3 ms·
Users don't need to directly interact with the language model. I think the article's saying that good UX shouldn't rely on the user having a great grasp on lang
by potatoman22 3y ago
Users don't need to directly interact with the language model. I think the article's saying that good UX shouldn't rely on the user having a great grasp on language. Engineers can abstract the function of the language model behind better things than a chat interface.
- bloaf 3y agoI am reminded of the scene in Idiocracy[1] where we see the receptionist picking conditions from a board containing a few dozen colorful icons. There certainly is a sense in which you could abstract e.g. medical diagnostics in a way that makes illiterate people better at getting an LLM to give advice to a patient. But to go back to my 3-part statement: do people who can't organize and express their thoughts really deserve to be in charge of your medical diagnosis? Or in other words, are UX designers catering to an audience that ought to not exist in the first place. [1] https://www.youtube.com/watch?v=LXzJR7K0wK0 https://www.youtube.com/watch?v=LXzJR7K0wK0
- potatoman22 3y agoYes. Most LLM-driven tools shouldn't need the user to engineer a prompt so they can function well. GitHub Co-pilot should work without you having to prompt the LLM in your code.
- cpill 3y ago"better things than a chat interface" like what? what interface comes more naturally to humans than the spoken word?