A good theory of mathematical beauty is more practical than any current observation, as new predictions about physical reality can be self-consistently verified. This belief applies to the current status of understanding deep neural networks including large language models and even the biological intelligence. Toy models provide a metaphor of physical reality, allowing mathematically formulating the reality (i.e., the so-called theory), which can be updated as more conjectures are justified or refuted. One does not need to present all details in a model, but rather, more abstract models are constructed, as complex systems such as the brains or deep networks have many sloppy dimensions but much less stiff dimensions that strongly impact macroscopic observables. This type of bottom-up mechanistic modeling is still promising in the modern era of understanding the natural or artificial intelligence. Here, we shed light on eight challenges in developing theory of intelligence following this theoretical paradigm. Theses challenges are representation learning, generalization, adversarial robustness, continual learning, causal learning, internal model of the brain, next-token prediction, and the mechanics of subjective experience.

Haiping Huang

Frontiers in Computational Neuroscience

FRONTIERS IN COMPUTATIONAL NEUROSCIENCE 是一本一流的电子期刊，致力于推广大脑功能的理论建模，促进理论和实验神经科学之间的跨学科互动。对大脑惊人能力的理解进展仍然有限，我们相信只有深入的理论思考和不同学科和方法之间的相互促进合作才能实现。因此，我们邀请对广泛主题的原创贡献，以展示这种合作的成果，或为未来的联盟提供刺激。我们的目标是为神经系统的前沿理论研究提供一个互动论坛，并向更广泛的神经科学界传播最好的理论研究。从生物物理驱动的神经元和突触的逼真模拟到推理和决策的高级抽象模型，各种风格和所有级别的模型都受到欢迎。虽然该杂志主要侧重于基于理论和驱动的研究，但我们欢迎验证和测试理论结论的实验研究。另外：COMP NEURO

Frontiers in Computational Neuroscience is a leading electronic journal dedicated to advancing theoretical modeling of brain functions and fostering interdisciplinary interactions between theoretical and experimental neuroscience. Our understanding of the brain's remarkable capabilities remains limited, and we believe that only through deep theoretical reflection and collaborative efforts across various disciplines and methods can we achieve breakthroughs. Therefore, we invite original contributions on a wide range of topics to showcase the results of such collaboration or to inspire future alliances. Our goal is to provide an interactive forum for cutting-edge theoretical research on the nervous system and to disseminate the best theoretical studies to the broader neuroscience community. From biophysically driven realistic simulations of neurons and synapses to advanced abstract models of reasoning and decision-making, a variety of styles and all levels of models are welcome. While the journal primarily focuses on theory-driven research, we also welcome experimental studies that validate and test theoretical conclusions.

《Frontiers in Computational Neuroscience》是一本顶尖的电子期刊，致力于推动大脑功能的理论建模，促进理论与实验神经科学之间的跨学科互动。尽管我们对大脑的惊人能力的理解仍然有限，但我们相信，只有通过深入的理论思考和不同学科及方法之间的相互促进合作，才能取得突破。因此，我们邀请对广泛主题的原创贡献，以展示这种合作的成果，或为未来的联盟提供启发。我们的目标是为神经系统的前沿理论研究提供一个互动论坛，并向更广泛的神经科学界传播最优秀的理论研究。从生物物理驱动的神经元和突触的逼真模拟，到推理和决策的高级抽象模型，各种风格和所有级别的模型都受到欢迎。虽然该期刊主要侧重于基于理论和驱动的研究，但我们也欢迎验证和测试理论结论的实验研究。

Category	Quartile
医学	4
医学, 数学与计算生物学	3
医学, 神经科学	4

Eight challenges in developing theory of intelligence

ivySCI AI Smartly Parses PDF, Answers Researchers' Questions, and Helps You Understand Papers in Seconds

Journal Info