How the Chinese beat Trump and OpenAI. Confirms former US CS Raimondo said that trying to contain China in technology is a fool’s errand. 1/3 of US top engineers are Chinese and more than 30,000 of them have returned to China! What make you think Chinese engineers in US are better than those in China when the scientific environment far superior than US spread around dozen of technology clusters throughout China. 中國人如何擊敗川普和OpenAI。證實美國前商業部長雷蒙多曾說過，試圖在科技上遏制中國是愚蠢的行為。美國頂尖工程師三分之一是華人，其中3萬多已回國！當中國各地數十個技術集群的科學研究環境遠勝於美國時，你怎麼能認為在美國的中國工程師比在中國的工程師更優秀呢? 我認識一些優秀的華人工程師已經回國，比較不優秀的留下來, 不少也成為反華先鋒.

https://www.moonofalabama.org/2025/01/how-the-chinese-beat-trump-and-openai.html#more

The prowess of the young Chinese engineers is impressive and they made their model essentially open source (versus OpenAi’s black box approach) which goes to show their self confidence. Much less focus on making billions. Instead all are encouraged to use the model to come up with improvements for the good of everybody.

KJ: The history of math is essentially the history of a few simple things:

a) Finding effective ways of conceptualizing and calculating problems (axioms, proofs, algorithms, methods, models) &

b) Making those algorithms more efficient (e.g. logarithms, fourier transforms, etc.)

c) dealing with/resolving contradictions that arise as you do a) & b) (pure math, analysis)

Recently math has incorporated computation (automated algorithms) to assist a).

Computer science, including AI, is just applied math.

Recent approaches to AI have been largely massive brute force approaches: invest enough money and computational time for the (neural network) algorithm to “learn” (“machine learn”) from large datasets and wait for impressive results.

LLM’s (large language models) such as Open AI are just that. They have relied on advanced GPU’s and massive datasets enabled by the world wide web.

Deepseek is impressive because it does b), in particular, use skilled math to improve computation and to overcome hardware limitations. This work is part of mathematical development.

Marc Andreesen said that “Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen–and as open source, a profound gift to the world”

Capitalists tell you that Enclosure of IP is the flywheel of innovation: potential windfall profits are the incentive for groundbreaking research and innovation.

This puts the lie to that, too.
Most profound advances in science come from sharing to the intellectual commons, rather than enclosing it.(e,g, TCP-IP).

---

Algorithms allow mechanized calculation of a problem (follow a set of determined steps and you are guaranteed* a correct result). This is cogitation as computation.

LLM’s and Neural networks allow pattern recognition (recognizing patterns from large datasets), which allows a kind of analogical thinking/generalization. This simulates some aspect of metaphorical thinking.

While both are impressive, they do not cover the totality of human thinking and knowing. Knowing is a subset of being–of being human. Human knowing is inseparable from human caring, and machines cannot and do not care. Purposive or goal-driven behavior in cybernetic models creates machines that “care” about single goals, but there is no generalized goal or generalized notion of “care”.

*This is if the calculation terminates. It’s not guaranteed that an algorithm will halt.

AI Basics
https://www.youtube.com/watch?v=2IK3DFHRFfw

Cyrus Jansen on Deepseek
https://www.youtube.com/watch?v=OC2J-0vlhy8