"Is there a single person on Earth who, like me, believes these two things at the same time? - AI risk is real, easy to imagine, and frankly terrifying - Regulation proposed is a bad idea because it won’t work or, more likely, will increase said risks https://gfodor.medium.com/to-de-risk-ai-the-government-must-accelerate-knowledge-production-49c4f3c26aa0 " Acceleration machine convergence chart https://twitter.com/quantizor/status/1764091695912620345 what do you think about alignment research being dual use, being in general controllability research, allowing for both good and bad actors to execute their goals on autonomous AI agents? https://akjournals.com/view/journals/2054/2/2/article-p97.xml [Researchers create AI worms that can spread from one system to another | Ars Technica](https://arstechnica.com/ai/2024/03/researchers-create-ai-worms-that-can-spread-from-one-system-to-another/) [Here Comes the AI Worm](https://sites.google.com/view/compromptmized) :thinking: [GitHub - lopusz/awesome-interpretable-machine-learning](https://github.com/lopusz/awesome-interpretable-machine-learning) Awesome Interpretable Machine Learning Opinionated list of resources facilitating model interpretability (introspection, simplification, visualization, explanation). Záleží jak definujete "rozumět". Například pro určitou část vědců je feature learning forma vnitřního rozmění. To dělají jak nebiologický stroje [GitHub - JShollaj/awesome-llm-interpretability: A curated list of Large Language Model (LLM) Interpretability resources.](https://github.com/JShollaj/awesome-llm-interpretability) [Feature learning - Wikipedia](https://en.wikipedia.org/wiki/Feature_learning) tak biologický stroje (lidi) v trochu jiných ale podobných formách. [Predictive coding - Wikipedia](https://en.wikipedia.org/wiki/Predictive_coding) Zvi Mowshowitz [Zvi Mowshowitz & Robin Hanson Discuss AI Risk - YouTube](https://www.youtube.com/watch?v=9XuVn6nljCM&pp=ygUOWnZpIE1vd3Nob3dpdHo%3D) [OpenAI Sora, Google Gemini, and Meta with Zvi Mowshowitz - YouTube](https://www.youtube.com/watch?v=ra3kpTVHjZs&pp=ygUOWnZpIE1vd3Nob3dpdHo%3D) [The Shift from Models to Compound AI Systems – The Berkeley Artificial Intelligence Research Blog](https://bair.berkeley.edu/blog/2024/02/18/compound-ai-systems/) Room Temperature Ambient Pressure Lead Apatite Superconductor https://twitter.com/mattparlmer/status/1764387559701188613 philosophy of mind https://pbs.twimg.com/media/GHxqvoPWUAAex6I?format=png&name=small Well of course "I" "know" "God", "he" might collectively and individually be everyone everywhere all at once at all times cyclically beyond all language, universe itself may be a gigantic neural network God that we're reverse engineering by physics embedded infinitely recursively in circular way as a simulation in a simulation, dream in a dream https://twitter.com/burny_tech/status/1764407623787393153 "Here’s a chain of assertions; - Climate change is happening - humans are the cause - It will destroy civilisation - We should de-grow the economy and only consume what we need to avoid it If you question any of this - you are labelled a denier. How about this? - climate change is happening - humans are the cause - it will gradually cause more issues over time - we can fix it - we will fix it - de-growth can’t solve it and would instead kill millions and make life miserable." https://twitter.com/PositivFuturist/status/1764244153792188929 Definitions of QTF https://twitter.com/Kaju_Nut/status/1764453203188531497?t=s5z4-euRvC0O9wQhDbDWqw&s=19 What is quantum field theory? Combination of classical field theory and quantum mechanics into one model! But physicists also use this label when it also combines special relativity in the standard model. But there I feel like there we should say instead that it uses "relativistic quantum field theory", or "special relativistic quantum field theory" to account for the fact that it uses a combination of classical field theory, quantum mechanics, and special relativity, but not general relativity. Symbol grounding Meaning in ML/linguistics [Prof. Geoffrey Hinton - "Will digital intelligence replace biological intelligence?" Romanes Lecture - YouTube](https://youtu.be/N1TEjTeQeg0?si=K2GNQWpKNFbpLuVa) https://arxiv.org/abs/2402.19450 https://fxtwitter.com/fly51fly/status/1764279536794169768?t=up6d06PPGeCE5fvIlE418Q&s=19 https://arxiv.org/abs/2402.18312 [OSF](https://osf.io/preprints/psyarxiv/d3gpf) Towards an Active Inference Account of Deep Meditative Deconstruction https://twitter.com/shawnprest/status/1764482933237383366?t=Ta9w9PhhOL0N7WMuB0Vslg&s=19 [Spontaneous collapse models lead to the emergence of classicality of the Universe | Journal of High Energy Physics](https://doi.org/10.1007/JHEP02(2024)193) https://twitter.com/GillVerd/status/1764590087927648534?t=wdFaCoZ_6600dZdOmbD56w&s=19 https://arxiv.org/abs/1912.01937 https://arxiv.org/abs/2402.10416 The fact that things exist and happen in the ways they do is just absolutely bizzare lmao what All generated images from phone and PC Galactic Architects comic https://twitter.com/burny_tech/status/1764609972732653625?t=eXpthXIGFKDt6IsgotnpGQ&s=19 I'm familiar with lots of this work but I dont see how is creativity related to it or to free will (whatever definition is used for it). Do you mean uncomputability of kologomov complexity? Or of exact bayesian inference? Both brains and neural nets approximate that in computable way. Or brains still work even because of halting problem. It seems to me that human brains (and other sources of other forms of cognition in organisms) are creative and they're a finite computational physical systems that can be theoretically replicated by engineering it, as one can see in Michael Levin's work for example. https://twitter.com/burny_tech/status/1764609972732653625 https://arxiv.org/abs/2403.00745 https://twitter.com/fly51fly/status/1764618330680574401?t=l6UZ0CCB004qntyqhY_FrA&s=19 "In our new work, we find that transformers develop in-context learning in discrete stages. Not only are these stages behaviorally and structurally unique, but they can be discovered in a model- and data-agnostic way by probing the geometry of the loss landscape, both in parameter space and in function space. Developmental interpretability is real." Paper: https://arxiv.org/abs/2402.02364 Twitter thread: https://x.com/jesse_hoogland/status/1755679791943147738 One reason why mathematics enjoys special esteem, above all other sciences, is that its laws are absolutely certain and indisputable, while those of other sciences are to some extent debatable and in constant danger of being overthrown by newly discovered facts. [Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses | The BMJ](https://www.bmj.com/content/384/bmj-2023-077310) [Nova classification - Wikipedia](https://en.wikipedia.org/wiki/Nova_classification)