Post B68JKwhnwsPbSkp49w by cassidy@blaede.family
 (DIR) More posts by cassidy@blaede.family
 (DIR) Post #B68JKvMUwbm5INqXJI by cassidy@blaede.family
       0 likes, 1 repeats
       
       LLMs aren’t conscious, even if giant companies incentivized to keep up the charade continue to name features after human experiences. LLMs are designed to trick you into thinking they’re conscious and human-like. They’ve encapsulated the worst qualities in the process: they know nothing and don’t think, yet they appear to be confident while outputting falsehoods. They output that they’re smart because they regurgitate the likely text response of being told they’re smart.https://www.wired.com/story/i-am-begging-ai-companies-to-stop-naming-features-after-human-processes/
       
 (DIR) Post #B68JKwhnwsPbSkp49w by cassidy@blaede.family
       0 likes, 0 repeats
       
       It turns out: LLMs *are* a pretty powerful technology for certain use cases! The frankly bonkers amount of data and training and power and scraping and iteration and ingesting copyrighted works and tuning and tooling has led to a technology that can often convincingly imitate humans and do lots of text-based stuff. But you must always remember that it fundamentally is a technology built around text prediction. It is nondeterministic. It does not think. It cannot be blindly relied on.