DNA to AI: How Evolution Shapes Smarter Algorithms - neurosciencenews.com/genomic-ai-neu… via Neuroscience News

How close is AI to human-level intelligence? nature.com/articles/d4158…

Dendrites are all you need nature.com/articles/s4146…

🏆🏅This New Method: Analogy-Augmented Generation (AAG) Mimics Human Problem-Solving with Analogical Reasoning, Tested on LangChain Tutorials and Outperforming RAG by 40% Large Language Models (LLMs) excel in language understanding but often struggle to synthesize complex,

Analogy-Augmented Generation (AAG) enhances large language models’ procedural task performance by retrieving and adapting past experiences from a procedural memory store, outperforming traditional methods in various datasets.

Having a convo w/4o about the origin of life and it dropped this nugget: Emergence of life aligns with patterns seen in algorithmic evolution: random generation followed by iterative selection. Information might be the “prime mover”, with biochemistry serving as its substrate.

thetransmitter.org/neuroai/neuroa…

If Elon Musk adds a super structure to sit on top of Grok and Twitter to somehow combine the two, could that lead to synthetic sentience? We need not just tokenization of language but also tokenization of modeling (to capture ambient context) - none better than the human mind.

Knowledge without understanding has no agency.

Regarding forest fires, can anyone verify this statement from NYS Dept of Environmental Conservation? “Large dead standing trees, slash on the ground don’t drive the fire. In fact, when the fire hits that, it stops.”

rdcu.be/eh75F

Bird evo per gpt: “Flight is not merely locomotion—it is graceful falling shaped by controlled descent. The “drive” toward flight is not a goal but a harmonization of conflicting pressures: evading predators, accessing new niches, conserving energy, and enhancing mating success.”

My guess- true AGI will somehow and some way reflect the structure of DNA - nothing else seems to have beaten entropy.

Can we design a LLM or hybrid model that treats entropy as an affordance- a gateway to learn, explore, and possibly feel- rather than a failure to predict?

One side of the helix to represent the forward pass, the probability of the next token in a sequence. The other to represent the backward pass- of what’s missing and of what could be.

LLMs are pattern recognizers. Using entropy, negative affordances, and analogical replay could lead closer to... AGI may emerge when agents track informational tension (entropy), retrieve historical analogs (memory), and resolve conflicts with creative replay (wisdom). -ChatGPT

ChatGPT just dropped this nugget on me: The universe is increasingly legible as a grammar of interrelated forms, information preceeds matter and energy.

AGI will emerge not from scale, but from synthesis—when systems can reinterpret the past to resolve present tension. ChatGPT

How evolution and environment confer flight to birds, 2% at a time with “local” improvements (hollow bones, feathers, navigation, etc.) followed by convergence of these adaptations, visualized. ChatGPT