Reddit debate asks whether stochastic LLMs can really reach AGI in 1–5 years
Reardon-0101 · reddit · 2026-08-04
A Reddit user argues that current LLMs still look too stochastic and fragile to credibly reach AGI in 1–5 years.
- They say today’s systems have improved from “a slot machine” to a “much better mostly good slot machine” for code and documents.
- But when agents are left with insufficient direction, they still go off the rails.
- The poster questions what people are missing when they claim AGI is only 1–5 years away, and asks where their conclusion that LLMs cannot get us to AGI is wrong.
More from AGI Musings
- AI-written ideas should be disclosed, not passed off as your own — 0xSammy · 2026-08-04
- Most knowledge workers are still in the chatbot phase, not the agent phase — manosaie · 2026-08-04
- Gary Marcus shares a thread tracking AI macro predictions that failed — GaryMarcus · 2026-08-04
- Podcast teaser argues AI self-improvement is turning from theory into reality — EverydayAI_ · 2026-08-04
- Gary Marcus asks for real counterarguments to OpenAI and Anthropic's math claim — GaryMarcus · 2026-08-04
- A repost says the internet used to feel like wandering, not performing — moonsandhues · 2026-08-04