There are bottlenecks, but It's not that improvement has stopped on training current AI models, if you read papers like this, "The Surprising Effectiveness of Test-Time Training for Abstract Reasoning"

https://arxiv.org/abs/2411.07279?utm_source=superhuman&utm_medium=newsletter&utm_campaign=chatgpt-meets-coding-apps&_bhlid=3421d13da369891acfd83127455653c0c20c8f24

#AI #LLM #reasoning