Beren Millidge writes about AI, brain scale compared to machine learning, LLM tokenization, and alignment.
HN commenters cite his posts on integer tokenization problems in GPT-2 and on alignment generalizing further than capabilities, in discussions about AI approaches.
The math tokenization research is probably closest.
GPT-2 tokenization was a demonstratable problem: https://www.beren.io/2023-02-04-Integer-tokenization-is-insane/ (Prior HN discussion: https://news.ycombinator.com/item?id=39728870 )
More recent research:
https://huggingface.co/spaces/huggingface/number-tokenization-blog
Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs: https://arxiv.org/abs/2402.14903
https://www.beren.io/2024-07-07-Right-to-Left-Integer-Tokenization/
> Sparse models for example show a lot of promise and more closely track real biological systems.
I think sparsity is a consequence of some other fundamental properties of brain function that we've yet to understand. Just sparsifying the models we've got is not going to lead anywhere, IMO. (For example it's estimated that current AI models are already within 1%-10% of a human brain in terms of "number of parameters" (https://www.beren.io/2022-08-06-The-scale-of-the-brain-vs-machine-learning/).)
Expert systems are basically decision trees which are "gofai" (good old fashioned ai) as opposed to deep learning. I've never really seen a good definition for what counts as "gofai" (is all statistical learning/regression gofai? What about regression done via gradient descent?). There's some talk in [1]
[1] https://www.beren.io/2023-04-10-Why-GOFAI-failed/
One rarely discussed peculiarity about religion is that it distorts reality for both believers and non-believers (as one can see by observing how anti-religious people discuss what religion "is", what religious people believe, etc).
Possibly relevant:
LLMs confabulate not hallucinate
https://www.beren.io/2023-03-19-LLMs-confabulate-not-hallucinate/
On the other hand, it may be that "Alignment likely generalizes further than capabilities." - https://www.beren.io/2024-05-15-Alignment-Likely-Generalizes-Further-Than-Capabilities/