neuronerd1

4 ms·

1. 1.▲
   LLMs extract high-level semantic concepts from SVG and ASCII art(transformer-circuits.pub)

   3 points by neuronerd1 11mo ago|1 comments
2. 2.▲
   by neuronerd1 11mo ago

   Anthropic interpretability researchers find neural representations of the semantic components of text-based visual depictions.
3. 3.▲
   Understanding how AI models think(youtube.com)

   2 points by neuronerd1 1y ago|0 comments
4. 4.▲
   by neuronerd1 1y ago

   Could you build AI using these principles?? And then what would be the equivalent of giving an AI model ketamine?
5. 5.▲
   Sustain in the brain: conserved neural mechanisms of emotion from mice to humans(med.stanford.edu)

   7 points by neuronerd1 1y ago|3 comments
6. 6.▲
   Resolving a 2500-year-old quandary with neural landscape diffusion(news.stanford.edu)

   4 points by neuronerd1 3y ago|0 comments
7. 7.▲
   by neuronerd1 3y ago

   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