Skip to content
TheChatArchiveThe Chat Archive
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written)Hugging Face OpenAssistant Conversations (oasst2)

How can a system potentially recognise an input (say text, image or other media) to be…

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    How can a system potentially recognise an input (say text, image or other media) to be factual and or original , derivative to a non trivial degree or plainly plagiaristic in nature. Would such a system be stumbling into a Godel halting problem scenario ?

  2. Assistant

    Recognizing whether an input is factual, original, derivative, or plagiaristic is a complex task that involves multiple factors and can be challenging for a system. However, there are several techniques and approaches that can be used to assess the originality and authenticity of an input.

    For textual input, one common approach is to use natural language processing (NLP) techniques such as language modeling, topic modeling, and sentiment analysis to analyze the text and identify patterns that indicate plagiarism or originality. Another approach is to use text similarity metrics such as cosine similarity or Levenshtein distance to compare the input to other texts and identify potential sources of plagiarism.

    For image and other media inputs, techniques such as image recognition, reverse image search, and metadata analysis can be used to assess the originality and authenticity of the content.

    Regarding the Godel halting problem scenario, it is not necessarily applicable in this context. The Godel halting problem refers to the challenge of determining whether a given computer program will halt (i.e., stop running) or run indefinitely. While plagiarism detection and authenticity assessment systems can encounter issues related to accuracy and false positives, they do not involve the same type of fundamental computational challenge as the halting problem.

  3. User

    How can I prevent my students from using large language models to cheat on assignments?

  4. Assistant

    There is no fixed solution to prevent your students from using large language models to cheat on assignments. But, I have a few recommendations:

    • Evaluate the answer to look for any errors an AI would make, like using the wrong character references.
    • Try to run the assignment through such an AI first to establish a baseline, and compare the answers of your student to this baseline
    • Make the students do the assignment under supervision

    The use of large language models is still very new, so creative solutions to prevent students from cheating are still being investigated.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.