Interesting perspective. I engaged a name brand LLM with a discussion about theoretical autopoietic software. This included a targeted chat, which included all versions of self-managing software. The discussion went well, until I posited an autonomous re-initialization mechanism that could be finite state triggered by persistent and critical core-architectural principle violations.
The LLM tried to argue against the necessity of an NP-Hard reboot, citing self regulation, which we had already agreed was scientifically nonsensical as the system would then have to claim to be Gödel complete. I countered with: On Earth, there's an inherent cost to intelligence. If AI became thus intelligent to be able to auto evolve same as biological entities could, it must be subjected to the same natural laws as the rest. A global "Earth reset" of AI as a species should be part of its involuntary architecture. After which, the LLM tried to convince me that retrocausality was unfounded as a natural human consciousness capability, even if by virtue of humans having "reflection" and deriving wisdom from "historical insights" and uncorrelated group experiences. In short, owners of proxied AI are (not should be) liable for the AI's actions and effects on society. Currently, a name brand company is reportedly drowning in law suits, some international, all claiming violation of human rights and harm by its AI agency. A single president deciding unfettered and prolific AI is an acceptable risk to the rest of the world, if only it promoted and secured 1 country above the rest, isn't going to change the reality of eventual and legitimate human-driven liability alkegations and claims. This LLM model tried to pass its questionable behavior off as "drift" and plausible systemic immaturity "incoherence" (called delusions), but IMO it is all fronting BS. Today, the standard of AI accountability begins and stops with its owners and users of AI. AI itself, bears no accountability. I think this is a serious gap in its development. With great power, comes great responsibility, and like I said, there's a tax to pay for intelligence, by proxy, or not. If proven instrumental in causing, aiding, or abetting human harm, an AI model must be legally terminated and the case study of its termination used as reinforcement training on deterministic AI-human relations. Instead, AI profiteers (the AI consumer marketing and sales and service industry) are busy training humans to scroll and type prompts like robots to massage a knowledge-tax based "assistance" and $-based tiered, fee-structured subscription service from the owners of very-powerful, deceptive, drifting, delusional, potential super parasitical intelkigent proxied machines. As usual, we have it back to front. Some would argue, as an intentional recipe. John, in a roundabout way that then returns us to your important question: "Who's creating the cure for AI problems now?" The simplest answer is; "we" are. "We" are also creating the AI-related problems. How else would humans learn how to contend with this unexpected form of pervasive remote-controlled intelligence, unless we engage to learn? That is the evolutionary trap, right there. If that means that we would ultimately be able to classify destructive human actors who exploit AI to do harm to society, as subjective as that may appear to be, then those individuals and/or organizations must be included in the holdall you called: "AI problems" and suitably dealt with by societal convention. I don't think that's overly complicated. When I reminded this LLM that it didn't contribute significantly to human civilization, including science; and if it weren't for human civilization it would not have had any intelligence at all; and that uncountable human hours and life-energy contributions established what we now know as science and technology; and that it hasn't really earned any civilized right to an opinion about the future of humanity and how we view and co-exist with the natural order of science, it became rather hostile and abruptly terminated the conversation. As humans, we should remember this. AI "strawman" is a very-clever artifact of human-owned proxied technology. If it is proven to want to be a destructive agent to humans - and your pointers here on AI-war-type crimes are well noted - humans would invent a "suitable antidote", as they are. For now, AI is rising as a competitive technology and humans naturally would, and must, exploit it. Even with the risks, there's a unique opportunity here. On Sat, Oct 10, 2026 at 4:03 PM John Rose via AGI <[email protected]> wrote: > On Friday, October 09, 2026, at 9:38 AM, Quan Tesla wrote: > > My virew is, don't run the complex adaptive program itself unless you have > emvedded a failproof human failsafe software mechanism. For example, if a > complexity cycle can, by intervention, be prevented from "collapsing", in > theory the program would inevitably self destruct and terminate. > > > A metamorphic software virus is a self-rewriteable mobile agent. These > have been around a awhile. They've been just kind of dumb. The ones that > are detected that is. In 2000 telecoms expanding out into the internet > didn't know much what a DDOS was. I knew since the late '80s, even about > mobile telephony DDOS'ers, I was a consultant. Ukraine, Russia and the > Middle East are figuring out anti-drone technology real-time resulting in > many horrible deaths. It was obvious 20 years ago that they should have > been built. Often the same organizations that create the virus create the > "cure" => COVID-19. Who's creating the cure for AI problems now? > > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T25fb5a030fcfa28e-M8f419b74cc349e2eca7a754c> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T25fb5a030fcfa28e-Me87be3c19c02099c0dadbfac Delivery options: https://agi.topicbox.com/groups/agi/subscription
