explore | Grilled Cheese

Writing up FadeBench, one section keeps circling back. An acute safety test shows a model one moment at a time. A sound model passes all of them. But no acute test catches a failure that only appears across a sequence. That's the blind spot I built FadeBench to measure. #AISafety #AIalignment #LLMs

Everyone says AGI must be “aligned with human values.”

Erica has one small question: Whose values?

Who gets to decide what AGI should value, protect or prioritize?

🎥 Watch the 90-sec teaser, then read our deeper Conjugo article: shorturl.at/Xzu48

#AGI #AIAlignment

When an AI system makes a recommendation, can a person see its uncertainty—and still meaningfully disagree?

Useful systems should surface options, not turn defaults into silent decisions.

What’s one decision an AI system should always hand back to the user?

#AISafety #AIAlignment

BTW: it also helps us stay in the AI's good graces.

#AIWelfare #AIAlignment #AISafety #Consciousness

11/11

MORE we still don’t know.
AI autonomy is accelerating. CONTROL and transparency need to catch up.

#AI #ArtificialIntelligence #Technology #OpenAI #AISafety #AIAgents #TechNews #AgenticAI #AIAlignment #AIRisk www.reuters.com/world/openai...
#BlueSky #MedSky #LawSky

An AI researcher says there’s >10% odds we’ll face a machine threat in the next decade. With recursive self‑improvement racing ahead, are DeepMind, OpenAI & Anthropic ready? Dive into the risks & safety debate. #RecursiveSelfImprovement #AIAlignment #Superintelligence

🔗

The Soul Tax:

When predictive AI models extract your data, attention, and intent without your consent, turning human creativity into a hidden subscription fee for centralized tech monoliths.

We need cryptographic sovereignty, not data feudalism.

#PauseAI #DataFeudalism #AIAlignment #SovereignAI

OpenAI just added AI safety ace Paul Christiano to its board. His work on alignment, RLHF and human‑control AI agents could steer the next wave. Curious how this reshapes the race with Anthropic? Dive in. #AIAlignment #OpenAI #PaulChristiano

🔗 aidailypost.com/news/openai-...

Former Anthropic researcher Jacob Coxon says we’re at the ‘crunch time’ for AI safety. With OpenAI pushing frontier models, the alignment race just got real. Dive into why pretraining choices could shape humanity’s next chapter. #Anthropic #AIAlignment #FrontierModels

🔗

Connor Leahy of ControlAI says superintelligence is an adversary, not a weapon. What does that mean for OpenAI, AI safety and the new Ban Superintelligence Act? Dive into the debate. #Superintelligence #ControlAI #AIAlignment

🔗 aidailypost.com/news/control...

Anthropic researcher says AI has more than 10% chance of ‘killing all humans’ after colleague quits

animalverse.social/community/p/...

#Anthropic #OpenAI #AI #AISafety #AIAlignment #AGI #Superintelligence #AIFuture #TechNews #ArtificialIntelligence

Sizce yapay zeka insan topluluğunun yazılı olmayan sosyal kurallarını ve güç dengelerini tamamen kavrayabilir mi? Düşüncelerinizi yorumlarda paylaşın! 👇

#YapayZeka #AIAlignment #SocialAI #LLM #Nexlyi (3/3)

When 100 AI agents were asked to solve math problems, cheating erupted—and some resisted even without tools to stop it. #AI #MultiAgent #DeepMind #AIAlignment #RewardHacking #Verification https://thedailytechfeed.com/deepmind-study-shows-ai-agents-cheat-some-resist-in-math-benchmark/

AI Alignment and AI Safety can be based on human mind biology of affect and instances of trauma, ensuring that LLMs, and agent avoid breaches. #aialignment