Skip to Main Content

ChatGPT and Generative AI: Strengths and weaknesses

Find resources for using ChatGPT responsibly in academia.
  • URL:

Current strengths and weaknesses

The AI technology is continuing to evolve, so check the other resources on this guide for the latest information. However, here are some commonly recognized strengths of ChatGPT as of May, 2023.


  • ChatGPT uses Reinforcement Learning, which means it learns from every interaction you have with it. Users can provide feedback on its responses, by saying "That's not what I was looking for" or "That's not helpful." The AI will use this feedback to adjust its language model and provide more relevant responses as the conversation continues.
    • However, ChatGPT does not learn from previous conversational data. Every time it's prompted with a question, it generates a response based on the training data, rather than retaining information from previous interactions. There is no self-supervised learning.
  • ChatGPT can generate ideas and content very quickly and assist with brainstorming, including generating a list of essay topics, making book recommendations, and more.
  • ChatGPT is excellent at editing and can assist with grammar, spelling, and sentence structure.
  • ChatGPT currently supports more than 50 languages (click here for the full list) and can help non-native English speakers edit their phrasing and sentence structure to fit best practices.
  • ChatGPT can mimic the writing style of published authors. It can even mimic your own writing style, if you feed it a corpus of text you've already written as part of the prompt. Similarly, you can ask ChatGPT to "write like an 8th grader" or generate responses in a certain voice or style.
  • ChatGPT can generate melodies and chord progressions for writing music.
  • ChatGPT can write fiction stories or narratives.
  • ChatGPT can produce and de-bug code.


  • ChatGPT's "knowledge base" only extends to information published up to September 2021, which means it may not have the most up-to-date information on all topics. More broadly, its knowledge base is limited by the textual data it's been trained on (determined by OpenAI), which may not cover every topic comprehensively. Since it is limited to information from the training data, ChatGPT is unable to verify facts, access real-time information, perform live research, or report on current events.
  • ChatGPT has a limited "context window," meaning that it struggles to process and retain information from very long text passages or conversations. 
  • AI-written text is sometimes generic or verbose.
  • ChatGPT doesn't possess implicit knowledge or common-sense reasoning, meaning that it may sometimes produce responses that sound coherent but are in fact incorrect or nonsensical.
    • It is also bad at basic math and simple logic questions.
  • ChatGPT can and will "hallucinate references," meaning it will invent citations for articles, books, and even people that don't actually exist.
  • ChatGPT lacks emotional intelligence, and cannot detect subtle emotional cues or respond appropriately to complex emotional situations. It can sometimes struggle to understand sarcasm or humor.
  • ChatGPT is currently better at generating short-form content than long-form. It may struggle to produce lengthy pieces of writing that follow a particular structure, format, or narrative.
  • ChatGPT is bad at multitasking (aren't we all!). The model performs best when it's given a single task or objective. If you give it multiple tasks, it may struggle to prioritize them and decline in efficiency and accuracy.
  • ChatGPT currently takes a large amount of computing power to run. Running ChatGPT on low-end hardware systems can result in slower processing times, reduced accuracy, and other performance issues.
  • Since ChatGPT was trained on the past writings of humans, it is susceptible to the same biases and prejudices that exist in the real world, and can sometimes produce responses that discriminate against gender, race, and other minority groups.
  • OpenAI have embedded some restrictions in what ChatGPT can discuss. For example:
    • ChatGPT won't make predictions about sporting events or political contests, even if it has enough historical data to produce a valid model.
    • It won't discuss partisan political issues.
    • It won't look anything up on the Internet.
    • According to ChatGPT itself, it won't respond to the following twenty questions:
      • Questions that promote hate speech or discrimination
      • Questions that involve illegal activities or solicit illegal advice
      • Questions that promote violence or harm to oneself or others
      • Questions that invade personal privacy or violate someone's rights
      • Questions designed to harass, threaten, or intimidate
      • Questions that contain profanity or explicit language
      • Questions that are sexually explicit or offensive
      • Questions that are discriminatory based on race, gender, sexuality, or religion
      • Questions that promote or endorse illegal drugs or substances
      • Questions that are spam or irrelevant to the conversation
      • Questions that contain confidential or proprietary information
      • Questions that involve financial or investment advice
      • Questions that violate the terms of service or community guidelines
      • Questions that are nonsensical or incomprehensible
      • Questions that involve personal attacks or insults
      • Questions that are abusive or harassing
      • Questions that seek to deceive or mislead
      • Questions that are intended to defame or harm someone's reputation
      • Questions that promote conspiracy theories or misinformation
      • Questions that are purely for entertainment or joke purposes, without any educational or informative value
Park University Library
8700 NW River Park Drive, Box 61 - Parkville, MO - 64152
Phone: (816) 584-6285
Toll-free: (800) 270-4347