Building the AI of tomorrow

• SETI @ Home
• God's Debris - Scott Adams
• "The Network is the Computer" - John Gage for Sun Microsystems in 1984

Numerous online articles proclaim the shortcomings of Large Language Models (LLMs) and the budgetary and resource constraints hindering our progress toward Artificial General Intelligence (AGI). While contemplating this, I began to consider how the human mind selectively retrieves information and activates specific brain regions based on sensory neural triggers. This led me to the realization that our minds activate only the necessary regions where long-term memories reside, but only after our mind and nervous system rapidly categorize and assimilate data in conjunction with multiple sensory inputs and memories. This assimilation process involves both hemispheres of the brain, which are demonstrably asymmetric. Studies such as "Brain activation in response to taste stimuli," "Brain activation related to the sensory modality," and "SH's brain activity during visual imagery" further illustrate this phenomenon.

These observations prompted me to explore the concepts of experiential context and its influence on bias and truth perception, drawing from my own experiences and readings.

1. Addressing the Contextual Activation Problem: I posit that no single model will universally solve this challenge. Regardless of advancements in data center capacity or physical storage, the perception of reality remains a complex influx of sensory input requiring interpretation based on individual and potentially nearby observations. This mirrors the challenges faced by radio telescope data collection and the subsequent development of distributed computing solutions like SETI@home. Similarly, managing the memory requirements for contextualization necessitates shifting the workload to individual devices. This involves securely storing multiple contextual scopes on consumer devices in numerical formats with edge-based agents, enabling long-term context storage without relying on continuous cloud storage. This approach allows for data passivation and activation to become a function of the network and the edge device. While the majority of computational requirements will remain centralized, the potential for agentic processing on the device emerges, reducing cloud computational costs and redundant token exchanges.
2. Embracing Distribution for AGI Development: Analogous to the hypothesis in "God's Debris," where mankind results from God's self-annihilation to create the universe, our AI strategy must embrace distribution to establish the foundational laws of its own physics. Mirroring the brain's structure, which relies on multiple hemispheres to hold the structures of truth, AGI will require the activation of multiple segments within a vast network to effectively perceive and communicate truth. This necessitates a novel protocol capable of interrogating and sharing signals from various sources (cellular, network, news, camera, etc.) in near real-time. By summing previously known trusted states, this protocol can derive the current status, preventing misinterpretations, such as mistaking fireworks for an explosion.
3. The Network as Intelligence: John Gage's statement, "The Network is the Computer," requires an amendment. We must recognize that "The network is the intelligence," as intelligence transcends lexical comprehension and next-sentence prediction. Our primary objective should be the unwavering pursuit of truthful observation, mitigating the inherent bias present in online written content. This necessitates a more secure IoT standard and standardized signal interpretation. Achieving this is possible, but significant changes are required before realization.

Conclusion: Only a select few companies possess the capacity to effectively and safely compete in this domain. We must transcend single-source contextualization and adopt Nature's approach. Sensory interpretation triggers multiple parts of the human mind, creating a consensus of reality based on past experiences and current circumstances. Distributing the workload to underutilized devices and extending the Cloud's context to create a more realistic observable platform offers the fastest and safest path forward. While the origin of consciousness remains unknown, ensuring safe execution by considering truth as a multi-faceted concept based on facts, not editorial interpretations, is our responsibility.

#AI #GenerativeAI #LLM #Google #GoogleCloud #Future #Futurist

References

(1) Kishi, Mayo & Sadachi, Hidetoshi & Nakamura, Junji & Tonoike, Mitsuo. (2017). Functional magnetic resonance imaging investigation of brain regions associated with astringency. Neuroscience research. 122. 10.1016/j.neures.2017.03.009.

(2) Multisensory Integration of Sounds and Vibrotactile Stimuli in Processing Streams for "What" and "Where" - Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/Brain-activation-related-to-the-sensory-modality-Auditory-in-red-and-tactile-in_fig3_26785978

(3) Visual imagery and functional connectivity in blindness: a single-case study - Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/SHs-brain-activity-during-visual-imagery-Activation-related-to-all-visual-imagery_fig2_272513065