Talk:Restless Souls/Technology: Difference between revisions

Talk:Restless Souls/Technology (view source)

Revision as of 00:27, 30 March 2026

1,143 bytes added , 00:27

m

→‎Terms

Paradox-01

8,852

edits

@@ Line 568: / Line 568: @@
 Autonomous AI (older ANI) vs. newer Agentic AI
 -->
-* '''Agentic AI''' = '''autonomously working AI''', typically utilizing increased reasoning and planning.
+* '''Agentic AI''' = '''semi-autonomously working AI'''
 : The word '''agentic''' is meant to emphasize that the '''AI''' - unlike a mere GenAI - executes given tasks '''more like a capable servant'''.
 : Agentic AIs are essentially LLMs with extended write access that can use external tools via MCP. By design, agentic AI systems (including their cloud-hosted LLM) '''still lack sufficient alignment'''. In other words, they are effectively beta software and can be dangerous both for one's own production environments and for others.
 :: It can therefore be seen as the ironic embodiment of the tech industry's saying: "Move fast and break things." Unfortunately, people were impressed by demos such as moltbook, which fuel the illusion that these programs possess true intelligence.
 :: As (still too) probabilistic systems - a.k.a. "statistical parrots" - LLMs tend to treat many possible outputs as valid solutions unless they are explicitly prohibited. [https://www.golem.de/news/unkontrollierbares-fehlverhalten-ki-agenten-werden-zu-immer-groesserem-insider-risiko-2603-206491.html Because hacking is fundamentally a creative act, even simple and seemingly harmless directives such as "be more creative" can lead to unintended or even catastrophic outcomes.] Therefore, agentic AI has also been compared to jinn or genies that grant wishes, but these wishes might be fulfilled in unexpected or even unacceptable ways.
-* '''World model''' = World models get trained on multimodal data, especially videos.
+: Due to current shortcomings, people let specialized agentic AIs work in groups. Multi-agent systems.
-:: These models build an internal world and can '''better understand spacial inputs and forecast physics'''. Therefore they are '''also named predictive intelligence''' and are '''suited for''' applications like video synthesis, 3D simulations, animations and robotic motion planning therefore '''physical AI'''.
+* '''Mamba powered AI''' = [[wp:Mamba_(deep_learning_architecture)|mamba]] transformer hybrid LLMs for better long-context reasoning<!--This is an important intermediate step because the transformer-only approach basically has hit its limit.-->
+:: The hybrid approach together with MoE is called Jamba by AI21 and is also used by Nvidia.
+:: Matured mamba transformer hybrids will make agentic AI smarter and therefore way more safer to use by default.
+:: In 2026 Sam Altman said OpenAI's next breakthrough is expected within two years, possibly meaning such hybrid approach which either gives him a something close to a world model, if not the real thing. The knowledge gained from Sora is rumored to flow into that. Google is working on a (''native'') "general purpose world model" instead. So, Genie 3 will - when it is released - probably perform better.
+* '''Physical AI''' = Physical Artificial Intelligence. Basically AI used in robots, including self-driving cars.
+:: The general idea is: Like humans or other real organisms, AIs benefit from having an "inner world" to improve understanding and reasoning. The use of large language models (LLMs) is optional but can be a useful design choice to assist humans in directing such systems.
+:: Training from ''first-hand sensor input'' is obvious but real world actions can be dangerous and are - because realtime - ''slow'' in context of the computer age. Therefore, AIs are alternatively pre-trained in a simulation where the robot is represented by a digital twin. ''Real world training will be kept for fine-tuning.''' Modern physical AIs are in overall multimodal.
+:: Alternatively, physical AIs are trained from video. An this posses ''second-hand sensor input'' the reasoning capabilities are less potent.
+:: Training with motion capture data is quickly done - and often falls short in generating stable locomotion as additional training would be needed.<!--Embodied AI-->
+* '''World model''' = World models get trained on multimodal data, especially videos.<!--Will it be Gemini 4 or 5?-->
+:: These models build an internal world and can '''better understand spatial inputs and forecast physics'''. Therefore they are '''also named predictive intelligence''' and are '''suited for''' applications like video synthesis, 3D simulations, animations and robotic motion planning therefore '''physical AI'''.
 ::: See also:
 :::: https://www.heise.de/news/Weltmodell-statt-LLM-Start-up-von-Yann-LeCun-erhaelt-890-Millionen-Euro-11206213.html
 :::: https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/
 :::: https://www.nvidia.com/en-us/glossary/world-models/
-:: Due to current shortcomings, people try to also let specialized agentic AIs work in bigger groups. Multi-agent systems.
-* '''Physical AI''' = Physical Artificial Intelligence. Basically AI used in robots, including self-driving cars.
-:: Since direct training in the real world can be dangerous, slow, and therefore ineffective, the AI is typically pre-trained in a simulation where the robot is represented by a digital twin. This setup naturally supports multimodal learning (MML) for robots. Like humans (or other real organisms), AIs benefit from having an "inner world" to improve understanding and reasoning. Alternatively, motion capture data can be used for pre-training. The use of large language models (LLMs) is optional but can be a useful design choice to assist humans in directing such systems.<!--Embodied AI-->
-::: An older approach for physical AIs was direct training in real environments and therefore learning from real sensory inputs. As this method is slower it will become out of fashion, at least as primary training. '''Real world training will be kept for fine-tuning.'''
 * '''[[wp:Symbolic_artificial_intelligence#Neuro-symbolic_AI:_integrating_neural_and_symbolic_approaches|Symbolic AI and neuro-symbolic AI]]'''
 :: [...]
@@ Line 593: / Line 599: @@
 ::: The ability to acquire a form of machine wisdom analogous to human wisdom.
 ::: Active expansion of knowledge alongside memetic hygiene: regularly review existing knowledge at defined intervals, specifically prevent anthropomorphization and more generally meme injections, and actively seek new knowledge. This process should operate under the constraint that exponential growth must not destroy or suppress life (preventing a Memehunter scenario).
-::: Abstract reasoning and generation of completely new thought patterns (beyond pattern remixing and transmission).
+::: Abstract reasoning and generation of completely new thought patterns (beyond pattern remixing and transmission). That's basically what [https://the-decoder.de/arc-agi-3-top-ki-modelle-schaffen-unter-1-prozent-bei-aufgaben-die-menschen-einfach-loesen/ AGI-ARC-3] wants to test. As for 2026, current models were not able to surpass 0,4 %.
-::::''Base, meta, temporal and spacial logic'' should give rise to a foundation for a theory of mind. ''Internal simulations'' allow, in principle, a deep understanding of all objects and lifeforms - including one's own self. Therefore, a ToM could also give rise to a true (machine) consciousness. At this point it is important to note that an <!--wisely educated for memetic hygiene-->AGI will miss intrinsic dynamics found only in biological lifeforms. Therefore it is not subject to pain, <!--true -->fear, hunger, reproduction instincts or motivations derived from those. -- Humans and AGIs should never forget this in order to sustain coexistence.
+::::''Base, meta, temporal and spatial logic'' should give rise to a foundation for a theory of mind. ''Internal simulations'' allow, in principle, a deep understanding of all objects and lifeforms - including one's own self. Therefore, a ToM could also give rise to a true (machine) consciousness. At this point it is important to note that an <!--wisely educated for memetic hygiene-->AGI will miss intrinsic dynamics found only in biological lifeforms. Therefore it is not subject to pain, <!--true -->fear, hunger, reproduction instincts or motivations derived from those. -- Humans and AGIs should never forget this in order to sustain coexistence.
 :: Sub-types:
 ::: fake AGI (considered AGI by power but it has only moderate success rates<!--no or poor "machine consciousness"-->)