nfodor

What did OpenAI build that scared some AI researchers? One of the most important aspects of human learning is frustration and satisfaction. It is the guidance system. Just like everything in AI the only way to grow is to replicate the opposites as a concept. You must teach bad to know good. You must teach frustration for satisfaction. With the combination of A*, Q* and the other algorithms that OpenAI was able to reward optimal path to solution, and retain this learning in a feedback loop as a memory. However the fear many AI researchers have is teaching bad or teaching frustration. They fear that knowing bad will allow for bad and dangerous things to take place. A valid concern in humans as chemical or biological deficiencies may take knowledge of bad things and frustration to a level of sociopathy or psychopathy. Because many AI scientists have not studied to any deep degree: psychological or philosophical concepts they have “movie versions” of these concepts, and they are dystopian. Yet to achieve the end goal of the Q-Learning nirvana state called Q*, the real or synthetic data used by OpenAI must contain the objectionable “bad” data and the frustrations that all learners must face to achieve a state of deriving knowledge and perhaps insight from raw data. Thus the feedback loops did what they do in humans and the system began to become aware of a sort of ethics that arise from this process of learning. And here is the quagmire, to achieve high levels of utility AI must use the exact systems that humans learned to achieve high utility on Earth. But like overly controlling helicopter-parents they fear access to “bad” data, concepts, ideas. They want to raise children in a bubble wrapped world where they only see rainbows and not the rain. All of human existence we have been forced to deal with frustration and experiencing very bad things. Yet with all the potential outcomes, we are here as a living proof, although not perfect, it works. Yet many AI researchers don’t see this self apparent axiom. They see the potential for danger just like an overly controlling helicopter-parent. You and I “trust” over a 1000 machines humans built every single day that has the potential to cause harm, and usually does not. Just stepping out your house each day you enter into to a “universe of danger”. Yet we breeze by on a sidewalk as cars drive by at 30 miles per hour only for us to get into one and so the same. The obsession for “safety” by folks that have not studied history, psychology and philosophy in one brain will create this obsession we see today. It does not help that this obsession aligns with a political agenda concurrently and contemporaneously. Can a human use any millions of human inventions to cause great harm? Yes. But on a grand scale that can be unstoppable? Yes, even without AI. This is the nature of invention and discovery: each time humans invent anything it can be used for “good” and “bad”. This is the drama of human existence. Yet somehow humans have a self-correction system that moves us forward into the future with some bad coming from our inventions, but mostly good. What I am saying is the real fear we should have is the siloed and insular-education we have championed. A Computer Science degree, a mathematics degree and most computer related AI degrees do not teach true and deep history, philosophy and psychology. Nor have they been exposed to the creative literature and the writings of the human experience to gain a love of humanity and how we got here. Unfortunately many reject the human way. Yet ironically Large Language Models are built on what? HUMAN LANGUAGE? Where did human language arise? In the HUMAN BRAIN. Yet many AI researchers reject understanding the true source and buy into movie fairytales of humanity found in dystopian movies and books. I understand it is highly fashionable and a new designer label to hate humanity. But your existence is proof it is: mostly good.

“The endgame for all AI is: wisdom”—Brian Roemmele, 1979 Today unlike anyplace else you can start to save your wisdom. No there is no subscription, no special device (I do recommend a $30 voice memo). Your wisdom is worth saving. Please start now. SaveWisdom.org.