Tutorial 3 - ChatGPT

Summary

Released in demo format on the 30th November 2022 and quickly picking up speed after that, ChatGPT is a large language model (LLM) produced by OpenAI, and represents a notable leap in both the interface and capabilities of conversational chatbots. While initially open for free access to anyone who created an account, later updates have been made accessible only to paid accounts. Other large tech companies have also started providing similar systems, such as Google's Bard, or customised interfaces, such as Microsoft's Bing Chat. This proliferation of powerful LLMs has had an impact across a a huge swathe of domains, and has raised a lot of both excitement and concern, over topics like misinformation or automated human experiences.

Core Readings

Lock, Samantha. (2022, December 5). What is AI chatbot phenomenon ChatGPT and could it replace humans? The Guardian.

https://www.theguardian.com/technology/2022/dec/05/what-is-ai-chatbot-phenomenon-chatgpt-and-could-it-replace-humans

Goetze, Trystan S. (2023, July 10). ChatGPT Reveals What We Value and What We Do Not. The Blog of the American Philosophical Association.

https://blog.apaonline.org/2023/07/10/chatgpt-reveals-what-we-value-and-what-we-do-not/

Alternative:

Heaven, Will D. (2023, March 3). The inside story of how ChatGPT was built from the people who made it. MIT Technology Review.

https://www.technologyreview.com/2023/03/03/1069311/inside-story-oral-history-how-chatgpt-built-openai/

Prompts (hah)

Each subgroup should discuss one of these prompts (with the letter matching your subgroup letter) before the tutorial.

  • B) ChatGPT has access to vast data from the internet in order to be able to process information about an enormous range of subjects, but it also required extensive data labelling work to try and identify harmful content. What problems, and for whom, can data usage on this scale cause?
  • C) Given that large language models have been accused of being well tuned for sounding convincing over being truthful, what consequences might the technology have for institutions built around reliability of information or other forms of trust? 
  • D) Many sources have started recommending the use of ChatGPT or other “generative” models for automating tasks such as writing (CVs, essays, etc.), question answering, or art. What might be lost to the user (or indeed wider society over time) in using these tools in these ways?
  • A) The release of ChatGPT took a lot of people, including in some ways its developers, by surprise. In what ways do you think the manner of release of a technology like this influences its potential to cause harm?
License
All rights reserved The University of Edinburgh