neuronerd1

Re:LLMs extract high-level semantic concepts from SVG and ASCII art on Friday October 24, 2025 @05:44PM (#45699412)

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

LLMs extract high-level semantic concepts from SVG and ASCII art

Posted by neuronerd1 on Friday October 24, 2025 @05:44PM
from the transformer-circuits.pub dept.

https://transformer-circuits.pub/2025/october-update/index.html#svg-cross-modal

( Read More... | 1 comment | 3 points | on HN )

Understanding how AI models think

Posted by neuronerd1 on Saturday August 16, 2025 @11:57PM
from the youtube.com dept.

https://www.youtube.com/watch?v=fGKNUvivvnc

( Read More... | no comments | on HN )

Re:Emergence of emotional response from sensory experience in humans a... on Saturday May 31, 2025 @06:22PM (#44147340)

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

Sustain in the brain: conserved neural mechanisms of emotion from mice to humans

Posted by neuronerd1 on Friday May 30, 2025 @01:02PM
from the med.stanford.edu dept.

https://med.stanford.edu/news/all-news/2025/05/emotions-eye-puff.html

( Read More... | 3 comments | 7 points | on HN )

Resolving a 2500-year-old quandary with neural landscape diffusion

Posted by neuronerd1 on Wednesday November 08, 2023 @01:37PM
from the news.stanford.edu dept.

https://news.stanford.edu/2023/11/08/mice-choose-eat-drink/

( Read More... | no comments | 2 points | on HN )

Re:“Self-reflecting” AI agents explore like animals on Thursday July 06, 2023 @08:07PM (#36625136)

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.