2 ms·
Why do people still ask LLMs niche factual knowledge, especially involving unique names? And even worse, judge its performance based off that? When someone do
by jdthedisciple 2y ago
Why do people still ask LLMs niche factual knowledge, especially involving unique names?
And even worse, judge its performance based off that?
When someone does that I immediately know that they have not the slightest clue about the underlying workings of any deep learning model, let alone basics of statistics.
- forgetfreeman 2y agoWhy? Because some combination of marketing flacks, influencers, and insipidly credulous technophiles keep advancing statements that strongly imply that these tools are intelligent and generally useful, despite being obviously curtailed in their utility and possessing no intellect at all. Next question?
- jdthedisciple 2y agoSo if the legitimately smartest person in the world happens to be ignorant about Brachiosaurus brancai then it means he's actually dumb?
- forgetfreeman 2y agoYou don't have to like the answer but your feels don't change the situation even slightly. Why are you emotionally invested here?
- criddell 2y agoBecause in some areas it works great. Ask it about Djikstra’s Algorithm and it does a pretty good job.
- soulofmischief 2y agoIt makes for great clickbait.