neuronerd1

Anthropic interpretability researchers find neural representations of the semantic components of text-based visual depictions.

Could you build AI using these principles?? And then what would be the equivalent of giving an AI model ketamine?

Author here, a key thing is that we didn't prescribe that the mechanism of exploration was the same, but rather we found that the AI agent explored poorly (i.e. unlike animals) until we included Curious Replay. Interestingly, we found that the benefits of Curious Replay also led to state of the art performance on Crafter.