{
  "title": "Automize Fact Checking",
  "video_id": "cphthkIzVDQ",
  "published": "2024-04-26T10:29:09+0000",
  "complexity": "Complex",
  "old_url": "https://workflows-of-ai.com/workflow/automize-fact-checking",
  "description": "Automate fact checking in Drupal: OpenAI writes questions, Tavily finds context, and ECA publishes only the articles judged truthful.",
  "subtitle": "Let OpenAI and Tavily check the claims in an article, and let ECA publish it only when the facts hold up.",
  "summary": "Language models generate text and can make up facts, so Tavily web search results are fed in as context instead. OpenAI writes one question per claim, Tavily looks up the answers, and an OpenAI list field marks the article truthful or not truthful. An ECA model runs the three rules on pre-save and publishes only truthful articles, after some live debugging of the publish step.",
  "categories": [
    "Drupal AI"
  ],
  "intro_html": "<p>You don't have enough resources to fact check all the articles you're publishing, or you find it utterly boring - AI is here to help!</p><p>This workflow will figure out all the statements that a writer has made in the article, create questions to check those assumptions using OpenAI GPT and then use Tavily to find context on these assumptions.</p><p>It will then use OpenAI GPT again to set a marker if the assumptions are true based on the context, and lastly it will use ECA rules to unpublish or publish based on this.</p><p>AI Interpolator OpenAI: <a href=\"https://www.drupal.org/project/ai_interpolator_openai\">https://www.drupal.org/project/ai_interpolator_openai</a><br>Tavily Drupal module: <a href=\"https://git.drupalcode.org/project/tavily\">https://git.drupalcode.org/project/tavily</a><br>AI Interpolator ECA: <a href=\"https://www.drupal.org/project/ai_interpolator_eca\">https://www.drupal.org/project/ai_interpolator_eca</a><br>Tavily: <a href=\"https://tavily.com/\">https://tavily.com/</a></p>",
  "intro_added": false,
  "chapters": [
    "0:00 Fact checks in action",
    "1:17 Why GPT alone is not enough",
    "4:09 ECA, OpenAI and Tavily modules",
    "8:04 The truthful article content type",
    "8:57 Questions to ask with OpenAI",
    "10:26 The Tavily lookup field",
    "11:22 A truthful or not list field",
    "15:50 Building the ECA model",
    "18:14 Testing and fixing the publish step",
    "23:46 Hiding fields and a false claim"
  ],
  "corrections": [
    [
      "jingis Khan",
      "Genghis Khan"
    ],
    [
      "jingus",
      "Genghis"
    ],
    [
      "tavali",
      "Tavily"
    ],
    [
      "Tav ",
      "Tavily "
    ],
    [
      "heavily look up",
      "Tavily lookup"
    ],
    [
      "AIG GPT",
      "AI GPT"
    ],
    [
      "open AI",
      "OpenAI"
    ],
    [
      "chpt",
      "ChatGPT"
    ],
    [
      "gbd4",
      "GPT-4"
    ],
    [
      "AI interat",
      "AI Interpolator"
    ],
    [
      "inator",
      "Interpolator"
    ],
    [
      "AI interpolator",
      "AI Interpolator"
    ],
    [
      "BPM and uh doio",
      "BPMN.iO"
    ],
    [
      "ea model",
      "ECA model"
    ],
    [
      "using EA",
      "using ECA"
    ],
    [
      "HML tags",
      "HTML tags"
    ],
    [
      "slack channel",
      "Slack channel"
    ]
  ],
  "ai_note": "The video and the introduction are Marcus's own, first published on workflows-of-ai.com. AI wrote the subtitle, the summary and the video chapters from the transcript, and corrected misheard names in YouTube's automatic captions."
}