Is the most intelligent use of the human brain single-threaded or multi-threaded?
Discussion
What if we have a conscious thread and a sub-concious thread. I find coding solutions come to me when not thinking about them
We all experience this situation. I think this is especially like the emergence of intelligence after a large amount of data and computing power are superimposed on large language models. Maybe in the case of a large amount of daily coding training, the connections between your brain's coding neurons become more and more numerous. However, in the current neural network where conscious single-threaded control exists, there is no way to better utilize the powerful capabilities of the neural network. And people always have flashes of inspiration when they are taking a bath or emptying their brains. This situation should be that the existing neurons are self-connected and there is an emergence of intelligence. It's like floating-point operations between tokens in a large language model.
Yoga meditation and Zen meditation both allow people to let go of themselves, just as calming the water can make it clear and transparent so that you can see the fish at the bottom of the lake, using your brain's wisdom.
Modern people are constantly exposed to various information due to the emergence of smartphones anytime and anywhere. We should put down our phones and give our brains time to unwind. Only an empty space can accommodate new things.
'Empty'
#brain #neuralnetwork #meditation #zen
There is no subconscious; what is referred to as the subconscious is similar to dreaming. It consists of random events produced by the free connection of neurons in different functional areas of the brain when it is not under conscious control.
智能是神经元对环境的反应规范，并由多巴胺、内啡肽和血清素等一系列神经递质和激素综合作用的结果。
机器智能是否应有情绪体验？
智慧生命的情绪体验和非对称信息动态博弈混合策略纳什均衡是什么关系？
大模型是否能模拟出多巴胺、内啡肽等化学物质的作用。
换言之，我们需要问大模型出现智能的前提条件是它的自然激励机制是什么？
更准确的来说，没有人的参与，大模型能不能仅靠追逐自身利益的自然动机所驱动而诞生出智能。
或者说，我们能不能将闪电网络L402协议和cashu等集合到大模型中去，
让它自行驱动并追寻自身利益过程中涌现出机器智能。
人类婴儿学习过程中并不需要像大模型那样输入过量庞杂的信息， 人类婴儿的智能是不需要辅助的仅由少量信息输入即可完成惊人的进步。
很显然，大脑神经元对大脑皮层环境的反应规范是无比高效的，几乎没有信息冗余。
而这一切依赖于ATP能量货币在神经元连接处作为信号分子的信息传递和能量供应的高效作用。
真正需要我们问的问题是，类似于生命世界，大模型的ATP能量货币是什么？
在此基础上，构建机器智能的内啡肽、多巴胺和血清素等神经递质和激素的类似物可能是什么？
我们能否通过带有ATP能量货币所驱动的计算机程序自然地实现内啡肽、多巴胺和血清素等功能。
假如你认为智能不是设计出来的，而是对环境的一种适应度，并通过能量货币降低系统内信息熵的方式实现智能的一种对环境的反应规范，
那么如今大模型通过亿级语料或对图像知识抽取，并堆叠和调整数千万亿参数来优化学习路径的研究方向是完全错误的。
大模型所谓的自监督学习和对无标签数据的预训练，实际上仍然包含了人的意图和价值观取向，并不是一个信息完备的自组织形式系统。
缺失能量货币的引入，大模型对大规模数据训练的每一步操作仍然属于伪随机事件或布朗运动。
大模型的图灵完备并非是在人的意志作用下完成的，而是符合能量最低原则和信息熵理论的底层物理规则下自然涌现出来的结果。
或许我们需要重新思考能量货币在机器智能的作用，思考人类花费十几年时间耗费巨大能量挖比特币的意义和价值。