Import AI’s latest issue introduces Humanity’s Sixth Sense (HSS), a benchmark created by Elorian and Scale AI to evaluate intuitive visual reasoning in multimodal models. HSS comprises 522 questions drawn from 288 images and 234 short video clips, totaling 17.6 hours of video, and covers four categories: temporal & causal dynamics, physical & spatial logic, abstract & contextual inference, and a fourth unspecified category. Human performance on the benchmark is high, with an aggregate pass@1 of 93.1%, while the best AI system, GPT‑6 Astra, achieves 53.6%. Chinese open‑weight models perform poorly, with GLM‑5.3‑flash at 26.8% and Kimi K3 at 25.5%. The most common AI errors are missing critical details (21%) and misidentifying objects or roles (20%). The article argues that HSS demonstrates that large language models can handle tasks previously thought to require specialized spatially aware world models, challenging assumptions about the necessity of separate visual reasoning systems. In addition, the newsletter reports on Reflection’s Beam, a 501‑billion‑parameter mixture‑of‑experts LLM with 23 billion active parameters, promoted for its token efficiency. The piece also discusses AI’s potential to accelerate the space economy, suggesting that autonomous systems could dominate space operations and create compounding advantages for early movers. Finally, a new benchmark, TasteVal from P‑Zero Research, measures AI research taste across eight tasks, showing that frontier models have doubled their experimental research taste every three months since December 2025, with Opus 5.5 surpassing human baselines by a compute multiplier of 2.30×. The author links these advances to the likelihood of recursive self‑improvement becoming feasible by 2027. All claims are attributed to the respective sources cited in the Import AI newsletter.
Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space
Import AI reports that the new Humanity’s Sixth Sense benchmark reveals significant gaps in AI visual reasoning, with GPT‑6 Astra scoring 53.6% pass@1 versus a human baseline of 93.1%. The benchmark includes 522 questions across images and short videos, testing temporal, spatial, and contextual inference. Chinese open‑weight models perform poorly, with GLM‑5.3‑flash at 26.8% and Kimi K3 at 25.5%. The article also highlights a 501‑billion‑parameter MoE model, Beam, from Reflection, touted for token efficiency, and discusses AI’s potential to accelerate space industrialization and recursive self‑improvement.
Why it matters
The benchmark shows current multimodal LLMs still lack intuitive visual reasoning, highlighting a key limitation for applications requiring real‑world perception. Beam’s token‑efficient design could influence future large‑scale model deployment strategies.