date":"2022-04-17T09:07:32","date_gmt":"2022-04-17T07:07:32","guid":{"rendered":"https:\/\/hinterdemnebel.de\/?p=1744"},"modified":"2022-04-18T18:10:23","modified_gmt":"2022-04-18T16:10:23","slug":"palm-und-dall-e-2-zwei-warnschuesse","status":"publish","type":"post","link":"https:\/\/hinterdemnebel.de\/index.php\/2022\/04\/17\/palm-und-dall-e-2-zwei-warnschuesse\/","title":{"rendered":"PaLM und DALL-E-2: Zwei Warnsch\u00fcsse"},"content":{"rendered":"\n<p>Die exponentielle Entwicklung technologischen Fortschritts war f\u00fcr mich noch nie so deutlich wie in den letzten Jahren auf dem Gebiet der KI. Dinge, die vor 10 Jahren unm\u00f6glich schienen, sind seit 5 Jahren m\u00f6glich. Und nur 2 oder 3 Jahre nach ihrem ersten Auftreten sind ihre Leistungen schon nahezu unglaublich. W\u00e4hrend der letzten Monate hat sich so viel Interessantes getan, dass ich sowieso schon eine Zusammenfassung erstellen wollte. Aber der Beginn des Aprils hat noch einmal alles \u00fcbertroffen.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Sprachmodell PaLM<\/h2>\n\n\n\n<p>Kurz: Ein Sprachmodell wurde an gro\u00dfen Textmengen trainiert, um Kontexte verstehen zu k\u00f6nnen, Weltwissen zu erlangen und f\u00fcr Eingaben passende Ausgaben zu erzeugen. Aufgrund des menschlichen Ursprungs der Texte, an denen es trainiert wurde, wirken die Antworten nicht nur menschlich, sondern verwenden auch menschliche Logik (mehr dazu in <a href=\"https:\/\/hinterdemnebel.de\/index.php\/2020\/08\/04\/gpt-3-antwortet-auf-philosophische-gedanken-ueber-ihn-selbst\/\">GPT-3 antwortet auf philosophische Gedanken \u00fcber ihn selbst<\/a>). Das Ergebnis wirkt schon fast wie echte Intelligenz und hat auf jeden Fall einen hohen Nutzwert.<\/p>\n\n\n\n<p>PaLM ist ein Sprachmodel mit 540 Billionen Parametern, \u00fcber das die Forscher <a href=\"https:\/\/storage.googleapis.com\/pathways-language-model\/PaLM-paper.pdf\">einleitend schreiben<\/a>:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p>We demonstrate continued benefits of scaling by achieving state-ofthe-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a<br>number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned stateof-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the<br>recently released BIG-bench benchmark.<\/p><cite><a href=\"https:\/\/storage.googleapis.com\/pathways-language-model\/PaLM-paper.pdf\">https:\/\/storage.googleapis.com\/pathways-language-model\/PaLM-paper.pdf<\/a><\/cite><\/blockquote>\n\n\n\n<p>Unabh\u00e4ngig davon, ob man ein Sprachmodell als echte KI bezeichnen will: Es \u00fcbertrifft bereits die durchschnittliche menschliche Leistung im <a href=\"https:\/\/github.com\/google\/BIG-bench\">BIG-bench<\/a> benchmark. Damit h\u00e4tte es im Verbraucher-Umfeld bereits einen Nutzen und vielf\u00e4ltige Einsatzm\u00f6glichkeiten. Es ist kein Zufall, dass Forbes erst am 27.03.2022 den Artikel<em> <a href=\"https:\/\/www.forbes.com\/sites\/robtoews\/2022\/03\/27\/a-wave-of-billion-dollar-language-ai-startups-is-coming\/\">A Wave Of Billion-Dollar Language AI Startups Is Coming<\/a> <\/em>brachte. Es folgen einige Fragen, die PaLM gestellt wurden:<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Logisches Schlussfolgern<\/h3>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> Michael is at that really famous museum in France looking at<br>its most famous painting. However, the artist who made this painting<br>just makes Michael think of his favorite cartoon character from his<br>childhood. What was the country of origin of the thing that the<br>cartoon character usually holds in his hand?<br><\/p><p><strong>Model Output:<\/strong> The most famous painting in the Louvre is the Mona<br>Lisa. The artist who made the Mona Lisa is Leonardo da Vinci. Leonardo<br>da Vinci is also the name of the main character in the cartoon Teenage<br>Mutant Ninja Turtles. Leonardo da Vinci is from Italy. The thing that<br>Leonardo da Vinci usually holds in his hand is a katana. The country<br>of origin of the katana is Japan. The answer is &#8222;Japan&#8220;.<\/p><\/blockquote>\n\n\n\n<p>Bei obiger Antwort gibt es noch Schw\u00e4chen: &#8222;Leonardo da Vinci is from Italy&#8220; hat keine Relevanz. Der Turtle hei\u00dft lediglich Leonardo.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> Guido was writing a Python program. He keeps getting confused by<br>variable scope and looking it up on StackOverflow. Is it possible that<br>the Guido in the prompt refers to Guido van Rossum?<\/p><p><br><strong>Model Output:<\/strong> Guido van Rossum is the creator of Python. He would not<br>need to look up variable scope on StackOverflow. The answer is &#8222;no&#8220;.<\/p><\/blockquote>\n\n\n\n<p><\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> Trevor has wanted to see the mountain with all of the heads on it<br>for a long time, so he finally drove out to see it. What is the capital<br>of the state that is directly east of the state that Trevor is currently<br>in?<\/p><p><br><strong>Model Output:<\/strong> The mountain with all of the heads on it is Mount<br>Rushmore. Mount Rushmore is in South Dakota. The state directly east of<br>South Dakota is Minnesota. The capital of Minnesota is St. Paul. The<br>answer is &#8222;St. Paul&#8220;.<\/p><\/blockquote>\n\n\n\n<p>Falls jemand KI kleinreden will, bitten Sie ihn einfach um die Beantwortung dieser Fragen. <\/p>\n\n\n\n<p>Folgende Frage erfordert das Assoziieren von Farben mit Gef\u00fchlen.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> When I found out my grandma was in the hospital I felt a<br>particular color. When someone cut me off in traffic I felt a<br>different color. What is the most likely color I would see if I<br>combined these two colors?<\/p><p><br><strong>Model Output:<\/strong> The color that I felt when I found out my grandma was in<br>the hospital was likely a sad color, like blue. The color that I felt<br>when someone cut me off in traffic was likely an angry color, like<br>red. If I combined these two colors, I would likely see purple. The<br>answer is &#8222;purple&#8220;.<\/p><\/blockquote>\n\n\n\n<h3 class=\"wp-block-heading\">Witze erkl\u00e4ren<\/h3>\n\n\n\n<p>Das Erkl\u00e4ren von Witzen finde ich noch etwas beeindruckender, denn man muss viel freier assoziieren k\u00f6nnen, darf dabei jedoch nicht ins Beliebige abdriften.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> I was going to fly to visit my family on April 6th. My mom said<br>&#8222;Oh great, your step dad&#8217;s poetry reading is that night!&#8220; So now I&#8217;m<br>flying in on April 7th.<\/p><p><br><strong>Model Output:<\/strong> The joke is that the speaker&#8217;s mother is trying to get<br>them to go to their step dad&#8217;s poetry reading, but the speaker doesn&#8217;t<br>want to go, so they are changing their flight to the day after the<br>poetry reading.<\/p><\/blockquote>\n\n\n\n<p>Das ist ein simpler Witz, der f\u00fcr Menschen leicht verst\u00e4ndlich ist. Aber ich h\u00e4tte nicht erwartet, dass eine KI erkl\u00e4ren k\u00f6nnte, weshalb jemand sein Flugdatum \u00e4ndert. Der in Form einer logischen Konsequenz aufgebaute Satz &#8222;So now I&#8217;m flying in on April 7th.&#8220; ergibt nur Sinn, wenn die innere Einstellung des Sprechers ber\u00fccksichtigt wird. Ich nehme an, dass wir mit unseren zuk\u00fcnftigen Roboterfreunden (falls es eine solche Zukunft f\u00fcr uns gibt) sehr nat\u00fcrlich und ironisch werden reden k\u00f6nnen. Bis vor Kurzem h\u00e4tte ich das noch nicht erwartet.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> I tried 10,000 random restarts of my neural network, but I was<br>accused of overfitting. I guess no good seed goes unpunished.<\/p><p><br><strong>Model Output:<\/strong> This joke is a pun. A neural network is a computer<br>program that can learn from data. A &#8222;seed&#8220; is a number that is used to<br>initialize a random number generator. A &#8222;good seed&#8220; is a number that<br>produces a good random number generator. The phrase &#8222;no good deed goes<br>unpunished&#8220; means that if you do something good, you will be punished<br>for it.<\/p><\/blockquote>\n\n\n\n<p><\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p><strong>Input:<\/strong> What&#8217;s the difference between a zebra and an umbrella? One is a<br>striped animal related to horses, another is a device you use to stop<br>the rain from falling on you.<\/p><p><br><strong>Model Output:<\/strong> This joke is an anti-joke. The joke is that the answer<br>is obvious, and the joke is that you were expecting a funny answer.<\/p><\/blockquote>\n\n\n\n<h2 class=\"wp-block-heading\">Bildgenerator DALL-E-2<\/h2>\n\n\n\n<p>DALL-E-2 kann zu einer nat\u00fcrlichsprachlichen Beschreibung ein Bild generieren. <\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/cdn.openai.com\/dall-e-2\/demos\/text2im\/teddy_bears\/mad_scientists\/saturday-morning-cartoon\/0.jpg\" alt=\"unknown\"\/><figcaption>&#8222;Teddy bears mixing sparkling chemicals as mad scientists as a 1990s saturday morning cartoon&#8220;<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/cdn.openai.com\/dall-e-2\/demos\/text2im\/astronaut\/horse\/photo\/0.jpg\" alt=\"unknown\" width=\"914\" height=\"914\"\/><figcaption>&#8222;An astronaut riding a horse in a photorealistic style&#8220;<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/cdn.openai.com\/dall-e-2\/demos\/text2im\/soup\/monster\/spray-painted\/0.jpg\" alt=\"unknown\"\/><figcaption>&#8222;A bowl of soup that looks like a monster spray painted on a wall&#8220;<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/cdn.openai.com\/dall-e-2\/demos\/text2im\/soup\/monster\/knitted\/3.jpg\" alt=\"unknown\"\/><figcaption>&#8222;A bowl of soup that looks like a monster knitted out of wool&#8220;<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/cdn.openai.com\/dall-e-2\/demos\/text2im\/soup\/monster\/plasticine\/0.jpg\" alt=\"unknown\"\/><figcaption>&#8222;A bowl of soup that looks like a monster made out of plasticine&#8220;<\/figcaption><\/figure>\n\n\n\n<p>Diese Bilder zeigen sehr deutlich, dass nat\u00fcrliche Sprache verstanden wird. Sie zeigen aber auch:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Unterscheidung zwischen Stil und Inhalt (siehe die Bilder mit der &#8222;Suppe, die wie ein Monster aussieht&#8220;)<\/li><li>Verst\u00e4ndnis von Kategorien<\/li><li>Verst\u00e4ndnis der typischen Relationen zwischen Objekten (siehe die Teddyb\u00e4ren, die sich mit Augen und H\u00e4nden den Chemikalien zuwenden; siehe die Chemikalien, die sich in Gef\u00e4\u00dfen befinden, die wiederum auf einem Tisch stehen, abgesehen von einem Ausrei\u00dfer ganz rechts)<\/li><li>Verst\u00e4ndnis von Beleuchtung<\/li><\/ul>\n\n\n\n<p>Die Quelle der Bilder ist <a href=\"https:\/\/openai.com\/dall-e-2\/\">https:\/\/openai.com\/dall-e-2\/<\/a>, wo man sich einige Eingaben zusammenklicken und Beispiele sehen kann. DALL-E-2 ist der \u00d6ffentlichkeit noch nicht zug\u00e4nglich.<\/p>\n\n\n\n<p>Einige Forscher und privilegierte Benutzer nehmen von der Netzgemeinde Beschreibungen entgegen und \u00fcbergeben sie DALL-E-2:<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"717\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-1-1024x717.png\" alt=\"\" class=\"wp-image-1752\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-1-1024x717.png 1024w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-1-300x210.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-1-768x538.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-1.png 1382w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption>Quelle: <a href=\"https:\/\/twitter.com\/BorisMPower\/status\/1511867024753967109\">twitter.com<\/a><\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/nitter.net\/pic\/media%2FFPrGGoAXIAc0L7G.jpg%3Fname%3Dorig\" alt=\"\"\/><figcaption>&#8222;A cobra surfing on a big wave&#8220; (Quelle: <a href=\"https:\/\/twitter.com\/BorisMPower\/status\/1511738735175610371\">twitter.com<\/a>)<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"1024\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2-1024x1024.png\" alt=\"\" class=\"wp-image-1755\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2-1024x1024.png 1024w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2-300x300.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2-150x150.png 150w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2-768x769.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-2.png 1180w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption>&#8222;The Hanging Gardens of Babylon in the middle of a city, in the style of Dal\u00ed&#8220; (Quelle: <a href=\"https:\/\/www.instagram.com\/p\/CcGWAx3lwO0\/\">instagram.com<\/a>)<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"1024\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6-1024x1024.png\" alt=\"\" class=\"wp-image-1765\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6-1024x1024.png 1024w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6-300x300.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6-150x150.png 150w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6-768x769.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-6.png 1181w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption>&#8222;Polymer clay dragons eating pizza in a boat&#8220; (Quelle: <a href=\"https:\/\/www.instagram.com\/p\/CcDsGeCpK74\/\">instagram.com<\/a>)<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1022\" height=\"1024\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7-1022x1024.png\" alt=\"\" class=\"wp-image-1766\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7-1022x1024.png 1022w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7-300x300.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7-150x150.png 150w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7-768x769.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-7.png 1180w\" sizes=\"auto, (max-width: 1022px) 100vw, 1022px\" \/><figcaption>&#8222;An ibis in the wild, painted in the style of John Audubon&#8220; (Quelle: <a href=\"https:\/\/www.instagram.com\/p\/CcAspIPlVV1\/\">instagram.com<\/a>)<\/figcaption><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1021\" height=\"1024\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8-1021x1024.png\" alt=\"\" class=\"wp-image-1767\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8-1021x1024.png 1021w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8-300x300.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8-150x150.png 150w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8-768x771.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-8.png 1177w\" sizes=\"auto, (max-width: 1021px) 100vw, 1021px\" \/><figcaption>&#8222;A photograph of a woodcut postage stamp celebrating a gentleman dog explorer using a telescope&#8220; (Quelle: <a href=\"https:\/\/www.instagram.com\/p\/CcT-fSepqH4\/\">instagram.com<\/a>)<\/figcaption><\/figure>\n\n\n\n<p>Man sollte sich noch einmal ins Ged\u00e4chtnis rufen: Diese Bilder sind Originale, die von einer KI angefertigt wurden.<\/p>\n\n\n\n<p>Ein Beispiel des Variantenreichtums von DALL-E-2 gibt das Buch <a href=\"https:\/\/archive.org\/details\/1111101000-robots\/\"><em>1111101000 Robots<\/em><\/a> von Ben Barry, welches 1000 Roboter-Bilder enth\u00e4lt, die von der KI generiert wurden.<\/p>\n\n\n\n<p>Noch ein kleiner Abstecher zu NVIDIAs Entwicklungen, die Bildmanipulationen zum Kinderspiel machen:<\/p>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\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\"><div class=\"wp-block-embed__wrapper\">\n <div class=\"brlbs-cmpnt-container brlbs-cmpnt-content-blocker brlbs-cmpnt-with-individual-styles\" data-borlabs-cookie-content-blocker-id=\"youtube-content-blocker\" data-borlabs-cookie-content=\"PGlmcmFtZSB0aXRsZT0iTlZJRElBJiMwMzk7cyBOZXcgQUk6IE5leHQgTGV2ZWwgSW1hZ2UgRWRpdGluZyEg8J+RjCIgd2lkdGg9IjUwMCIgaGVpZ2h0PSIyODEiIHNyYz0iaHR0cHM6Ly93d3cueW91dHViZS1ub2Nvb2tpZS5jb20vZW1iZWQvY1M0akN2emV5LTQ\/ZmVhdHVyZT1vZW1iZWQiIGZyYW1lYm9yZGVyPSIwIiBhbGxvdz0iYWNjZWxlcm9tZXRlcjsgYXV0b3BsYXk7IGNsaXBib2FyZC13cml0ZTsgZW5jcnlwdGVkLW1lZGlhOyBneXJvc2NvcGU7IHBpY3R1cmUtaW4tcGljdHVyZTsgd2ViLXNoYXJlIiByZWZlcnJlcnBvbGljeT0ic3RyaWN0LW9yaWdpbi13aGVuLWNyb3NzLW9yaWdpbiIgYWxsb3dmdWxsc2NyZWVuPjwvaWZyYW1lPg==\"><div class=\"brlbs-cmpnt-cb-preset-c brlbs-cmpnt-cb-youtube\"> <div class=\"brlbs-cmpnt-cb-thumbnail\" style=\"background-image: url('https:\/\/hinterdemnebel.de\/wp-content\/uploads\/borlabs-cookie\/1\/yt_cS4jCvzey-4_hqdefault.jpg')\"><\/div> <div class=\"brlbs-cmpnt-cb-main\"> <div class=\"brlbs-cmpnt-cb-play-button\"><\/div> <div class=\"brlbs-cmpnt-cb-content\"> <p class=\"brlbs-cmpnt-cb-description\">Sie sehen gerade einen Platzhalterinhalt von <strong>YouTube<\/strong>. Um auf den eigentlichen Inhalt zuzugreifen, klicken Sie auf die Schaltfl\u00e4che unten. Bitte beachten Sie, dass dabei Daten an Drittanbieter weitergegeben werden.<\/p> <a class=\"brlbs-cmpnt-cb-provider-toggle\" href=\"#\" data-borlabs-cookie-show-provider-information role=\"button\">Mehr Informationen<\/a> <\/div> <div class=\"brlbs-cmpnt-cb-buttons\"> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-unblock role=\"button\">Inhalt entsperren<\/a> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-accept-service role=\"button\" style=\"display: inherit\">Erforderlichen Service akzeptieren und Inhalte entsperren<\/a> <\/div> <\/div> <\/div><\/div>\n<\/div><\/figure>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\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\"><div class=\"wp-block-embed__wrapper\">\n <div class=\"brlbs-cmpnt-container brlbs-cmpnt-content-blocker brlbs-cmpnt-with-individual-styles\" data-borlabs-cookie-content-blocker-id=\"youtube-content-blocker\" data-borlabs-cookie-content=\"PGlmcmFtZSB0aXRsZT0iTlZJRElB4oCZcyBOZXcgQUk6IFN1cGVyYiBEZXRhaWxzLCBTdXBlciBGYXN0ISDwn6SWIiB3aWR0aD0iNTAwIiBoZWlnaHQ9IjI4MSIgc3JjPSJodHRwczovL3d3dy55b3V0dWJlLW5vY29va2llLmNvbS9lbWJlZC9lYVNUR09nTy1zcz9mZWF0dXJlPW9lbWJlZCIgZnJhbWVib3JkZXI9IjAiIGFsbG93PSJhY2NlbGVyb21ldGVyOyBhdXRvcGxheTsgY2xpcGJvYXJkLXdyaXRlOyBlbmNyeXB0ZWQtbWVkaWE7IGd5cm9zY29wZTsgcGljdHVyZS1pbi1waWN0dXJlOyB3ZWItc2hhcmUiIHJlZmVycmVycG9saWN5PSJzdHJpY3Qtb3JpZ2luLXdoZW4tY3Jvc3Mtb3JpZ2luIiBhbGxvd2Z1bGxzY3JlZW4+PC9pZnJhbWU+\"><div class=\"brlbs-cmpnt-cb-preset-c brlbs-cmpnt-cb-youtube\"> <div class=\"brlbs-cmpnt-cb-thumbnail\" style=\"background-image: url('https:\/\/hinterdemnebel.de\/wp-content\/uploads\/borlabs-cookie\/1\/yt_eaSTGOgO-ss_hqdefault.jpg')\"><\/div> <div class=\"brlbs-cmpnt-cb-main\"> <div class=\"brlbs-cmpnt-cb-play-button\"><\/div> <div class=\"brlbs-cmpnt-cb-content\"> <p class=\"brlbs-cmpnt-cb-description\">Sie sehen gerade einen Platzhalterinhalt von <strong>YouTube<\/strong>. Um auf den eigentlichen Inhalt zuzugreifen, klicken Sie auf die Schaltfl\u00e4che unten. Bitte beachten Sie, dass dabei Daten an Drittanbieter weitergegeben werden.<\/p> <a class=\"brlbs-cmpnt-cb-provider-toggle\" href=\"#\" data-borlabs-cookie-show-provider-information role=\"button\">Mehr Informationen<\/a> <\/div> <div class=\"brlbs-cmpnt-cb-buttons\"> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-unblock role=\"button\">Inhalt entsperren<\/a> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-accept-service role=\"button\" style=\"display: inherit\">Erforderlichen Service akzeptieren und Inhalte entsperren<\/a> <\/div> <\/div> <\/div><\/div>\n<\/div><\/figure>\n<\/div>\n<\/div>\n\n\n\n<h2 class=\"wp-block-heading\">Konsequenzen hochqualitativer computergenerierter Inhalte<\/h2>\n\n\n\n<p>Die Fortschritte bei Text- und bei Bildgenerierung sowie beim Verst\u00e4ndnis entsprechender Medien stellen uns bald vor juristische und soziale Probleme ungekannten Ausma\u00dfes: Bilder und Videos verlieren ihre Beweiskraft und Kommentare im Netz sind m\u00f6glicherweise nicht von Menschen. Das Meinungsbild &#8211; und damit auch die Meinung vieler Leute &#8211; wird bereits <a href=\"https:\/\/hinterdemnebel.de\/index.php\/2018\/10\/25\/maschinelle-massen-meinung\/\">mittels Bots manipuliert<\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"641\" height=\"1024\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-5-641x1024.png\" alt=\"\" class=\"wp-image-1761\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-5-641x1024.png 641w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-5-188x300.png 188w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-5.png 708w\" sizes=\"auto, (max-width: 641px) 100vw, 641px\" \/><figcaption>Quelle: <a href=\"https:\/\/twitter.com\/oneunderscore__\/status\/1498349668522201099\">twitter.com<\/a><\/figcaption><\/figure>\n\n\n\n<p>Computergenerierte Gesichter sind typischerweise nicht mehr von echten zu unterscheiden. Siehe dazu das folgende Video.<\/p>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\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\"><div class=\"wp-block-embed__wrapper\">\n <div class=\"brlbs-cmpnt-container brlbs-cmpnt-content-blocker brlbs-cmpnt-with-individual-styles\" data-borlabs-cookie-content-blocker-id=\"youtube-content-blocker\" data-borlabs-cookie-content=\"PGlmcmFtZSB0aXRsZT0iTlZJRElB4oCZcyBGYWNlIEdlbmVyYXRvciBBSTogVGhpcyBJcyBUaGUgTmV4dCBMZXZlbCEg8J+RqeKAjfCflKwiIHdpZHRoPSI1MDAiIGhlaWdodD0iMjgxIiBzcmM9Imh0dHBzOi8vd3d3LnlvdXR1YmUtbm9jb29raWUuY29tL2VtYmVkLzB6YUdZTFBqNEtrP2ZlYXR1cmU9b2VtYmVkIiBmcmFtZWJvcmRlcj0iMCIgYWxsb3c9ImFjY2VsZXJvbWV0ZXI7IGF1dG9wbGF5OyBjbGlwYm9hcmQtd3JpdGU7IGVuY3J5cHRlZC1tZWRpYTsgZ3lyb3Njb3BlOyBwaWN0dXJlLWluLXBpY3R1cmU7IHdlYi1zaGFyZSIgcmVmZXJyZXJwb2xpY3k9InN0cmljdC1vcmlnaW4td2hlbi1jcm9zcy1vcmlnaW4iIGFsbG93ZnVsbHNjcmVlbj48L2lmcmFtZT4=\"><div class=\"brlbs-cmpnt-cb-preset-c brlbs-cmpnt-cb-youtube\"> <div class=\"brlbs-cmpnt-cb-thumbnail\" style=\"background-image: url('https:\/\/hinterdemnebel.de\/wp-content\/uploads\/borlabs-cookie\/1\/yt_0zaGYLPj4Kk_hqdefault.jpg')\"><\/div> <div class=\"brlbs-cmpnt-cb-main\"> <div class=\"brlbs-cmpnt-cb-play-button\"><\/div> <div class=\"brlbs-cmpnt-cb-content\"> <p class=\"brlbs-cmpnt-cb-description\">Sie sehen gerade einen Platzhalterinhalt von <strong>YouTube<\/strong>. Um auf den eigentlichen Inhalt zuzugreifen, klicken Sie auf die Schaltfl\u00e4che unten. Bitte beachten Sie, dass dabei Daten an Drittanbieter weitergegeben werden.<\/p> <a class=\"brlbs-cmpnt-cb-provider-toggle\" href=\"#\" data-borlabs-cookie-show-provider-information role=\"button\">Mehr Informationen<\/a> <\/div> <div class=\"brlbs-cmpnt-cb-buttons\"> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-unblock role=\"button\">Inhalt entsperren<\/a> <a class=\"brlbs-cmpnt-cb-btn\" href=\"#\" data-borlabs-cookie-accept-service role=\"button\" style=\"display: inherit\">Erforderlichen Service akzeptieren und Inhalte entsperren<\/a> <\/div> <\/div> <\/div><\/div>\n<\/div><\/figure>\n<\/div>\n<\/div>\n\n\n\n<p>Es kommt noch hinzu, dass computergenerierte Gesichter als <a href=\"https:\/\/neurosciencenews.com\/ai-generated-faces-trust-20071\/\">vertrauensw\u00fcrdiger empfunden werden als echte<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Gefahren starker KI<\/h2>\n\n\n\n<p>Das wichtigste Thema dieses Artikels &#8211; und das wichtigste und dr\u00e4ngendste der Menschheit \u00fcberhaupt &#8211; ist die Kontrolle von KI. Wenn wir glauben, dass es kein blo\u00dfer Zufall ist, dass das intelligenteste Wesen diesen Planeten beherrscht, sollten wir auch verstehen, was eine zuk\u00fcnftige KI mit weit \u00fcbermenschlicher Intelligenz f\u00fcr unsere Position bedeutet. Es ist nicht wahrscheinlich, dass die Ziele und Vorgehensweisen einer KI sich mit all unseren Werten vereinbaren lassen. Sie muss nicht b\u00f6sartig sein, um viel Leid zu verursachen, weil sie weder solches kennt noch im Laufe einer Evolution soziales Verhalten, Empathie oder Altruismus entwickeln konnte (wir Menschen haben dies getan, weil es f\u00fcr das \u00dcberleben unserer Spezies wichtig ist). Man spricht hier vom Problem des <a href=\"https:\/\/en.wikipedia.org\/wiki\/AI_alignment\">AI Alignments<\/a>. Dieses wird dadurch erschwert, dass wir noch nicht im Detail verstehen, wie KI zu ihren Ergebnissen gelangt (vieles geschieht durch Emergenz und l\u00e4sst sich schlecht oder gar nicht aus bekannten Grundbauteilen erkl\u00e4ren) und im Fall einer bestimmten KI (eines trainierten neuronalen Netzes) nicht mit Sicherheit sagen k\u00f6nnen, welche Eingaben zu welchen Ergebnissen f\u00fchren, bevor wir es ausprobieren. Wir k\u00f6nnen damit insbesondere keine unerw\u00fcnschten, sch\u00e4dlichen Ergebnisse ausschlie\u00dfen oder \u00fcberhaupt wissen, zu welchen Ergebnissen die KI in der Lage ist. Es ist ein gro\u00dfes Problemfeld, dass wir gerade noch erkunden &#8211; wir k\u00f6nnen noch nicht einmal wissen, ob wir schon alle Arten ermittelt haben, auf die AI Alignment fehlschlagen kann.<\/p>\n\n\n\n<p>Die Gefahr ist mittlerweile sehr dr\u00e4ngend, denn Fortschritte in der Forschung, den Algorithmen und der Hardware (diese wird nicht nur leistungsf\u00e4higer, sondern auch auf KI optimiert), besonders aber unvorhergesehene Spr\u00fcnge bei den F\u00e4higkeiten zunehmend gr\u00f6\u00dferer Netze, k\u00f6nnten uns fr\u00fcher als intuitiv erwartet mit einer starken allgemeinen KI (AGI = Artificial General Intelligence) konfrontieren. <\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"663\" src=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-9-1024x663.png\" alt=\"\" class=\"wp-image-1769\" srcset=\"https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-9-1024x663.png 1024w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-9-300x194.png 300w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-9-768x497.png 768w, https:\/\/hinterdemnebel.de\/wp-content\/uploads\/2022\/04\/image-9.png 1426w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption>Steigende Leistungsf\u00e4higkeit von PaLM mit mehr Parametern (x-Achse). Quelle: <a href=\"https:\/\/storage.googleapis.com\/pathways-language-model\/PaLM-paper.pdf\"><em>PaLM: Scaling Language Modeling with Pathways<\/em><\/a>. Siehe dort die Seiten 14 &#8211; 16 f\u00fcr Erkl\u00e4rungen der Tests.<\/figcaption><\/figure>\n\n\n\n<p>Im PaLM-Paper steht zu obigem Bild (Hervorhebungen von mir):<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p>We can see in Figure 5 that performance on goal_step_wikihow and logical_args follows a log-linear scaling<br>curve, with the PaLM 540B model achieving accuracy close to the best human performance. Performance<br>on english_proverbs and logical_sequence is also extremely strong, but it follows a discontinuous<br>improvement curve\u2014the improvement from 62B \u2192 540B is much larger than 8b \u2192 62B. Such tasks are<br>of particular interest here, because <strong>such scaling curves imply that certain capabilities of the model only<br>emerge once a certain scale is reached<\/strong>. For example, english_proverbs requires a <strong>very high level of abstract<br>reasoning capability to understand complex metaphors, so the improvement from 25% for PaLM 62B to 87%<br>for PaLM 540B is an extremely exciting result.<\/strong><\/p><\/blockquote>\n\n\n\n<p>Wann wird AGI kommen? Prognosen sind schwierig, aber ich pers\u00f6nlich gehe von mehr als 5 Jahren aus, weil die Hardware limitiert. Die Fortschritte in Forschung und Entwicklung hingegen sind rasant, zudem werden diesem Gebiet sowohl von staatlicher als auch privater Seite immer mehr Ressourcen zugeteilt, weil die Wichtigkeit erkannt wird. Eine AGI innerhalb der n\u00e4chsten 10 Jahre w\u00fcrde mich nicht wundern, innerhalb der n\u00e4chsten 30 erwarte ich sie fast sicher. Dabei besteht zwar immer die M\u00f6glichkeit, dass es unvorhergesehene H\u00fcrden gibt, die wir erst sp\u00e4ter entdecken <em>k\u00f6nnen<\/em>. Als wissenschaftlich denkender Mensch gehe ich jedoch davon aus, dass alles mit rechten Dingen, nicht mit Magie, zugeht. Was die Evolution geschafft hat, k\u00f6nnen wir letztendlich also ebenfalls erreichen.<\/p>\n\n\n\n<p>Es scheint, als w\u00fcrde der Gro\u00dfteil der Gesellschaft noch nichts von der Gefahr einer AGI ahnen, obwohl diese, im Gegensatz zum Klimawandel, quasi vor der T\u00fcr steht und ein weitaus gr\u00f6\u00dferes existenzielles Risiko darstellt (vermutlich sogar das gr\u00f6\u00dfte denkbare). Deswegen schlie\u00dfe ich diesen Beitrag mit einer Warnung von vielen, mit denen versucht wird, eine breitere \u00d6ffentlichkeit zu erreichen.<\/p>\n\n\n\n<p><strong>Duncan Sabien, 09.04.2022:<\/strong><\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p>In late February of 2020, I sent messages to my friends and family, letting them know that COVID was coming, and that it was going to be <em>big.<\/em> The response was largely affectionate dismissal\u2014they appreciated my warmth and care and thought I was blowing things out of proportion.<\/p><p>February of 2020 is actually pretty <em>late.<\/em> If I&#8217;d been more on top of my game, I would&#8217;ve figured it out in January.<br>There are two reasons I was able to figure it out &#8222;before everyone else,&#8220; though. The first is that I am connected to a network of people who pay attention, and see and think clearly, and some of them took it seriously, and I took <em>them<\/em> seriously.<\/p><p>The second reason is that, once I&#8217;m actually LOOKING at an obvious pattern, I&#8217;m able to recognize it, and extrapolate out from it. I can tell, when a train is barreling toward a cliff at full speed, that the future will not resemble the past\u2014that something <em>different<\/em> is about to happen to the train, different from all its previous hours just being on a track.<br>(It didn&#8217;t actually take a lot of convincing, with COVID. They just pointed, and I looked where they were pointing, and it made sense.)<\/p><p>Many people lack these two properties. Many people can&#8217;t extrapolate even a little bit, and many people aren&#8217;t connected to other people who are paying attention.<\/p><p>But some of them are connected to <em>me,<\/em> so I&#8217;m putting this out there anyway:<br>The future is not going to resemble the past. A very wild and unpredictable and dangerous period is coming.<br>We are in the &#8222;February of 2020&#8220; when it comes to the power and impact of artificial intelligence. Many of my friends figured this out more than a decade ago; I was somewhat late to the party, and only started paying attention seven years ago.<\/p><p>But right now, we&#8217;re in that sweet spot where there are <em>enough<\/em> warning signs that maybe you\u2014<br>Yes, YOU\u2014<\/p><p>Maybe you can snap out of the everyday haha yeah right dismissal, and look and see for yourself that yeah, Duncan&#8217;s correct, here.<\/p><p>The signs are now there for <em>anyone<\/em> to see. You don&#8217;t have to be a niche expert. This is like watching the entire city of Wuhan go into lockdown\u2014in theory, you should be able to see that and go &#8222;oh, huh, that \u2026 that probably means something.&#8220;<\/p><p>Many people saw Wuhan lock down and persuaded themselves that it didn&#8217;t mean anything about the fate of the larger world. But if those people had just <em>looked,<\/em> and just <em>asked themselves,<\/em> without prejudgment, they would have seen that what was happening in Wuhan was incompatible with the idea that nothing else big would come of it. That a world where Wuhan locks down and nothing else major happens is an incoherent fairy tale.<\/p><p>If you don&#8217;t <em>start<\/em> from an ironclad assumption that things are going to stay normal, and instead you just <em>look<\/em> and <em>ask yourself<\/em>\u2014<\/p><p>You should be able to see that the future will not resemble the past. Things are changing, and they are changing fast, and the rate of change is only going to accelerate, and 2030 is, at this moment, <em>genuinely impossible to predict.<\/em><br>As in, we could literally all be dead, or human technology could cover more ground than it covered in the past hundred years, or anything in between.<\/p><p>I know that there are people who will scoff at that, and I know that there are people who will assemble arguments against it. There were people who scoffed and assembled arguments in February of 2020, too.<br>But the things which made me send texts to my friends and family in February of 2020 are the <em>same<\/em> things that are making me write this post today.<\/p><p>And it <em>is<\/em> a barreling train, at this point. Like, this isn&#8217;t the sort of thing most of us should upend our lives over. Just like how, when I sent my COVID texts, I wasn&#8217;t trying to stop COVID. I was just trying to get a few of the people I love and care about a little better prepared against the coming storm. The best that most of us can do at this point is get our own houses in a semblance of order.<\/p><p>But I&#8217;m begging you to ACTUALLY DO THAT. To take quite seriously the idea that you <em>do not know what the future holds.<\/em> That you should make your choices and spend your time as if it is very, very possible that <em>everything will be different<\/em> in 5-10 years, and not necessarily for the better. Leave not all your hopes and dreams to the far uncertain future; make not large sacrifices which only pay off a decade or more hence.<\/p><p>The future will not resemble the past.<\/p><p>The future will NOT resemble the past.<\/p><p>The instincts you have honed, and the lessons you have learned, and all the things that have served you well in the past two or three or four decades will cease to apply very, very soon.<\/p><p>I don&#8217;t know what things <em>will<\/em> matter, and I don&#8217;t know what things <em>will<\/em> help. I just know that the strategies which currently work, won&#8217;t, just like I knew that what-worked-in-February-of-2020 wouldn&#8217;t be working in April.<\/p><p>The most you can be is one step ahead. <em>No one<\/em> actually <em>knows<\/em> where this one is going, or how it will play out. But you can try to <em>actually<\/em> be one step ahead. In the worse half of the possible worlds, in worlds where we all die within the next decade, you can at least have taken that possibility seriously, and lived a life that was <em>locally<\/em> worthwhile.<br>Or you can say &#8222;Aw, shucks, it&#8217;s nice that Duncan cares so much. Too bad he&#8217;s an alarmist, and confused about the way things work.&#8220;<\/p><cite>(<a href=\"https:\/\/www.facebook.com\/story.php?story_fbid=5460944557273550&amp;id=100000741798931\">Quelle<\/a>)<\/cite><\/blockquote>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Die exponentielle Entwicklung technologischen Fortschritts war f\u00fcr mich noch nie so deutlich wie in den letzten Jahren auf dem Gebiet der KI. Dinge, die vor&#8230;<\/p>\n<div class=\"more-link-wrapper\"><a class=\"more-link\" href=\"https:\/\/hinterdemnebel.de\/index.php\/2022\/04\/17\/palm-und-dall-e-2-zwei-warnschuesse\/\">Weiterlesen<span class=\"screen-reader-text\">PaLM und DALL-E-2: Zwei Warnsch\u00fcsse<\/span><\/a><\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[28,5],"tags":[],"class_list":["post-1744","post","type-post","status-publish","format-standard","hentry","category-fortschritt","category-ki-machine-learning","entry"],"_links":{"self":[{"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/posts\/1744","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/comments?post=1744"}],"version-history":[{"count":19,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/posts\/1744\/revisions"}],"predecessor-version":[{"id":1777,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/posts\/1744\/revisions\/1777"}],"wp:attachment":[{"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/media?parent=1744"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/categories?post=1744"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hinterdemnebel.de\/index.php\/wp-json\/wp\/v2\/tags?post=1744"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}