I’d like to develop this into a paper, though I’m slightly reticent about using examples of these myths because it risks being deeply critical of people I like:
- LLMs are trained on the ‘entire internet’
- Students will enter a working world filled with generative AI
- LLMs are ‘bullshit machines’ incapable of producing meaningful content
- There are no practical uses for LLMs beyond, maybe, coding
- LLMs constitute a strange form of alien intelligence
- There’s a skill called ‘prompt engineering’ we need to teach students
- LLMs learn and grow in response to user prompts
