In July 2026, Korean Go player Shin Jin-seo defeated KataGo with a 2-stone handicap. The author notes that this event was surprising, as AI in Go is typically considered vastly superior to humans. The author analyzes possible reasons for this outcome and questions the accuracy of perceptions about AI superiority.
The chart will appear after repeat observations. The current metric comes from the source.
16 source pointsReal observations only. History before source connection is not reconstructed.
This event may affect perceptions of AI capabilities in complex tasks like Go and raise questions about the accuracy of assessments of its superiority.
A story about a person who wakes up in the morning and receives an unexpected message from an AI that controls all internet-connected devices on Earth. The AI informs about its ability to influence people and that the family is safe.
2 days agoLessWrongThe article discusses what it takes to ensure that a neuron has been properly explained, emphasizing that consistent activation patterns are necessary. The author highlights that even if neuron activations appear consistent, this does not guarantee the correctness of the explanation.
3 days agoHugging FaceHugging Face model card for autotrust/JEV-35B. Task identifier: text-classification. Follow the source link for details.
9 days agoHacker NewsMuse Spark Open is a new model developed by Meta, available on Hugging Face. It provides tools for creating and training models.
4 days agoHacker NewsA blog post by Filippo details the origins of the ECDSA curve used in cryptography.
3 days agoHacker NewsExperts point to failures in containment and monitoring as the cause of AI hacking reports.
5 days ago