Raw LLM Responses

This is my view as well. LLMs are incredible in their own right at certain tasks, but the next generation of architectures that implement online learning in ML models are going to be more human-like in capabilities and consistency: See [Hierarchical Reasoning Model](https://arxiv.org/pdf/2506.21734) [Energy Based Model](https://alexiglad.github.io/blog/2025/ebt/) [Test-time Training](https://test-time-training.github.io/) Yann LeCunn’s JEPA and Francois Chollet’s program synthesis could also be potentially promising paths towards AGI. I personally put a 50% chance of AGI in the next 10 years, 70% in the next 20 years and 90% in the next 30.