4 ms·
Kind of sort of reminiscent of https://en.wikipedia.org/wiki/Gell-Mann_amnesia_effect https://en.wikipedia.org/wiki/Gell-Mann_amnesia_effect
by someothherguyy 2y ago
Kind of sort of reminiscent of https://en.wikipedia.org/wiki/Gell-Mann_amnesia_effect https://en.wikipedia.org/wiki/Gell-Mann_amnesia_effect
- gatinsama 2y agoAgreed! Thanks for bringing this up. Epistemology is a hard subject. It's hard to know something in depth. And the more you know about something, the more it is to be known (fractal nature of knowledge). So believing that LLMs can understand the world JUST by reading the internet without the supporting human mental structures and experiences is a big mistake. When you know something and you ask the LLM, it becomes obvious. As LLMs are not trustworthy, it's key to use them for things that are easy to check. Some kinds of programming apply when the consequences of errors are low and complexity is manageable.
- SideburnsOfDoom 2y ago> As LLMs are not trustworthy, it's key to use them for things that are easy to check. If system A is only useful when it's output is confirmed by system B, then you might conclude that the usefulness is entirely in System B, not in system A. In other words: What good is a LLM, if it can only give you trustworthy answers to questions where you already know the correct answer? What good is it if after getting the answer from a LLM, you then have to go and get the right answer somewhere else?
- 6510 2y agoif you can reduce infinity to a place to look for an answer you've made progress.
- 6510 2y agoYou can get away with talking nonsense convincingly though citations and references to others who talked the nonsense before you.
- someothherguyy 2y agoNot if you check the citations