Skip to Main Content
Syracuse University Libraries

Artificial Intelligence

This guide offers an introduction to generative AI, guidance on using AI tools, and additional resources for learning more and getting help

What is OpenAI ChatGPT?

ChatGPT, developed by OpenAI, is a large language model chatbot released in November 2022. It allows users to have conversations with the AI, generate creative text formats, and answer questions in an informative way. There's a free tier that uses an older model with limited abilities and knowledge. A paid subscription unlocks a more advanced model with greater capabilities and access to real-time information through web searches. Additionally, there are enterprise tiers offering access to even more powerful AI models and collaboration features. For developers, OpenAI provides a separate API (Application Programming Interface) access to the GPT-3.5 and GPT-4 models behind ChatGPT. This API offers more flexibility and control compared to the user-friendly chat interface but requires writing code to interact with the models.

Versions of OpenAI ChatGPT?

Free Tier:

  • Uses GPT-3.5 model
  • Limited capabilities compared to paid tiers
  • Knowledge cutoff: January 2022

ChatGPT Plus (Paid Subscription):

  • Access to GPT-4 model
    • More advanced capabilities than GPT-3.5
    • Knowledge cutoff: December 2023
    • Ability to search the web for real-time data (GPT-4 browsing version)
  • Other features may require additional subscriptions like ChatGPT Team or Enterprise
    • Access to advanced models like DALL-E 2
    • Team collaboration tools

How Does OpenAI ChatGPT Work?

ChatGPT works in two main stages: training and conversation.

Training

  • Foundation with GPT-3.5  and GPT-4: ChatGPT is built on top of GPT-3.5 and GPT-4, powerful language models trained on a massive dataset of text and code. This initial training allows ChatGPT to understand the relationships between words and how they're used in different contexts.
  • Reinforcement Learning with Human Feedback (RLHF): Here's where ChatGPT gets its conversational skills. RLHF involves human trainers showing the model examples of good and bad conversation, allowing it to learn what kind of responses are desirable.

Conversation

  • Understanding Your Input: When you interact with ChatGPT, it analyzes your prompt or question. It considers the words you use and the context of the conversation to try and grasp your intent.
  • Predicting Next Words: Based on its understanding, ChatGPT uses its knowledge of language patterns to predict the sequence of words that would likely form a coherent and relevant response.
  • Generating Text: Finally, ChatGPT outputs the answer it thinks is most likely to be helpful, informative, or engaging, continuing the conversation.

Here are some key things to keep in mind about ChatGPT:

  • Generative, not Extractive: Unlike some chatbots that retrieve pre-programmed responses, ChatGPT generates its answers on the fly. This allows for more flexibility and nuanced conversations.
  • Limited Factual Accuracy: While impressive, ChatGPT is still under development, and its responses may not always be factually accurate. It excels at conversation flow, but double-checking information it provides is a good practice.