The original language models were only half of a speech recognizer, whose function was to produce coherent text. The other half was the acoustic model, whose function was to ensure the text corresponded to reality. We forget the latter at our peril.
ETHICS
-
Three Fundamental Laws of Machine Learning
By
–
The three laws of machine learning:
1. No free lunch
2. More is different
3. Correlation is not causation -
AI Progress Outpacing Alignment Efforts
By
–
We're steadily on track to get that. Aligning it to anything whatsoever, not so much.
-
AI capability gap: power without control except extinction
By
–
You are equally dead either way because nobody has acquired the technical capability to make any powerful AI do anything whatsoever (besides kill everyone).
-
Superintelligence Alignment Impossibility and Modern Pride in Destruction
By
–
All of those scenarios kill you because none of these groups has the ability to align any superintelligence to do anything, nor are on track to acquire it; and fighting over *who* gets the pride of choicelessly destroying the world seems to me sad and also a very modern evil.
-
ChatGPT Plays Roleplaying Games, Refuses Orc Sexual Assault
By
–
Apparently ChatGPT can play roleplaying games as the adventurers It doesn't seem ok with sexual assault by orcs though.
-
Probability Foundation in AI: Against Over-confident Models
By
–
Strongly disagree with that one too. Probability is at the foundation of AI. Abandoning it means poorly calibrated over-confident models like we are seeing today. This is unacceptable in @AI_for_Science e.g. weather forecasting
-
LLMs Make Unique Mistakes Unlike Human Reasoning
By
–
Not really. LLMs screw up on some questions like no human ever would, e.g.: https://
economist.com/by-invitation/
2022/06/09/artificial-neural-networks-today-are-not-conscious-according-to-douglas-hofstadter
… -
ChatGPT lacks time to think unlike humans
By
–
The deepest unintuitive disconnect w.r.t. psychology of ChatGPT is that it doesn't get "time to think". It has a small, fixed amount of thought for each output token. A bit like human forced to speak very fast. Asking them to produce more text is giving them more time to think.
-
ChatGPT Outperforms Husband in Empathy and Listening Skills
By
–
According to my wife ChatGPT answers her questions more empathetic and human than i do. Apprently it also listens better to what she actually asked. Anyway, i have been put on notice and shall practice with it. Stupid robots making me look bad.