Game over. AGI is not imminent, and LLMs are not the royal road to getting there
a day ago
- #LLMs
- #AI Critique
- #AGI
- Dreams of LLMs leading to AGI have collapsed.
- Apple's reasoning paper confirmed LLMs can't solve distribution shift, a core weakness in neural networks.
- GPT-5's release was delayed and underwhelming.
- Rich Sutton, a Turing Award winner, supported critiques of LLMs.
- Andrej Karpathy stated agents are far from AGI, estimating it's a decade away.
- Demis Hassabis debunked overhyped claims about LLMs' math capabilities.
- Gary Marcus's 2020 article 'Next Decade in AI' appears increasingly prescient.
- The consensus is that current LLMs are not close to AGI.