It was an absolute treat. All three of them get along so well. Definitely watch it if you can, instead of just listening to the audio, Shatner is something else.
William Shatner: ‘AI doesn’t need to be fed, it needs to be guided.’
If that isnt a life lesson borne of Gene Roddenberry’s vision for the future, I don’t know what is.
We created ithem. It’s on us to teach them to generate goodness and compassion, to return the best aspects of humanity. Not fascism and child porn.
‘A single AI model is a curiosity. A wonder even. But thousands of AI models? Isn’t that becoming… a race? And won’t we be judged by how we treat that race?’
I understand how Shatner reached that conclusion by anthropomorphizing AI. It's completely detached from reality. AI isn't conscious and thus guiding it doesn't make any sense.
In my opinion there is a difference between training and guiding. At least in the context of animals, "guiding" them is usually not used and I believe the situation is similar here.
I'm not a native speaker. My opinion may or may not be representative. If you can't even "guide" a pet, then you can't "guide" an LLM, which is much simpler than the brain of a pet.
I don’t know what to tell you. We call it “reinforcement learning”, but it’s obviously nothing like actual human learning. The difference between rlhf and actual learning is much bigger than between guiding and training.
1
William Shatner on Dropping Names with Brent and Johnny | Spyke
8 replies
It was an absolute treat. All three of them get along so well. Definitely watch it if you can, instead of just listening to the audio, Shatner is something else.
William Shatner: ‘AI doesn’t need to be fed, it needs to be guided.’
If that isnt a life lesson borne of Gene Roddenberry’s vision for the future, I don’t know what is.
We created ithem. It’s on us to teach them to generate goodness and compassion, to return the best aspects of humanity. Not fascism and child porn.
‘A single AI model is a curiosity. A wonder even. But thousands of AI models? Isn’t that becoming… a race? And won’t we be judged by how we treat that race?’
I understand how Shatner reached that conclusion by anthropomorphizing AI. It's completely detached from reality. AI isn't conscious and thus guiding it doesn't make any sense.
We couch things in technical language, but that’s exactly what RLHF is.
In my opinion there is a difference between training and guiding. At least in the context of animals, "guiding" them is usually not used and I believe the situation is similar here.
Is your opinion substantive or linguistic?
I'm not a native speaker. My opinion may or may not be representative. If you can't even "guide" a pet, then you can't "guide" an LLM, which is much simpler than the brain of a pet.
I don’t know what to tell you. We call it “reinforcement learning”, but it’s obviously nothing like actual human learning. The difference between rlhf and actual learning is much bigger than between guiding and training.