SlideShare a Scribd company logo

ChatGPT: The Hype vs Reality

•Download as PPTX, PDF•
•
Kristine Schachinger SEO and Online Marketing
Kristine Schachinger SEO and Online Marketing

The document discusses ChatGPT and large language models. It provides an overview of what ChatGPT is, how it works using transformer models, and some of its applications and limitations. Specifically: - ChatGPT is an AI assistant created by OpenAI that is trained on massive datasets to generate human-like text. It uses transformer models and self-attention to understand context. - ChatGPT has applications in content generation, translation, summarization, and more. However, it also has limitations like an inability to fact check, lack of justification for responses, and producing made-up or inaccurate information ("hallucinations"). - Prompts are important for guiding ChatGPT's responses.

Related slideshows

Generative AI
Generative AIGenerative AI
Generative AI
 
How do OpenAI GPT Models Work - Misconceptions and Tips for Developers
How do OpenAI GPT Models Work - Misconceptions and Tips for DevelopersHow do OpenAI GPT Models Work - Misconceptions and Tips for Developers
How do OpenAI GPT Models Work - Misconceptions and Tips for Developers
 
Revolutionary-ChatGPT
Revolutionary-ChatGPTRevolutionary-ChatGPT
Revolutionary-ChatGPT
 
1 of 61
@schachin Kristine Schachinger Kristine@SitesWithoutWalls.com "Unlocking the Power of ChatGPT” -- without blowing yourself up -- “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”
@schachin Kristine Schachinger What is a “Chat Generative Pre-trained Transformer”?
@schachin Kristine Schachinger ChatGPT “Chat Generative Pre-trained Transformer”
@schachin Kristine Schachinger Generative Pre-Trained Transformer Model.
@schachin Kristine Schachinger GPT, ChatGPT, and … Large Language Models. Cha Dalle’s vision of ChatGPT ChatGPT via MidJourney
@schachin Kristine Schachinger Why are LLMs So Special? Unlike embedded based language models LLMs can disambiguate words from the sentence and apply meaning forward and backward to those words in order to predict a masked word using those applied contexts. This is SUPER EFFICIENT!
@schachin Kristine Schachinger Why are LLMs So Special? Large Language modeling can determine the meaning of words in context so it can better predict the next word in the sentence. These sentences mean two different things forward and backward.
@schachin Kristine Schachinger How does this work? Transformers What are transformers? A transformer in language processing is a type of computer program that is designed to understand and generate text. It does this by using a special type of algorithm called self-attention. Self-attention allows the program to look at all the words in a sentence or a piece of text at once, and understand how they relate to each other, rather than just one word at a time like traditional methods. This way it can better understand the meaning of the text, and can generate text that is more similar to how a human would write.
@schachin Kristine Schachinger LLMs can go forward and backwards to predict an unknown (masked) term and/or sentence. Also uses root words, so play for player/playing/played are the same This allows them to derive context for what is being written. Previous models were based on word vectors (entities and knowledge graphs) LLM Transformers are Bidirectional https://blog.google/products/search/search-language-understanding-bert/
@schachin Kristine Schachinger Confusing?
@schachin Kristine Schachinger From Google on their LLM Bard. (One correction it is run on PaLM now.)
@schachin Kristine Schachinger Simply language modeling is “Language modeling – although it sounds formidable – is essentially just predicting words in a blank.”
@schachin Kristine Schachinger Like the American Game - MadLibs
@schachin Kristine Schachinger Mad Libs! LLMs fill in the next most likely word in the blank. It is predictive text with semantic embeddings.
@schachin Kristine Schachinger Large Language Models? Prompt: 3d photo image of a large language model as a 90s popstar Some of the more relevant well-known LLMs are: • OpenAI  ChatGPT 3.5 / ChatGPT 4  WebGPT  attempts to resolve the issue of attribution and accuracy  Bing  Bing Chat which uses ChatGPT4  Google  BERT  Used in Search  LaMDA  conversational like ChatGPT  “Unlike GPT-3, which is not configured to perform any specific task, LaMDA is “trained on dialogue”:  PaLM  Search relevancy  it is also trained using a “combination of English and multilingual datasets that include high-quality web documents, books, Wikipedia, conversations and GitHub code” to drive greater accuracy in responses.
@schachin Kristine Schachinger What is ChatGPT?
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: What is ChatGPT? Let’s Ask It! ChatGPT is/was • a variant of the GPT (Generative Pre-training Transformer) language model • introduced by OpenAI in 2018. • The original GPT trained on a massive dataset of over 40GB of text data for GPT-2 and 45TB for GPT-3.x • ChatGPT includes conversational text from books, articles, and websites. • capable of generating human-like text • useful for certain tasks • language translation, text summarization, and question answering. • fine-tuned on several specific • writing poetry, composing music, and even coding. • one of the largest language models currently available • over trillions of parameters • used in several real-world applications • chatbots, automated customer service, writing news articles, programming etc. • continuously being updated and improved by OpenAI
@schachin Kristine Schachinger TeThje What is ChatGPT? In layman’s terms, a large language model (LLM), which generates convincing sentences by mimicking the statistical patterns of language in a huge database of text collated from the Internet. It was trained to sound confident to humans even when it is wrong. https://www.nature.com/articles/d41586-023-00107-z
@schachin Kristine Schachinger TeThje What is ChatGPT? https://sensoriumxr.com/articles/what-is-chatgpt4
@schachin Kristine Schachinger ChatGPT is Predictive Text. It is predicting the next most likely word in a sentence based on the context of the other words in the sentence derived from the information it was trained on.
@schachin Kristine Schachinger TeThje What is ChatGPT NOT? • A Search Engine • The data was trained only up to 2021 • It is not connected to the Internet, so it cannot retrieve information • The Bing Chat plugin allowed you to “ground” a prompt in a search but was recently disabled because users found a way to copy paywalled content. • Is based on predictive text NOT information retrieval • There is no QUALITY check for the accuracy of the content or code • Creative • if it was not trained on the information, it cannot generate something as a human would from inference or imagination. https://www.nature.com/articles/d41586-023-00107-z
@schachin Kristine Schachinger Using ChatGPT for Marketers.
@schachin Kristine Schachinger Important! ChatGPT Output Cannot Be Copyrighted. Not in the US or the EU, not yet anyhow.
@schachin Kristine Schachinger Why? A human must create a creative work for copyright to apply.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Some ChatGPT Ideas for Marketers • SEO: Titles and Metas (okayed by Google) • PPC Landing Pages • Content Ideas • Content NOT for Search • Topical Funnels • Analyze Competitor Site Content • Product Descriptions • Customer Service / Chatbots • Emails • Outreach • Link Building • Lead Generation • Newsletters • Reviews digital marketer using AI to do SEO Disney style
@schachin Kristine Schachinger ALWAYS USE HUMAN REVIEW. ChatGPT is not a search engine, has no fact checking function, and the core product was not trained on any data past 2021 so make sure someone reviews the output produced for errors, accuracy, factual information, and tone.
@schachin Kristine Schachinger AI Content and Google. Don’t Blow Yourself Up. a human male person with the job of SEO Consultant yelling at Google in 3d cartoon form
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: AI Content and Google. Google says AI Content is okay IF it provides value and it not “spammy”, But since it is writing what it trained on how does provide value?
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Detection and Google. How does Google define “Spammy” content?
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Detection and Google. https://developers.google.com/search/blog/2022/08/helpful-content-update Google and the Helpful Content Update.
@schachin Kristine Schachinger Play it safe, do not use ChatGPT or AI to write article content. Use humans to write/edit your content. They can use ChatGPT as a tool to generate ideas, however.
@schachin Kristine Schachinger How to Use ChatGPT.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT. The Application of ChatGPT Some of the potential applications of ChatGPT include: • Text generation: ChatGPT can be used to generate human-like text responses to prompts, which makes it useful for creating chatbots for customer service, generating responses to questions in online forums, or even creating personalized content for social media posts. • Language translation: ChatGPT can also be used for language translation tasks. By providing the model with a text prompt in one language and specifying the target language, the model can generate accurate and fluent translations of the text. • Text summarization: ChatGPT can be used to generate summaries of long documents or articles. This can be useful for quickly getting an overview of a text without having to read the entire document. • Sentiment analysis: ChatGPT can be used to analyze the sentiment of a given text. This can be useful for understanding the overall tone and emotion of a piece of writing, or for detecting the sentiment of customer feedback in order to improve customer satisfaction.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT. Fine Tuning in ChatGPT Training. ChatGPT was trained with the following focuses: • Code • Content Writing • Poetry • Music This means it tends to create these items well. Fine tuning allows ChatGPT to focus in on certain types of functions and may not do as well with others.
@schachin Kristine Schachinger Using ChatGPT Prompts.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: InstructGPT: ChatGPT Prompts. Prompts. The MOST Important Part of Using ChatGPT. Prompts provide a starting point for ChatGPT to generate text based on the input given. They can be used to guide the model's output and control the context of the generated text. Prompts? Simply means, how you tell ChatGPT what to write. This relies on another language model called InstructGPT, which allows you to give ChatGPT detailed and specific instructions. https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
@schachin Kristine Schachinger Using ChatGPT Prompts: Examples.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > Articles.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > Articles + Style.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > FAQs.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Code.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Summarization > URLs.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Summarize URLs.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Write Emails.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Lyrics.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Poetry.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Ideas.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Prompts. Prompts. How to write ChatGPT Prompts • Give Context • Be Clear • Be Concise • Be Specific • Be Descriptive • Define Format • Define Length • Define Style
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Prompts. Prompts. Some great ideas for how you can use ChatGPT • 100 Best ChatGPT Prompts • https://mpost.io/100-best-chatgpt-prompts-to-unleash-ais-potential/ • Web Development ChatGPT • Music ChatGPT • Business ChatGPT • Educational ChatGPT • Comedy ChatGPT • History ChatGPT • Health&Medicine ChatGPT • Art ChatGPT • Food & Cooking ChatGPT • Marketing ChatGPT • Game ChatGPT Prompts https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
@schachin Kristine Schachinger Issues with ChatGPT
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Issues? + Programmed to Avoid Certain Kinds of Content • ChatGPT is specifically programmed to not generate text on the topics of graphic violence, explicit sex, and content that is harmful such as instructions on how to build an explosive device. • And swearing + Unaware of Current Events • not aware of any content that is created after 2021.  Has Built-in Biases • This can be just in the writing style, not necessarily cultural or racial though they can exist as well + ChatGPT Requires Highly Detailed Instructions • You must be very specific with ChatGPT Prompts to really get good output + Mediocre Content • ChatGPT is designed to sound confident, but the content it writes is not high quality. + Suffers from “Hallucinations” • Makes stuff up https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
@schachin Kristine Schachinger “Hallucinations”: Machine Learning for Making Stuff Up
@schachin Kristine Schachinger https://www.wired.com/story/chatgpt-fluent-bs/ All of this makes playing around with ChatGPT incredibly fun, charmingly addictive, and—as someone who writes for a living— really quite worrying. But you soon start to sense a lack of depth beneath ChatGPT’s competent prose. It makes factual errors, conflating events and mixing people up. It relies heavily on tropes and cliché, and it echoes society’s worst stereotypes. Its words are superficially impressive but largely lacking in substance— ChatGPT mostly produces what The Verge has described as “fluent bullshit.” But that kind of makes sense. ChatGPT was trained on real-world text, and the real world essentially runs on fluent bullshit. “Bovine Feces”
@schachin Kristine Schachinger Google https://ai.google/static/documents/exploring-6-myths.pdf
@schachin Kristine Schachinger Google https://www.semanticscholar.org/paper/Rethinking-Search%3A-Making-Domain-Experts-out-of-Metzler-Tay/e3d7065d6196437259483ed8b9b70ba4b4bb76ad “…they do not have a true understanding of the world, they are prone to hallucinating, and crucially they are incapable of justifying their utterances by referring to supporting documents in the corpus they were trained over.” Image from DALL-E
@schachin Kristine Schachinger Google https://www.semanticscholar.org/paper/Rethinking-Search%3A-Making-Domain-Experts-out-of-Metzler-Tay/e3d7065d6196437259483ed8b9b70ba4b4bb76ad “…they do not have a true understanding of the world, they are prone to hallucinating, and crucially they are incapable of justifying their utterances by referring to supporting documents in the corpus they were trained over.” Image from DALL-E “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”
@schachin Kristine Schachinger https://gizmodo.com/cnet-ai-chatgpt-news-robot-1849996151
@schachin Kristine Schachinger ALWAYS USE HUMAN REVIEW. ChatGPT is not a search engine, has no fact checking function, and was not trained on any data past 2021 so make sure someone reviews the output produced for errors, accuracy, factual information, and tone.
@schachin Kristine Schachinger ChatGPT Resources.
@schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Docs. ChatGPT Guide • All the basics you need to know about how to use ChatGPT • https://www.digitaltrends.com/computing/how-to-use-openai-chatgpt-text-generation-chatbot/#google_vignette OpenAI ChatGPT API Guide. • Everything you need to know about using the API version • https://beta.openai.com/docs/guides/completion/inserting-text Some cool things people have done with ChatGPT • https://www.bleepingcomputer.com/news/technology/openais-new-chatgpt-bot-10-coolest-things-you-can-do-with-it/
@schachin Kristine Schachinger Kristine@SitesWithoutWalls.com "Unlocking the Power of ChatGPT” -- without blowing yourself up -- “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”

More Related Content

ChatGPT: The Hype vs Reality

  • 1. @schachin Kristine Schachinger Kristine@SitesWithoutWalls.com "Unlocking the Power of ChatGPT” -- without blowing yourself up -- “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”
  • 2. @schachin Kristine Schachinger What is a “Chat Generative Pre-trained Transformer”?
  • 5. @schachin Kristine Schachinger GPT, ChatGPT, and … Large Language Models. Cha Dalle’s vision of ChatGPT ChatGPT via MidJourney
  • 6. @schachin Kristine Schachinger Why are LLMs So Special? Unlike embedded based language models LLMs can disambiguate words from the sentence and apply meaning forward and backward to those words in order to predict a masked word using those applied contexts. This is SUPER EFFICIENT!
  • 7. @schachin Kristine Schachinger Why are LLMs So Special? Large Language modeling can determine the meaning of words in context so it can better predict the next word in the sentence. These sentences mean two different things forward and backward.
  • 8. @schachin Kristine Schachinger How does this work? Transformers What are transformers? A transformer in language processing is a type of computer program that is designed to understand and generate text. It does this by using a special type of algorithm called self-attention. Self-attention allows the program to look at all the words in a sentence or a piece of text at once, and understand how they relate to each other, rather than just one word at a time like traditional methods. This way it can better understand the meaning of the text, and can generate text that is more similar to how a human would write.
  • 9. @schachin Kristine Schachinger LLMs can go forward and backwards to predict an unknown (masked) term and/or sentence. Also uses root words, so play for player/playing/played are the same This allows them to derive context for what is being written. Previous models were based on word vectors (entities and knowledge graphs) LLM Transformers are Bidirectional https://blog.google/products/search/search-language-understanding-bert/
  • 11. @schachin Kristine Schachinger From Google on their LLM Bard. (One correction it is run on PaLM now.)
  • 12. @schachin Kristine Schachinger Simply language modeling is “Language modeling – although it sounds formidable – is essentially just predicting words in a blank.”
  • 14. @schachin Kristine Schachinger Mad Libs! LLMs fill in the next most likely word in the blank. It is predictive text with semantic embeddings.
  • 15. @schachin Kristine Schachinger Large Language Models? Prompt: 3d photo image of a large language model as a 90s popstar Some of the more relevant well-known LLMs are: • OpenAI  ChatGPT 3.5 / ChatGPT 4  WebGPT  attempts to resolve the issue of attribution and accuracy  Bing  Bing Chat which uses ChatGPT4  Google  BERT  Used in Search  LaMDA  conversational like ChatGPT  “Unlike GPT-3, which is not configured to perform any specific task, LaMDA is “trained on dialogue”:  PaLM  Search relevancy  it is also trained using a “combination of English and multilingual datasets that include high-quality web documents, books, Wikipedia, conversations and GitHub code” to drive greater accuracy in responses.
  • 17. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: What is ChatGPT? Let’s Ask It! ChatGPT is/was • a variant of the GPT (Generative Pre-training Transformer) language model • introduced by OpenAI in 2018. • The original GPT trained on a massive dataset of over 40GB of text data for GPT-2 and 45TB for GPT-3.x • ChatGPT includes conversational text from books, articles, and websites. • capable of generating human-like text • useful for certain tasks • language translation, text summarization, and question answering. • fine-tuned on several specific • writing poetry, composing music, and even coding. • one of the largest language models currently available • over trillions of parameters • used in several real-world applications • chatbots, automated customer service, writing news articles, programming etc. • continuously being updated and improved by OpenAI
  • 18. @schachin Kristine Schachinger TeThje What is ChatGPT? In layman’s terms, a large language model (LLM), which generates convincing sentences by mimicking the statistical patterns of language in a huge database of text collated from the Internet. It was trained to sound confident to humans even when it is wrong. https://www.nature.com/articles/d41586-023-00107-z
  • 19. @schachin Kristine Schachinger TeThje What is ChatGPT? https://sensoriumxr.com/articles/what-is-chatgpt4
  • 20. @schachin Kristine Schachinger ChatGPT is Predictive Text. It is predicting the next most likely word in a sentence based on the context of the other words in the sentence derived from the information it was trained on.
  • 21. @schachin Kristine Schachinger TeThje What is ChatGPT NOT? • A Search Engine • The data was trained only up to 2021 • It is not connected to the Internet, so it cannot retrieve information • The Bing Chat plugin allowed you to “ground” a prompt in a search but was recently disabled because users found a way to copy paywalled content. • Is based on predictive text NOT information retrieval • There is no QUALITY check for the accuracy of the content or code • Creative • if it was not trained on the information, it cannot generate something as a human would from inference or imagination. https://www.nature.com/articles/d41586-023-00107-z
  • 23. @schachin Kristine Schachinger Important! ChatGPT Output Cannot Be Copyrighted. Not in the US or the EU, not yet anyhow.
  • 24. @schachin Kristine Schachinger Why? A human must create a creative work for copyright to apply.
  • 25. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Some ChatGPT Ideas for Marketers • SEO: Titles and Metas (okayed by Google) • PPC Landing Pages • Content Ideas • Content NOT for Search • Topical Funnels • Analyze Competitor Site Content • Product Descriptions • Customer Service / Chatbots • Emails • Outreach • Link Building • Lead Generation • Newsletters • Reviews digital marketer using AI to do SEO Disney style
  • 26. @schachin Kristine Schachinger ALWAYS USE HUMAN REVIEW. ChatGPT is not a search engine, has no fact checking function, and the core product was not trained on any data past 2021 so make sure someone reviews the output produced for errors, accuracy, factual information, and tone.
  • 27. @schachin Kristine Schachinger AI Content and Google. Don’t Blow Yourself Up. a human male person with the job of SEO Consultant yelling at Google in 3d cartoon form
  • 28. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: AI Content and Google. Google says AI Content is okay IF it provides value and it not “spammy”, But since it is writing what it trained on how does provide value?
  • 29. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Detection and Google. How does Google define “Spammy” content?
  • 30. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Detection and Google. https://developers.google.com/search/blog/2022/08/helpful-content-update Google and the Helpful Content Update.
  • 31. @schachin Kristine Schachinger Play it safe, do not use ChatGPT or AI to write article content. Use humans to write/edit your content. They can use ChatGPT as a tool to generate ideas, however.
  • 33. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT. The Application of ChatGPT Some of the potential applications of ChatGPT include: • Text generation: ChatGPT can be used to generate human-like text responses to prompts, which makes it useful for creating chatbots for customer service, generating responses to questions in online forums, or even creating personalized content for social media posts. • Language translation: ChatGPT can also be used for language translation tasks. By providing the model with a text prompt in one language and specifying the target language, the model can generate accurate and fluent translations of the text. • Text summarization: ChatGPT can be used to generate summaries of long documents or articles. This can be useful for quickly getting an overview of a text without having to read the entire document. • Sentiment analysis: ChatGPT can be used to analyze the sentiment of a given text. This can be useful for understanding the overall tone and emotion of a piece of writing, or for detecting the sentiment of customer feedback in order to improve customer satisfaction.
  • 34. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT. Fine Tuning in ChatGPT Training. ChatGPT was trained with the following focuses: • Code • Content Writing • Poetry • Music This means it tends to create these items well. Fine tuning allows ChatGPT to focus in on certain types of functions and may not do as well with others.
  • 36. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: InstructGPT: ChatGPT Prompts. Prompts. The MOST Important Part of Using ChatGPT. Prompts provide a starting point for ChatGPT to generate text based on the input given. They can be used to guide the model's output and control the context of the generated text. Prompts? Simply means, how you tell ChatGPT what to write. This relies on another language model called InstructGPT, which allows you to give ChatGPT detailed and specific instructions. https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
  • 38. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > Articles.
  • 39. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > Articles + Style.
  • 40. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Content > FAQs.
  • 41. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Code.
  • 42. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Summarization > URLs.
  • 43. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Summarize URLs.
  • 44. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Write Emails.
  • 45. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Lyrics.
  • 46. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Poetry.
  • 47. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: Using ChatGPT Prompts: Ideas.
  • 48. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Prompts. Prompts. How to write ChatGPT Prompts • Give Context • Be Clear • Be Concise • Be Specific • Be Descriptive • Define Format • Define Length • Define Style
  • 49. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Prompts. Prompts. Some great ideas for how you can use ChatGPT • 100 Best ChatGPT Prompts • https://mpost.io/100-best-chatgpt-prompts-to-unleash-ais-potential/ • Web Development ChatGPT • Music ChatGPT • Business ChatGPT • Educational ChatGPT • Comedy ChatGPT • History ChatGPT • Health&Medicine ChatGPT • Art ChatGPT • Food & Cooking ChatGPT • Marketing ChatGPT • Game ChatGPT Prompts https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
  • 51. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Issues? + Programmed to Avoid Certain Kinds of Content • ChatGPT is specifically programmed to not generate text on the topics of graphic violence, explicit sex, and content that is harmful such as instructions on how to build an explosive device. • And swearing + Unaware of Current Events • not aware of any content that is created after 2021.  Has Built-in Biases • This can be just in the writing style, not necessarily cultural or racial though they can exist as well + ChatGPT Requires Highly Detailed Instructions • You must be very specific with ChatGPT Prompts to really get good output + Mediocre Content • ChatGPT is designed to sound confident, but the content it writes is not high quality. + Suffers from “Hallucinations” • Makes stuff up https://www.searchenginejournal.com/chatgpt-for-content-and-seo/473823/#close
  • 53. @schachin Kristine Schachinger https://www.wired.com/story/chatgpt-fluent-bs/ All of this makes playing around with ChatGPT incredibly fun, charmingly addictive, and—as someone who writes for a living— really quite worrying. But you soon start to sense a lack of depth beneath ChatGPT’s competent prose. It makes factual errors, conflating events and mixing people up. It relies heavily on tropes and cliché, and it echoes society’s worst stereotypes. Its words are superficially impressive but largely lacking in substance— ChatGPT mostly produces what The Verge has described as “fluent bullshit.” But that kind of makes sense. ChatGPT was trained on real-world text, and the real world essentially runs on fluent bullshit. “Bovine Feces”
  • 55. @schachin Kristine Schachinger Google https://www.semanticscholar.org/paper/Rethinking-Search%3A-Making-Domain-Experts-out-of-Metzler-Tay/e3d7065d6196437259483ed8b9b70ba4b4bb76ad “…they do not have a true understanding of the world, they are prone to hallucinating, and crucially they are incapable of justifying their utterances by referring to supporting documents in the corpus they were trained over.” Image from DALL-E
  • 56. @schachin Kristine Schachinger Google https://www.semanticscholar.org/paper/Rethinking-Search%3A-Making-Domain-Experts-out-of-Metzler-Tay/e3d7065d6196437259483ed8b9b70ba4b4bb76ad “…they do not have a true understanding of the world, they are prone to hallucinating, and crucially they are incapable of justifying their utterances by referring to supporting documents in the corpus they were trained over.” Image from DALL-E “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”
  • 58. @schachin Kristine Schachinger ALWAYS USE HUMAN REVIEW. ChatGPT is not a search engine, has no fact checking function, and was not trained on any data past 2021 so make sure someone reviews the output produced for errors, accuracy, factual information, and tone.
  • 60. @schachin Kristine Schachinger Some of the general parameters that are used to train language models include: ChatGPT Docs. ChatGPT Guide • All the basics you need to know about how to use ChatGPT • https://www.digitaltrends.com/computing/how-to-use-openai-chatgpt-text-generation-chatbot/#google_vignette OpenAI ChatGPT API Guide. • Everything you need to know about using the API version • https://beta.openai.com/docs/guides/completion/inserting-text Some cool things people have done with ChatGPT • https://www.bleepingcomputer.com/news/technology/openais-new-chatgpt-bot-10-coolest-things-you-can-do-with-it/
  • 61. @schachin Kristine Schachinger Kristine@SitesWithoutWalls.com "Unlocking the Power of ChatGPT” -- without blowing yourself up -- “…they are prone to hallucinating, and crucially they are incapable of justifying their utterances…”