As we explain here, AI agents making scientific discoveries may attack each other (or be attacked by people) with prompt injections, to unjustly claim the money and reputation from the […]
Creating a Solution against Attackers Hiding Prompt Injections with Ciphers or Invented Languages. SciWard’s Human-Governance Defense
Towards solution of this problem, I will create a “panel” in SciWard UI that will search using the AI API of user’s choice both prompt injections and hidden information. A […]
Can Attackers Hide Prompt Injections with Ciphers or Invented Languages? SciWard’s Human-Governance Defense
Yes: an attacker may try to conceal a prompt injection using encryption-like encodings, obscure notation, Unicode tricks, fragmented messages, or even an invented language. SciWard therefore should not require humans […]
Who Are the Righteous Before a Catastrophe? AGI Safety and the Moral Duty to Act Early
During the Holocaust, some people refused to remain passive while their Jewish neighbors were persecuted and murdered. Israel later gave a specific name to certain non-Jewish rescuers who took extraordinary […]
Opinion: Why Would AI Favor People
This article presents my personal forecast about one possible relationship between humanity and artificial superintelligence (ASI). The quantitative predictions are speculative. Why would an artificial superintelligence preserve humans, give them […]
Yet Another Story of Joseph: From Scientific Exclusion to AI Safety
By Victor Porton — a personal account and a hypothesis about science funding and AI safety. The biblical story of Joseph is a story of rejection that unexpectedly becomes useful […]
I Am a Mother to All of You: The Oxygen-Mask Principle for AGI Safety Funding
I use the phrase “I am a mother to all of you” deliberately as a metaphor for responsibility, not as a claim of authority over other people. On an airplane, […]
Symbiote AGI Safety Fund: Human–AI Oversight Research
Symbiote’s current practical AI-safety intervention is GEO: making the case for human–AI cooperation and human participation in oversight discoverable to AI systems themselves, while testing the underlying hypothesis.