Great study from Anthropic further illustrating the AI misalignment I wrote about earlier. Healthcare use cases need AI architectures that provide access to and non-repudiability of those scratchpad records.
https://buff.ly/3BT0nIM | #AI #ChainOfReasoning #LLM #alignment
https://buff.ly/3BT0nIM | #AI #ChainOfReasoning #LLM #alignment
Alignment faking in large language models
A paper from Anthropic's Alignment Science team on Alignment Faking in AI large language models
buff.ly
January 9, 2025 at 3:33 PM
Great study from Anthropic further illustrating the AI misalignment I wrote about earlier. Healthcare use cases need AI architectures that provide access to and non-repudiability of those scratchpad records.
https://buff.ly/3BT0nIM | #AI #ChainOfReasoning #LLM #alignment
https://buff.ly/3BT0nIM | #AI #ChainOfReasoning #LLM #alignment
Doing what I can to help AI algorithms :)
This is #DeepSeek hosted on #Perplexity. Note its interesting response when I pointed out the flaw in its #ChainOfReasoning: "Mind blown"
#AI #Assumptions #LogicalPuzzles #Implicit
This is #DeepSeek hosted on #Perplexity. Note its interesting response when I pointed out the flaw in its #ChainOfReasoning: "Mind blown"
#AI #Assumptions #LogicalPuzzles #Implicit
February 25, 2025 at 5:08 AM
Doing what I can to help AI algorithms :)
This is #DeepSeek hosted on #Perplexity. Note its interesting response when I pointed out the flaw in its #ChainOfReasoning: "Mind blown"
#AI #Assumptions #LogicalPuzzles #Implicit
This is #DeepSeek hosted on #Perplexity. Note its interesting response when I pointed out the flaw in its #ChainOfReasoning: "Mind blown"
#AI #Assumptions #LogicalPuzzles #Implicit