{"id":4405,"date":"2023-05-03T08:55:00","date_gmt":"2023-05-03T14:55:00","guid":{"rendered":"https:\/\/singularityumexico.com\/?p=4405"},"modified":"2023-05-02T16:58:12","modified_gmt":"2023-05-02T22:58:12","slug":"chatgpt-is-greatyoure-just-using-it-wrong","status":"publish","type":"post","link":"https:\/\/singularityumexico.com\/en\/chatgpt-is-greatyoure-just-using-it-wrong\/","title":{"rendered":"ChatGPT Is Great\u2014You\u2019re Just Using It Wrong"},"content":{"rendered":"

It doesn\u2019t take much to get\u00a0ChatGPT<\/a>\u00a0to make a factual mistake. My son is doing a report on US presidents, so I figured I\u2019d help him out by looking up a few biographies. I tried asking for a list of books about Abraham Lincoln, and it did a pretty good job:<\/p>\n\n\n\n

\"\"
A reasonable list of books about Lincoln. Image Credit:\u00a0Screen capture by Jonathan May,\u00a0CC BY-ND<\/a><\/em><\/figcaption><\/figure>\n\n\n\n

Number 4 isn\u2019t right. Garry Wills famously wrote \u201cLincoln at Gettysburg,\u201d and Lincoln himself wrote the Emancipation Proclamation, of course, but it\u2019s not a bad start. Then I tried something harder, asking instead about the much more obscure William Henry Harrison, and it gamely provided a list, nearly all of which was wrong.<\/p>\n\n\n\n

\"\"
Books about Harrison, fewer than half of which are correct. Image Credit:\u00a0Screen capture by Jonathan May,\u00a0CC BY-ND<\/a><\/em><\/figcaption><\/figure>\n\n\n\n

Numbers 4 and 5 are correct; the rest don\u2019t exist or are not authored by those people. I repeated the exact same exercise and got slightly different results:<\/p>\n\n\n\n

\"\"
More books about Harrison, still mostly nonexistent. Image Credit:\u00a0Screen capture by Jonathan May,\u00a0CC BY-ND<\/a><\/em><\/figcaption><\/figure>\n\n\n\n

This time numbers 2 and 3 are correct and the other three are not actual books or not written by those authors. Number 4, \u201cWilliam Henry Harrison: His Life and Times\u201d is a real book<\/a>, but it\u2019s by James A. Green, not by Robert Remini, a well-known historian<\/a> of the Jacksonian age.<\/p>\n\n\n\n

I called out the error, and ChatGPT eagerly corrected itself and then confidently told me the book was in fact written by Gail Collins (who wrote a different Harrison biography), and then went on to say more about the book and about her. I finally revealed the truth, and the machine was happy to run with my correction. Then I lied absurdly, saying during their first hundred days presidents have to write a biography of some former president, and ChatGPT called me out on it. I then lied subtly, incorrectly attributing authorship of the Harrison biography to historian and writer Paul C. Nagel, and it bought my lie.<\/p>\n\n\n\n

When I asked ChatGPT if it was sure I was not lying, it claimed that it\u2019s just an \u201cAI language model\u201d and doesn\u2019t have the ability to verify accuracy. However, it modified that claim by saying, \u201cI can only provide information based on the training data I have been provided, and it appears that the book \u2018William Henry Harrison: His Life and Times\u2019 was written by Paul C. Nagel and published in 1977.\u201d<\/p>\n\n\n\n

This is not true.<\/p>\n\n\n\n

Words, Not Facts<\/h2>\n\n\n\n

It may seem from this interaction that ChatGPT was given a library of facts, including incorrect claims about authors and books. After all, ChatGPT\u2019s maker, OpenAI, claims it trained the chatbot on \u201cvast amounts of data from the internet written by humans<\/a>.\u201d<\/p>\n\n\n\n

However, it was almost certainly not given the names of a bunch of made-up books about one of the most mediocre presidents<\/a>. In a way, though, this false information is indeed based on its training data.<\/p>\n\n\n\n

As a computer scientist<\/a>, I often field complaints that reveal a common misconception about large language models like ChatGPT and its older brethren GPT3 and GPT2: that they are some kind of \u201csuper Googles,\u201d or digital versions of a reference librarian, looking up answers to questions from some infinitely large library of facts, or smooshing together pastiches of stories and characters. They don\u2019t do any of that\u2014at least, they were not explicitly designed to.<\/p>\n\n\n\n

Sounds Good<\/h2>\n\n\n\n

A language model like ChatGPT, which is more formally known as a \u201cgenerative pre-trained transformer\u201d (that\u2019s what the G, P, and T stand for), takes in the current conversation, forms a probability for all of the words in its vocabulary given that conversation, and then chooses one of them as the likely next word. Then it does that again, and again, and again, until it stops.<\/p>\n\n\n\n

So it doesn\u2019t have facts, per se. It just knows what word should come next. Put another way, ChatGPT doesn\u2019t try to write sentences that are true. But it does try to write sentences that are plausible.<\/p>\n\n\n\n

When talking privately to colleagues about ChatGPT, they often point out how many factually untrue statements it produces and dismiss it. To me, the idea that ChatGPT is a flawed data retrieval system is beside the point. People have been using Google for the past two and a half decades, after all. There\u2019s a pretty good fact-finding service out there already.<\/p>\n\n\n\n

In fact, the only way I was able to verify whether all those presidential book titles were accurate was by Googling and then verifying the results<\/a>. My life would not be that much better if I got those facts in conversation, instead of the way I have been getting them for almost half of my life, by retrieving documents and then doing a critical analysis to see if I can trust the contents.<\/p>\n\n\n\n

Improv Partner<\/h2>\n\n\n\n

On the other hand, if I can talk to a bot that will give me plausible responses to things I say, it would be useful in situations where factual accuracy isn\u2019t all that important<\/a>. A few years ago a student and I tried to create an \u201cimprov bot,\u201d one that would respond to whatever you said with a \u201cyes, and\u201d to keep the conversation going. We showed, in a paper<\/a>, that our bot<\/a> was better at \u201cyes, and-ing\u201d than other bots at the time, but in AI, two years is ancient history.<\/p>\n\n\n\n

I tried out a dialogue with ChatGPT\u2014a science fiction space explorer scenario\u2014that is not unlike what you\u2019d find in a typical improv class. ChatGPT is way better at \u201cyes, and-ing\u201d than what we did, but it didn\u2019t really heighten the drama at all. I felt as if I was doing all the heavy lifting.<\/p>\n\n\n\n

After a few tweaks I got it to be a little more involved, and at the end of the day, I felt that it was a pretty good exercise for me, who hasn\u2019t done much improv since I graduated from college over 20 years ago.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Sure, I wouldn\u2019t want ChatGPT to appear on \u201cWhose Line Is It Anyway?<\/a>\u201d and this is not a great \u201cStar Trek\u201d plot (though it\u2019s still less problematic than \u201cCode of Honor<\/a>\u201d), but how many times have you sat down to write something from scratch and found yourself terrified by the empty page in front of you? Starting with a bad first draft can break through writer\u2019s block and get the creative juices flowing, and ChatGPT and large language models like it seem like the right tools to aid in these exercises.<\/p>\n\n\n\n

And for a machine that is designed to produce strings of words that sound as good as possible in response to the words you give it\u2014and not to provide you with information\u2014that seems like the right use for the tool.<\/p>\n\n\n\n

\n\n\n\n

This article is republished from The Conversation<\/a> under a Creative Commons license. Read the original article<\/a>.<\/em><\/p>\n\n\n\n

Image Credit:\u00a0Justin Ha<\/a>\u00a0\/\u00a0Unsplash<\/a><\/em><\/p>\n\n\n\n

Author:<\/strong>
Jonathan May<\/a>https:\/\/www.cs.usc.edu\/directory\/faculty\/profile\/?lname=May&fname=Jonathan<\/a><\/p>\n\n\n\n

Jonathan May is a research associate professor at the University of Southern California’s Information Sciences Institute, where he directs the Center for Useful Techniques Enhancing Language Applications Based on Natural and Meaningful Evidence. His research areas include natural language processing, specifically machine translation, dialogue, semantic parsing, and formal language theory. Jonathan May received his Ph.D. in Computer Science from USC in 2010. Prior to re-joining USC and the Information Sciences institute in 2014, he was a research scientist at SDL Language Weaver.<\/p>\n\n\n\n

Original Article<\/a><\/strong><\/p>","protected":false},"excerpt":{"rendered":"

It doesn\u2019t take much to get\u00a0ChatGPT\u00a0to make a factual mistake. My son is doing a report on US presidents, so I figured I\u2019d help him out by looking up a few biographies. I tried asking for a list of books about Abraham Lincoln, and it did a pretty good job: Number 4 isn\u2019t right. Garry […]\n","protected":false},"author":1,"featured_media":4415,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"episode_type":"","audio_file":"","podmotor_file_id":"","podmotor_episode_id":"","cover_image":"","cover_image_id":"","duration":"","filesize":"","filesize_raw":"","date_recorded":"","explicit":"","block":"","footnotes":""},"categories":[13],"tags":[18,233,235,236,234],"series":[],"class_list":["post-4405","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-articulos-ingles","tag-inteligencia-artificial","tag-chatgpt","tag-gpt-2","tag-gpt-3","tag-lenguaje-models"],"episode_featured_image":"https:\/\/singularityumexico.com\/wp-content\/uploads\/2023\/05\/chatbot-openai-chatgpt.jpeg","episode_player_image":"https:\/\/singularityumexico.com\/wp-content\/uploads\/2023\/05\/11711533-1673157178559-89a95be153719-4-scaled.jpg","download_link":"","player_link":"","audio_player":false,"episode_data":{"playerMode":"dark","subscribeUrls":{"apple_podcasts":{"key":"apple_podcasts","url":"","label":"Apple Podcasts","class":"apple_podcasts","icon":"apple-podcasts.png"},"stitcher":{"key":"stitcher","url":"","label":"Stitcher","class":"stitcher","icon":"stitcher.png"},"google_podcasts":{"key":"google_podcasts","url":"","label":"Google Podcasts","class":"google_podcasts","icon":"google-podcasts.png"},"spotify":{"key":"spotify","url":"","label":"Spotify","class":"spotify","icon":"spotify.png"}},"rssFeedUrl":"https:\/\/singularityumexico.com\/en\/feed\/podcast\/the-feedback-loop-by-singularity","embedCode":"

ChatGPT Is Great\u2014You\u2019re Just Using It Wrong<\/a><\/blockquote>