Select Page

AI Authoring Tools

Faculty will need to use caution when considering AI Authoring tools with their students.  The following should be considered before implementing them in a course: 

  • Any software or tool that requires students to enter information must go through the Courseware Review process to ensure that university privacy, security, accessibility standards are met. 
  • The instructor must provide clear guidelines about the acceptable and unacceptable use of AI in their courses. These guidelines should be stated in the syllabus and the assignment instructions. 
  • The instructor should include a module on Artificial Intelligence Literacy to ensure that students are aware of the benefits and risks associated with using the AI. 
  • The instructor should only use assignments where students enter level 1 information per PSU policy AD95. 
  • The instructor must permit students to opt out of using AI and provide an alternate assignment. 

Following are examples of AI generation tools available as of January 8, 2023. Venture capital firm NFX identifies over 450 startups building generative AI (Currier, 2022). Keywords: AI generation tools, AI content creation tools; AI content generator + text, art, image, video

Text

Image

Video

Music

Other

  • Elicit: AI research assistant for research tasks
  • PHIND: AI search engine for developers
  • Perplexity: AI search engine
  • ELIZA: an early AI chatbot for psychotherapy (1960s)
  • Debuild: AI for writing code for React apps
  • Krisp: Removes background voices, noises, and echo from calls
  • Cleanvoice: Automatically edit podcast episodes
  • Podcastle: Studio-quality recording from your computer
  • Flair: Design-branded content
  • Patterned: Unique patterns, the digital version of cloth swatches
  • CopyMonkey: Create Amazon listings
  • Ocoya: Create social media content
  • Unbounce: Write digital ads, blog posts, sales emails
  • Vidyo: Make short videos from long ones
  • Maverick: Create video for boosting ecommerce
  • QuickChat: For building AI assistants
  • PuzzleLabs: Create an AI-powered glossary
  • Mem: Ai-powered self-organizing workspace
  • LawDroid: AI legal assistant to research legal issues, draft emails/letters, summarize docs
  • NoCode: Create personalized GPT-3 model
  • MarketingBlocks: Create marketing collateral
  • Hyyype: AI marketing assistants
  • Explainpaper: Upload a paper, highlight confusing text, get an explanation
  • Tome: Create presentation slides
  • Calligraphy.ai: Creates handwriting from typed text
  • Nolej: AI eLearning authoring tool
  • Verse by Verse: Free, experimental AI for writing poetry, from Google AI Semantic Experiences

See also: Antler: The Generative AI Landscape

AI Text Detection Tools (“AI-giarism detection)

  • GPT-2 Output Detector (From Open AI, the makers of ChatGPT) Claims a detection rate of 05% for machine-generated text using GPT-2. OpenAI recommends using in conjunction with additional approaches. (OpenAI claims 95% accuracy)
  • Giant Language Model Text Room (GLTR) (From Harvard and IBM) (Strobelt & Gehrmann, 2019) Developed to test for GPT-2. Detects likelihood that words were predicted by a bot. Generates a certain level for each. Color-coded results to aid interpretation (green and yellow, high likelihood of automated generation; red and purple, unlikely predictions and strong indicators of human-written text). 
  • Originality.AI, combines plagiarism and AI detection, pricing $0.01 per credit, 1 credit scans 100 words. Trained to predict content generated with GPT versions at 94% accuracy. 
  • GPTZeroX: Created by Princeton University student Edward Tian for educators. Supports large text inputs and file uploading, claims to identify portions written by AI. Scores on “perplexity” (measure of randomness determined by how likely the next word could have been suggested by a bot–the lower the score, the more likely computer generated) and “burstiness” (measures spikes of perplexity–variations in sentence length and complexity are more likely written by humans). The closer to 0 score, the greater the likelihood that the text was computer-generated. False-positive rate estimated 2%. Now in 2nd version.
  • AI Classifier: (From Open AI, the makers of ChatGPT) Indicates likelihood that text is automated or human written. Reliability not guaranteed: 26% true positive rate; 9% false positive rate. Reliability improves with longer text, but still not guaranteed. Errs on the side of human-written. Unreliable for non-English text or code. Not recommended for use as a primary decision-making tool. 

Related Tools

Keras Tokenizer: https://keras.io/api/preprocessing/text/ Quillbot: https://quillbot.com/ (text spinning tool)

Thanks to the Center for eLearning Initiatives at Penn State Erie, the Behrend College for their support in compiling this list.