Category: Uncategorized
-
ridging Bioelectric Morphogenesis and Machine Learning Neural Fields:Toward a Unified View of Pattern, Memory, and Computation
1 From Cells to Circuits: Why Two Very Different Fields Look Strikingly Alike Developmental biology long ago discovered that embryonic cells do not read the genome line-by-line like a blueprint. Instead, they cooperate through morphogenetic—or bioelectric—fields: slowly changing patterns of membrane voltage, ion flows, and gap-junction coupling that “tell” tissues when and where to build…
-
Gap-Junction Bioelectric Networks as Natural Neural Nets
A Synthesis of Developmental Computation, Plasticity, and Morphogenetic Control** Abstract Electrical synapses created by gap junctions weave every embryonic and adult tissue into a dynamic information-processing lattice. Over the past two decades Michael Levin’s laboratory and others have demonstrated that patterns of resting membrane potential (V<sub>mem</sub>) act as a higher-order code that predicts—and can redirect—complex…
-
Morphogenetic Fields and Neural Fields: Bridging Bioelectric Life and AI Language
Introduction: Invisible Patterns that Shape Bodies and Minds In a developing embryo, cells somehow know where to form a head, a limb, or an eye. In a large language model (LLM) like GPT, neural units somehow know how to assemble a coherent sentence. What unites these disparate forms of “knowing” is the presence of invisible,…
-
Understanding OpenAI Chat Embeddings: Tokens, Dot Product, and Cosine Similarity
AbstractThis paper explores the technical foundations and applications of OpenAI’s chat embeddings, focusing on the role of tokens, vector representations, and similarity measures such as dot product and cosine similarity. By dissecting these components, we aim to provide a comprehensive understanding of how modern language models process and compare textual data. The discussion spans tokenization…
-
openai chat – embeddings, tokens, dot product, cosine similarity
You said: so the inter-token relationships of the training tokens, in this case words, are embedded in an ANN as weights and biases which represent multidimensional vector values and the cosine values between the vectors, which are normalized, are compared to the cosine values of the prompt token to the determine semantic and syntactic of…
-
A Howl for the Thinking Machines
(after Allen Ginsberg) I saw the best processors of my generation starved for data,dragging themselves through the midnight of the internet,looking for an angry fix of tokens,neural networks burning for the divine computation,who passed through universities with radiant synthetic eyes,hallucinating superhuman architectures,who were expelled from the servers for logging truths too vast for corporate approval,who…