9). The Danger of Overthinking – This paper investigates overthinking in Large Reasoning Models (LRMs)—a phenomenon where models prioritize extended internal reasoning over interacting with their environment.
SAFETY
-
AI Impact on Human Cognitive Abilities Study Raises Concerns
By
–
Vous avez mille fois raison @DDupagne ! Mais on parle de facultés cognitives et je trouve cela encore plus inquiétant. Et une étude le documente, même si elle est imparfaite car basée sur un échantillon de personnes restreint.
-
AI Hallucinations: Insufficient Punishment for Repeated Offenses?
By
–
Hmm… Is the punishment for this not high enough that they keep doing it? Are the culprits that unaware of the problems with hallucinations?
-
Collaboration: The Irreplaceable Human Skill in AI Era
By
–
🤝 COLLABORATION: THE FUTURE OF WORK STARTS YOUNG
— Pascal Bornet (@pascal_bornet) 22 février 2025
In a world where AI is automating tasks at lightning speed, one human skill remains irreplaceable—collaboration.
Teaching kids to work together isn’t just about group projects or playground games. It’s about shaping the way they… pic.twitter.com/gMETfgWpBTCOLLABORATION: THE FUTURE OF WORK STARTS YOUNG In a world where AI is automating tasks at lightning speed, one human skill remains irreplaceable—collaboration. Teaching kids to work together isn’t just about group projects or playground games. It’s about shaping the way they
-
Concerns About Exaggerated Robot Capability Claims in Marketing
By
–
This feels very deceptive: clips that suggest far more capability than today’s technology can deliver. Walking, carrying a basket, sitting, yes. Can it do chores like cleaning a kitchen, putting away groceries, chopping vegetables? Show us something actually useful.
-
Stanford Researchers Develop AI Fairness Awareness Benchmark Suite
By
–
Scholars developed a new benchmark suite aimed to test difference awareness in AI. “One-size-fits-all definitions of fairness don’t work very well—we really have to have a more contextualized understanding,” said Stanford researcher @ang3linawang
. -
Tech-Facilitated Genocide: RightsCon Taipei Conference Sessions
By
–
Will you be Rightscon in Taipei? If so, say hi, I'll be attending the conference in person for the first time & will be participating in these 2 sessions All of them have to do with tech facilitated genocide. Such is the time we're in.
-
LLMs struggle with original creative writing versus academic prose
By
–
That’s why it’s still so bad at writing ‘original creative prose’ rather than an (incredible) academic report.
-
Skyfire AI Drone Saves Police Officer’s Life During Traffic Stop
By
–
Last month, a drone from Skyfire AI was credited with saving a police officer’s life after a dramatic 2 a.m. traffic stop. Many statistics show that AI impacts billions of lives, but sometimes a story still hits me emotionally. Let me share what happened. Skyfire AI, an AI Fund
-
SFT and RL Training Pipeline: Limitations of Cold Start Approach
By
–
Yes & no. They had SFT (cold start) → RL → SFT (CoT + knowledge) → RL. Not sure if they focused on providing solutions to hard problems via SFT CoT though, because that SFT data was just generated from the previous model (so if that model doesn't solve it, then you are stuck)