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

Alignment faking in large language models A paper from Anthropic's Alignment Science team on Alignment Faking in AI large language models

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