i think an additional thing to note is that we call them hallucinations because of the propensity of these ideas to mislead their holder(or in the case of AI, reader). this means that we need to come up with new names for hallucinations if they are a feature
the statistical limitations of creating intelligence are well captured in this year's Breakthrough Discuss. Check out the playlist, especially Roger Penrose's remarks on why intelligence is not a computational process: https://www.youtube.com/playlist?list=PLyF3OMOiy3nEM6rWMGZ9oTbFX2Kt5W_I3
i have a few more thoughts on the subject, let me school myself and put it into a lesswrong
Yeah man, can't wait to read it. I love LessWrong myself.
i think an additional thing to note is that we call them hallucinations because of the propensity of these ideas to mislead their holder(or in the case of AI, reader). this means that we need to come up with new names for hallucinations if they are a feature
Word—that's why we are preferring "confabulations," as it captures the true essence.
the statistical limitations of creating intelligence are well captured in this year's Breakthrough Discuss. Check out the playlist, especially Roger Penrose's remarks on why intelligence is not a computational process: https://www.youtube.com/playlist?list=PLyF3OMOiy3nEM6rWMGZ9oTbFX2Kt5W_I3
Awesome. Thanks. Let me look it up!