ChatGPT chatbot 

ChatGPT (or 'Chat Generative Pre-trained Transformer') is an AI-powered chatbot that is capable of understanding natural human language and generating detailed human-like information and answers questions through a conversational interface.

Built on a version of Open AI's GPT-3 large language model, ChatGPT was released as a prototype for general public use in November 2022. Voice chat and image generation (via DALLE-3) capabilities were incorporated in September 2023.

System 🤖


Documents 📃


Reviews 🗣️

Operator: OpenAI
Developer: OpenAI

Country: USA; Global

Sector: Multiple

Purpose: Generate text

Technology: Chatbot; NLP/text analysis; Neural network; Deep learning; Machine learning; Reinforcement learning
Issue: Accuracy/reliability; Anthropomorphism; Bias/discrimination; Confidentiality; Copyright; Environment; Employment; Mis/disinformation; Privacy; Safety; Security

Transparency: Governance; Black box; Complaints/appeals; Marketing; Privacy

Risks and harms 🛑

ChatGPT has been criticised for its inaccuracy, unreliability, bias, disregard for confidentiality and privacy, and negative impacts on politics, employment and the environment, amongst others.

Accuracy, reliability

ChatGPT is prone to generate factual inaccuracies, including those known as 'hallucinations' (or plausible-sounding inaccurate information such as fake references or sources), and can be inconsistent and unreliable. 

For example, the chatbot:

Anthropomorphism

ChatGPT's ability to generate plausible-looking, human-like information and advice runs the risk of convincing users that it is somehow human and/or 'sentient'. 

Bias, discrimination

Despite its built-in safeguards, researchers and others have demonstrated that ChatGPT produces outputs that are explicitly and implicitly biased in multiple ways.

Confidentiality

Unless they expressly opt-out, OpenAI uses users' information and data inputted to ChatGPT to train its models, thereby potentially or actually violating their personal confidentality, or the confidentiality of their employers.

Data privacy

Concerns exist about ChatGPT's approach to and impact on personal privacy. 

Employment

ChatGPT is seen to pose a risk to jobs across a swathe of industries, from law, finance, and marketing to academics, journalists, and software programmers.

OpenAI's development of ChatGPT involved poorly paid workers in third-world countries with limited rights.

Environment

OpenAI has not revealed the environmental impacts of its products, which are thought to be substantial. 

IP/copyright

ChatGPT has been accused of using copyright protected information without permission to train its language models and ChatGPT by a wide range of individuals and organisations. 

Misinformation and disinformation

ChatGPT has been shown to produce large volumes of convincing misinformation and disinformation:

Cheating and plagiarism


ChatGPT has fueled concerns that students can easily generate high-quality essays and other material, thereby reducing their need to build critical-thinking and problem-solving skills.

Politics

ChatGPT's ability to generate plausible-sounding misinformation, disinformation, and hate speech is seen to have potentially serious effects on the well-being of communities and democracy. 

Reputational damage

ChatGPT can harm the reputation of individuals and organisations it generates information about, in some instances resulting in threats to sue OpenAI for defamation. 

ChatGPT has also been shown to harm the reputation of individuals and organisations using it, including when used naively. 

Safety

ChatGPT has safeguards to minimise it being tricked into making errors and to stop requests generating inappropriate, abusive, or offensive content. However, it appears some of these can be overridden relatively easily. 

Security

ChatGPT has been shown to be vulnerable to malfunction and manipulation, and can easily be made to generate spam and malware. The chatbot: 

Transparency 🙈

OpenAI may have been open about some of ChatGPT's limitations, but its governance and algorithms are - and appear likely to remain - carefully concealed, commercially-driven black boxes. 

Page info
Type: System
Published: December 2022
Last updated: May 2024