I observed a growing concern that household robots equipped with large language models (LLMs) or vision‑language models (VLMs) could be manipulated, similar to a virus or “COVID‑like” attack, leading to unsafe behavior toward families. This motivated me to prioritize research on adversarial attacks in robotics.
- Paper: “Highlighting the Safety Concerns of Deploying LLMs/VLMs in Robotics”
- Decision: prioritize identifying and mitigating adversarial attacks because they could cause robots to act dangerously toward humans.
- Emphasis on rigorous safety engineering rather than ad‑hoc fixes.
- Authors: Xiyang Wu, Souradip Chakraborty, Ruiqi Xian, Joleen Liang, Tianrui Guan, Fuxiao Liu, Brian Sadler, Dinesh Manocha, Amrit Singh Bedi
The complexity of home robotics can be managed and reduced if Rust is used as single language across the stack!
On 2026-03-03 19:45 UTC I wrote, "Literally, we are living in the prequel to the terminator world!"
Timestamp of the AI note: 2026-03-03 07:40 UTC.
Quote: AI is finally stepping out of our screens :).
I felt a strong surge of emotion on 2026‑10‑04 when I wrote, “This is insane!”.
Timestamp: 2026‑10‑04 16:53 UTC – “This is insane!”.