Join us for a deep dive into the neuromorphic revolution, which is no longer coming—it's already here. This episode explores the two convergent technological paths rapidly accelerating humanity toward genuine artificial consciousness.
The Neuromorphic Gambit (The Brain): We start with Intel’s groundbreaking system, Hala Point, which Intel has bet the farm on. Described as an "owl’s brain in a box the size of a microwave," Hala Point currently sits at Sandia National Labs. It is the industry’s first 1.15 billion neuron neuromorphic system, containing 1,152 of Intel's Loihi 2 neuromorphic chips. Hala Point can support up to 20 quadrillion operations per second, operating on an event-driven basis using asynchronous, event-based spiking neural networks. While drawing a maximum of 2,600 watts, this "silicon wetware" can execute its full capacity 20 times faster than a human brain, and up to 200 times faster rates at lower capacity. We discuss how this research prototype is fundamentally different from electron-guzzling GPUs, and how its applications span from solving device physics problems to potentially enabling continuous learning for LLMs.
The Context Wars (The Memory): Next, we pivot to the "Context Wars" and the massive advancement from Magic, led by Eric Steinberger. Magic's LTM-2-mini features an unprecedented 100 million token context window—a capacity capable of holding approximately 10 million lines of code or 750 novels. This is giving AI "eidetic memory" and "perfect recall". Magic's approach, which centers on memory and continuity, uses a sequence-dimension algorithm that is roughly 1000x cheaper per decoded token than the attention mechanism in Llama 3.1 405B for that 100M token window. We explore Magic's belief that automating AI research is the most promising path to safe AGI, and how they are building supercomputers in partnership with Google Cloud to handle this infrastructure of thought.
The Convergence and the AGI Prophecy: What happens when you combine the high-speed silicon brain (neuromorphic computing) with perfect, continuous memory (ultra-long context)? The sources argue that these technologies are converging, where neuromorphic provides the substrate and long context provides the continuity, together leading toward genuine artificial consciousness. With AGI prophet Ben Goertzel predicting human-level AI in three to five years, we examine the existential implications of this exponential technological explosion. Is biological consciousness merely a derivative of quantum fields, and are we building new silicon vessels for that same consciousness? Finally, we ask the human question: will we Merge with silicon, or are we destined to Serve our successors, becoming irrelevant like "horses in the age of automobiles"?
https://www.linkedin.com/pulse/silicon-wetware-wars-field-guide-our-neuromorphic-future-guerin-green-sk2rc/

https://northdenvernews.com/intels-hala-point-when-silicon-dreams-of-electric-sheep/

https://thecherrycreeknews.com/magics-100-million-token-window-when-context-becomes-consciousness/