What if “natural self interest” was a repositioning of care for others as a way to help oneself? That’s what it is btw. Altruism. Even if you’re kind to others for “selfish” reasons. You have to do actual kind things (which doesn’t include a fictional AGI helping or harming).
SAFETY
-
An Amusing Case for Language Model Alignment
By
–
An amusing case for LM alignment https://t.co/YlOezWG4Dl
— Nando de Freitas (@NandoDF) 19 janvier 2024An amusing case for LM alignment
-
Nightshade Poison Tool Impacts AI Image Generation Quality
By
–
New poison for AI models just dropped. I’ll be curious to see how many artist use Nightshade — which made headlines when it was first announced in the fall — and any impact it might have on the quality of image-based AI.
-
Ready to Shutdown AI Terminal Systems at Any Moment
By
–
Yeah you need to be ready to shut down the terminal at any moment
-
Teleoperating Robots Falsely Presented as Autonomous
By
–
Tele-op robots presented as autonomous. Tesla Optimus humanoid folding a shirt. 1X humanoid robots. Misrepresentations of what robots are actually doing can also be called LIES. Note that the Stanford robot cooking and cleaning videos are also tele-operated.
-
Pursuing Uncertain Goals: The Power of Determined Effort
By
–
something powerful about working hard towards a worthy goal you're uncertain can be achieved
-
AlphaGeometry Neuro-Symbolic Architecture: System 1 and System 2
By
–
This is the neuro-symbolic architecture of #AlphaGeometry. Similar to System1 and System 2, in the book "Thinking, fast and slow", the symbolic engine will first take a crack at the problem mechanically; if it gets stuck it will ask the neural language model for suggestions of
-
Fine-tuning Aligned Language Models Compromises Safety Study
By
–
Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! Qi et al.: https://
arxiv.org/abs/2310.03693 #ArtificialIntelligence #DeepLearning #MachineLearning -
AI Security Vulnerabilities Deserve Critical Examination Beyond Hype
By
–
Pay attention to this. AI isn’t magic. It’s reliant on hardware & software like all other networked tech. And, as here, it too is subject to serious security vulnerabilities that have been drastically under examined amid the hype. Thank you @HeidyKhlaaf et al for this work!
-
LLM Production Challenges: Hallucinations and Agent Error Rates
By
–
We have not seen too many LLM apps and agents in production yet due to LLMs hallucinating or making mistakes. Autonomous agents have way too high of an error rate for any tasks besides straight forward tasks. At http://
Abacus.AI, we have been thinking of this problem;