The switch from cybernetics (control theory) to computation (automata theory) was profound. Semiotics is mostly useless wordcel stuff and did not substantially contribute to CS.
RESEARCH
-

Nine Algorithms That Changed Computing History
By
–
Nine Algorithms That Changed the Future — The Ingenious Ideas That Drive Today's Computers: https://
amzn.to/41kXVnW -
AI Model Reasoning Traces Transparency and Steering Comparison
By
–
Yes, you can. Unfortunately, OpenAI doesn’t make their real reasoning traces available, so it’s really hard to know if you’re doing it right with their models. Other models are different, you can see the actual reasoning so you can tell if it’s being steered properly!
-
Router needs 4-5x more parameter capacity
By
–
u get it there needs to be probably 4-5x more ooms of params in the router too
-
Academia innovates, industry commercializes and claims success
By
–
Academic research takes you to the tipping point. Industry takes the last step and gets all the credit.
-
AI Growth Exponential: Industry Veteran Reflects on Rapid Expansion
By
–
I've been in AI since the 90s, and every few years I ask myself, "How much bigger could this get?" And then it gets way bigger.
-

local-deepthink: AI Agents Collaborating as Neural Network Neurons
By
–
local-deepthink: Democratizing Deep Thought local-deepthink ia a system that runs a novel algorithm called Qualitative Neural Network (QNN). In a QNN, different AI agents are treated like "neurons" that collaborate and critique each other to refine ideas, trading slower
-

GPT-5 Energy Consumption: 18 Wh Per Response Analysis
By
–
GPT-5 consommerait 18 Wh par réponse. Huit fois plus que GPT-4.
Et bien loin des 0,34 Wh annoncés par OpenAI, puisque GPT-4 atteindrait en réalité 2,12 Wh selon l’étude de l’Université de Rhode Island. Au delà du chiffre, et de sa méthode de calcul par hypothèse, c'est une -
Efficiency improvements keeping pace with model demand and quality
By
–
Yes, I think the efficiency improvements are doing a pretty good job of keeping up with demand (and what people care about is not necessarily bigger models, but better models).