Limitations of ChatGPT

Understanding ChatGPT

James Chapman

Curriculum Manager, DataCamp

ChatGPT under the hood

Demystifying the LLM

 

Demystifying the LLM

 

Demystifying the LLM

Demystifying the LLM

Demystifying the LLM

Demystifying the LLM

Demystifying the LLM

Limitation 1 - Knowledge cutoff

 

 

• Trained on data from up to a certain date:
  • GPT 3.5: January 2022
  • GPT 4: April 2023
• Isn't aware of events beyond this date

 

Limitation 2 - Training data bias

 

• ChatGPT was on a huge text dataset, including:

  • Books
  • Articles
  • Websites

• Model may learn the biases from the training data

• Could bias the responses

Limitation 3 - Context tracking

 

Limitation 3 - Context tracking

 

Limitation 3 - Context tracking

 

• Struggles to keep track of the context if the focus shifts
• Can lead to inaccurate or irrelevant results
• Tip: Keep conversations to a single topic

 

Limitation 4 - Hallucination

 

• Model confidently provides inaccurate information
• Often occurs when trying to go beyond the model's knowledge or abilities

¹ https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

Limitation 5 - Legal and ethical considerations

 

• Example: Creating a song in the style of an existing artist
  • Who owns the new song?

 

Limitation 5 - Legal and ethical considerations

 

• Example: Creating a song in the style of an existing artist
  • Who owns the new song?

 

Limitation 5 - Legal and ethical considerations

 

• Example: Creating a song in the style of an existing artist
  • Who owns the new song?

 

• Easy to fall into a legal gray area
• Ownership and privacy → Chapter 2

 

Let's practice!

Understanding ChatGPT