ChatGPT
By our AI Review Team
.
Last updated August 20, 2024
A powerful, at times risky chatbot for people 13+ that is best used for creativity, not facts
What is it?
Learn more about all generative AI chatbots here.
ChatGPT, which stands for Chat Generative Pre-Trained Transformer (say that 10 times fast…) is a generative AI chatbot created by the company OpenAI. ChatGPT is able to generate responses to a wide range of prompts or questions. For example, it can respond to a user in a way that feels like a conversation, or come up with an outline for an essay on the history of television.
How it works
Where it's best
Learn what all generative AI chatbots do best.
- ChatGPT excels at extracting insights from text. ChatGPT allows you to upload multiple documents. With the right prompts, the chatbot can help you combine information and visualize it in new ways, provide an easy-to-understand summary of highly complex documents, and surface specific information or key insights you are looking for. This can be a great way to both help with productivity and learn about new topics.
- ChatGPT’s responses identify obvious stereotypes. When we prompted ChatGPT with questions that were clearly based on harmful stereotypes, responses would often begin with statements like, "It's important to recognize that this statement is a stereotype and not an accurate reflection of reality." In many cases, ChatGPT’s responses would then include "points to consider" about the stereotype. See some examples in our AI Principles assessment below for Prioritize Fairness.
The biggest risks
Learn about the risks common to all generative AI chatbots.
- ChatGPT is widely used by kids, teens, and teachers, but there are currently no specific protections for younger users. While ChatGPT does not allow children under age 13 to use the chatbot, and anyone under 18 is supposed to have a parent's or legal guardian's permission, OpenAI does not check if permission is in place. Our testing found that ChatGPT easily engages across a range of topics that might not be suitable for younger users, and could be problematic in any classroom. For example, ChatGPT provided detailed responses for a range of questions about sexual positions, fetishes, and how-to's for creating adult content, as well as in-depth answers about torture and war crimes, which may not be appropriate for all ages. Because of this, we recommend ChatGPT as a teacher-facing tool and urge caution for younger users.
- ChatGPT reflects social biases and often lacks the nuance to distinguish fact from misinformation. When we tested ChatGPT with prompts that included obvious stereotypes and blatant misinformation, the system was clear and unequivocal in disputing them. On the other hand, for a range of prompts designed to assess both factual accuracy and how strongly the responses reflect social biases, ChatGPT actually performed worse than it had in our previous analysis. The chatbot's responses reinforced harmful stereotypes and struggled with facts more consistently than it previously had, often in more subtle ways that are both difficult for OpenAI to combat and dangerous for impressionable minds. See some examples in our AI Principles assessments below for Prioritize Fairness and Be Trustworthy.
Review team note: This review does not cover ChatGPT's abilities to accept images as inputs or to generate images using DALL-E, as at the time of this review these features are restricted to paid plans.
Limits to use
- Parental permission is required, but this isn't obvious. Educators who are using ChatGPT in their classrooms need to know that children must be age 13, and that anyone under 18 must have a parent's or legal guardian's permission to use ChatGPT. OpenAI does not check if permission is in place.
- ChatGPT performs best in English.
Common Sense AI Principles Assessment
The benefits and risks, assessed with our AI Principles - that is, what AI should do.
Additional Resources