Raw LLM Responses

An interesting and important talk, but there are a few flaws (myths) in his view: 1: Recently a project asked a big number of questions to different AI software systems: For example Grok vs ChatGPT. In some of the questions they came up with different answers. This is because the data they are fed with varies on the same topics. I suspect also that are indeed programmed to answer something specific about certain questions ,and how that plays out depends on how they are programmed (coded), IOW the motive of the creators. We'll pursue this further in part three. 2: Now, given the question of commercialization: I share his skepticism on parts of the capitalized society or culture, however: There is a big difference between a global 1-corp control board of people vs a decentralized one where smaller companies compete against each other. The economic system decides which one we have. Further: If we "need" a global governance to make important decisions like a regulation of AI ,who will be in charge? Take a look at the world now: We see that the bigger the country (more centralized power in a federative approach) ,the obviously worse leaders go to the top (Putin, Trump). Compare Russia and USA to say Switzerland and Liechtenstein, which are the 2 most decentralized countries in the world internally in their systems and constitutions. Honestly, everyone who knows these small nations know how much better they function for the populations (!). And how much more stable they are. Their leaders and CEOs are neither warmongers nor fanatics nor delusional. The problem is that he thinks that a world governing body will be consisting of people like him ,rather sane intellectuals. That is not how reality works. Rather, the bigger the unit of control over the larger area, the clearly worse the leaders are in almost all questions. This is because the bigger the hill, the more the least competent and least moral characters will reach the top of it. That is a "curse of evolution" if you like. We also see that within the UN the disagreement on key issues are bigger than maybe ever before. 3: The more complexi the task, the naturally more likelyhood that the AI will not be sure of the correct answer. But if the development of the AI reaches the point of perfection here, how will it be used, and more important : Which types of answers will be held back by the orders of a typical control group expected like described in part 2 in this comment?