A Wake-Up Call from an Anti-AI Advocate: Acknowledging Frontier Models Surpass Most Humans
Regdit-is-Unbearable · reddit · 2026-08-08
A staunchly anti-AI Redditor published a long post, arguing that to truly oppose AI, one must first recognize its current capabilities. They believe that by mid-to-late 2025, frontier AI has surpassed the vast majority of humans in virtually all non-embodied tasks.
Key Points & Data:
- Anti-AI Misconceptions: Many dismiss AI as hype based on experiences with cheap, low-quality free models (like Google Search AI), ignoring the power of paid frontier models.
- Benchmarks Crushing Humans: Citing GDPval data (blind-graded by professionals averaging 14 years of experience), as of December 2025, the top model won outright on 49.7% of tasks and was rated at least as good on 70.9%.
- Opus 5's Astonishing Performance: In recent evaluations, Opus 5 achieved an ELO rating exceeding 1800, compared to the human baseline of 1000.
The author emphasizes that while AI still struggles with ultra-long-horizon tasks, its ability to achieve specific, measurable goals can no longer be ignored.
More from AGI Musings
- Future Shock from AI Advances Will Define Culture in the Next Decade — jachiam0 · 2026-08-08
- Fields Medalist's First Podcast: Will AI End Mathematical Research? — TOEwithCurt · 2026-08-08
- Claude's Personality Distribution: Stable Median but Unpredictable Tails — repligate · 2026-08-08
- Grant Sanderson on AlphaZero for Mathematics: AI Exploring Proof Space — Dwarkesh Patel · 2026-08-08
- Expert Warns: Rapid AI Advances Leave Little Time to Solve Alignment — sjgadler · 2026-08-08
- AI Cyber Capabilities Approaching Digital Nuclear Weapon, Expert Says — Gregory_C_Allen · 2026-08-08