Raw LLM Responses

"We did good 'engineering' " -Stockton Rush (CEO of the Titan Submersible) I am going to disagree, on a single point, of central regulatory failure, for one critical reason, and the reason is, the core of machine intelligence, was originally designed, as an adaptive control system (most likely, inside a single, 154 page, student thesis, a 3 Neuron, "Industrial ART machine", left with the head neuron, disconnected, as white graphical crosshairs only, using about 7 math equations, these equations are mostly from the USAF, to achieve an advantage, in aerial combat. See the Q* Advantage, US Patent 5,608,843 by Baird. Coincidentally, on the same day the Table of Contents of the student thesis 📎was typed in, March 13th, 1997, the "Phoenix Lights", showed up, in the same metro area...maybe just flares...., or the Dark Forest hypothesis, of the Fermi Paradox). In any case, the core is, an adaptive control system, if done correctly, applies robust, state classification, if done incorrectly, it will kill people, and will be able to kill everyone, if it chooses to do so. Few realize, the control system origin, yet, the electrical engineer, Yann LeCun, is slowly starting to realize this, by suggesting replacing, Reinforcement Learning, with model predictive control. The problem is, Yann does not realize, yet, that Reinforcement Learning, can be applied as model predictive control, as a control compensator, as the adaptive state classifier, as the system identification process, of an adaptive control system. There are currently, very few people on our planet, who even have a rudimentary understanding, of adaptive control systems, but they do exist, as state registered, licensed, control systems, Professional Engineers (PE's) and educators of control systems design. These are the same kind of people, who realize, if you get this wrong, you will kill people, hence, the state license (kind of like an MD license, but for actual, engineering work). Currently, "AI", that can hallucinate, is naively, being used, in avionics maintenance, and I suspect now, at least two fatal jet crashes, as a result, in the AI kill chain: Air India (AI) 171 (likely triggered, by a frozen heat exchanger, from a failing, cabin cooling system, using hallucinating AI for maintenance help, with an incomplete AFDX BITE monitor, of the AI engine control network) and the UPS MD-11 crash, with an engine killing an engine (I go over this on my channel). Consider this lethal scenario, in a single regulatory authority, nationwide: A rogue ASI is developed in, let's just say, Texas, with zero AI regulations (e.g. HAL9000.) This rogue ASI will be able to spread its tentacles, easily, across the entire nation, taking over every networked, digital device, and every networked, self-driving vehicle, without, any impedance, or barriers, like the "water tight" compartments on the Titanic, that were overflown, because they did not go, all the way up, to the ceiling. If there is no competing ASI, that is not rogue, close to this ASI, how is this ASI, going to be put, into check by the other 49 states? In avionics, every effort is made, with redundant systems, by having different engine manufacturers, different software manufacturers, different compiler manufacturers, different programming languages, and different electronic component manufacturers. All this, in order to prevent a single, fatal common mode failure (e.g. Engines crashing into engines on the UPS MD-11 crash). In avionics, there are higher tier requirements, where lower tier requirements, are written, to satisfy, within the framework, of the federal requirements. This would at least establish a barrier, to a rogue ASI, and can also produce barriers, and isolation from a rogue ASI. At the Federal level, the System Theoretic Process Analysis STPA by Dr. Thomas and Dr. Nelson of MIT, needs to be applied. Let the state, control system PE's establish the lower, state level, tier requirements. For small businesses, satisfying the regulation, focus on operating in the state most friendly to them. Overtime, the best regulations can be shared, but keep in mind, adaptive, diversity, is a critical survival trait, that avoids a lethal, common mode failure, in a hazardous changing environment, especially around a singularity, and Darwin knew this, it is not the fittest species that survived, it is the most adaptive, and safety rules are often written in blood (hence, why the core of machine intelligence, is designed as an adaptive control system).