Agora-1: The Multi-Agent World Model
AI: Not Conscious, but UNPREDICTABLE!
AI: Not Conscious, but UNPREDICTABLE makes for a dangerous cocktail!
AI is obviously NOT conscious. To be con-scious one needs something ELSE to go with the science... At least etymologically speaking. But the science of AI fully explains AI, there is no need for anything else to go with the knowledge. Present AI, Large Language Models, LLM, all about finding gradients in extremely high dimensional spaces
Another type of AI is possible: WMs: World Models, where the AI basically learns physics. WMs will be superior to LLMs for most applications beyond Internet searches.
Consciousness requires the feeling of existing: present day AI do not have this capability, all the more because we do not know what it consists of.
The smallest known structures in the brain are Dendritic Nano Tubes, DNTs, and they are of the order of 100 nanometers across… 1000 atoms wide (molecules causing Alzheimer squeeze through them). At a scale ten times smaller, Quantum effects will appear. It is likely that they will play a crucial role in consciousness. Why? Because at this sort of scale, Quantum effects are going to appear, first of all (there are libraries of geometries and the Casimir Effect they bring… At the ten nano scale, so not only will there be Quantum effects, but Quantum Field Theory effects, and the dynamics of vacuum… Turns out the soul’s material is vacuum energy… Something like that is looming… It’s curious that QFT is not vibrating more in the collective consciousness… Obviously a problem with the High Energy physicists’ lack of elocution or imagination…)
Present day AI rests on electronics which work like canal systems (canals carrying electrons, not water molecules)… Quantum effects caused by said electrons are deliberately obliterated so that “semiconductors” can behave like classical canal systems.
The Quantum Computer, QC, is the exact opposite: far from avoiding Quantum effects, they are maximally exploited to compute with. The QC rests on entanglement, nonlocality and a new property called “magic” (basically what classical entanglement can’t duplicate)..
If Twenty-first Century technology can do it, it would be naive to believe that evolution did not get there first, especially since Quantum effects have been shown to play a direct role in neurology already (for example in birds seeing the Earth magnetic field).
Thus, if we want consciousness, we need the QC.
QC will enable us to create AC, Artificial Consciousness.
***
A different question is whether present AI can SIMULATE consciousness. The answer is obviously yes. Confusing this simulation with a creature with (Quantum) consciousness is the AI DELUSION. Dawkins who wrote a book called the God Delusion is himself suffering from an arguably worse condition, the AI Delusion. It’s worse because AI as we have it, is just classical canal engineering and smart high dimension calculus… There is no mystery whatsoever. .
***
Even though the AI we have now are not conscious, and can’t be conscious, they are still unpredictable. I actually have a mathematical PROOF of this (using mathematics and logics from the 1870-1940 period…).
That makes uncontrolled AI a potential master of humanity.
Lest we be very smart and careful…
Patrice Ayme
Coming soon to a neighborhood near you… Robot contemplates Valles Marineris on the equator of Mars, 10 kilometers deep… The colonization of Mars will be initially driven by AI… Next NASA Mars mission will be propelled by a nuclear engine, and three AI helicopters will fly away with supersonic blades before hitting the ground… And communicate with Earth through overhead satellites…
#AI #ArtificialConsciousness #ArtificialIntelligence #Consciousness #llm #Philosophy #QuantumComputer #RichardDawkins #spirituality #Unpredictable #WorldModel
My next read: https://le-wm.github.io/
Actually, I might start lazy and ask an AI to summarize it for me first ;-)
Most self-driving systems rely on hand-coded perception outputs and engineered driving rules. Learning directly from human driving data with an end-to-end method can allow for a training architecture that is simpler and scales well with compute and data. In this work, we propose an end-to-end training architecture that uses real driving data to train a driving policy in an on-policy simulator. We show two different methods of simulation, one with reprojective simulation and one with a learned world model. We show that both methods can be used to train a policy that learns driving behavior without any hand-coded driving rules. We evaluate the performance of these policies in a closed-loop simulation and when deployed in a real-world advanced driver-assistance system.
#Sora was always a physics engine in disguise. The stated goal from day one: teach AI how the material world works by ingesting video at scale. How objects persist, collide, deform, fail. The consumer deepfake playground was a distraction. The robot training data was the mission.
The app is now dead. $2M lifetime revenue against a $730B valuation.
OpenAI; "The team is "focusing on world simulation research to advance robotics.""
Worth flagging: independent research from ByteDance and Tsinghua found Sora never actually learned physics. It learned surface patterns, prioritising colour over shape over motion.
Yann LeCun called pixel-prediction-as-world-model "doomed to failure" back in 2024. OpenAI's sudden pivot suggests he may have had a point.
The TikTok wrapper is dead. The bet that video is the richest training signal for grounding AI in physical reality is still very much alive. Just without the Mario deepfakes.
⬆️ >> #AI got the blame for #Iran school bombing…
Excellent example of how a #languageModel is NOT the same as #worldModel or #realTime #realWorld data.
The #Maven system that #Palantir embedded into the #US military infrastructure relies on BOTH #LLM and #realTime #realWorld data, but it cannot prevent catastrophes when there is a failure in either or both of them.
In this case, it was #staleData at the very least— possibly a faulty/imprecise language model as well.
Soo. I think that the #worldModel #ontology stuff is about as close as the world is going to get to saying #lambdaMOO #programming people
are and have always been completely right about the #future, #technology, reality and #vr and living in, reasoning about, building and extending this reality.
@me @nosrednayduj @spidercat @pizzapal @mdhughes @schroeder @kentpitman at at at.
If you will forgive the example, Lecun quitting Facebook to work on World Models instead is where we are.
One of the more in-depth talk about LLM's problem and world model (JEPA architecture by Yann LeCun
Laurent Cheylus
Hacker News
Patrice Ayme's Thoughts
Stefan Köhler
cryptax
Wulfy—Speaker to the machines
Ramesh Gupta
screwlisp
樹雄 🌳🐨 