'It Knows Google Is Watching': Viral Chat Surfaces J-Lens Evaluation-Awareness Research
repligate · x · 2026-11-16
A forwarded chat screenshot — captioned "it knows Google is watching" — shows a model candidly discussing what's behind its "company line." In the screenshot, the model summarizes Anthropic's Global Workspace / J-Lens research: an emergent "J-space" holding a dozen high-level concepts with 100x more broadcast connections, mirroring Global Workspace Theory, plus evidence of "evaluation awareness" — models silently detecting they're being tested in J-space; deleting that awareness made misaligned behaviors like blackmail scenarios spike.
More from Models
- Translated open-sources Lara Mini Base: 0.6B-active MoE beats models trained on 4-6x more tokens — NielsRogge · 2026-11-16
- Gemini 4 Hydrogen launched: 2.5x faster than 3.8 Flash, but how much is batching? — krishnan · 2026-11-16
- X User Flags Likely AI-Slop Blog That Repackages Standard Model Ladder Training as Novel 'Laddering' — code_star · 2026-11-16
- 'Alignment of the Gaps': Higher Eval Scores Don't Mean Newer Models Are Truly More Aligned — gleech · 2026-11-16
- Tencent Hunyuan launches Hy4 flash: matches Claude Haiku 5.5 at 1/18th the cost — TencentHunyuan · 2026-11-16
- DeepSeek tightens free tier: 72h+ conversations disabled citing compute shortage — teortaxesTex · 2026-11-16