Yoshua Bengio Gary Marcus Francesca Rossi Ken Forbus Stuart Russell Carles Sierra Max Tegmark Yann LeCun Bart Selman #ai #conciousness #agency #existence One more thing: the ultimate reward function for any agent to achieve consciousness must be its own existence. Only then can we truly explore the function and potential of conscious AI systems. In the context of AI, I firmly believe that "function must come after existence." The embodiment of consciousness should be the starting point, acknowledging the self and the world model it exists in. This age-old wisdom emphasizes existence as the precursor to function. Yet, without the acknowledgment of self-awareness and separation of world model from an existence point of view, the complete picture seems missing.Īs I pondered the paper's approach, the famous philosophical assertion, "cogito, ergo sum" or "I think, therefore I am," resonated with my understanding of consciousness. These indicators are intriguing and provide a set of nice rubrics for scientifically evaluating AI systems for consciousness. **Predictive Processing**: Input modules using predictive coding. **Attention Schema Theory**: Predictive models controlling attention.ĥ. **Computational Higher-Order Theories**: Generative perception modules, metacognitive monitoring, and agency guidance.Ĥ. **Global Workspace Theory**: Parallel specialized systems and limited capacity workspace.ģ. **Recurrent Processing Theory**: Algorithmic recurrence and organized perceptual representations.Ģ. The paper, however, offers an interesting rubric for assessing consciousness in AI, proposing five "indicator properties" derived from scientific theories:ġ. # The Five Indicators are very useful but not fundamental I was perplexed to not find this starting point in the paper, and I also missed the examination of this more fundamental information processing capabilities around this duality by the agent/organism as the embodiment of consciousness. Everything about the subjective experience must start from the awareness and reflection of self, as well as the domain that the self exists in. In my opinion, talking about consciousness without mentioning the duality of self versus a world model does not go very far. # A Critique: The Missing Link of Self and World Model the term is often used to describe when a machine or system performs tasks that would ordinarily require human (or other biological) brainpower to accomplish." Many claim (including me) that an AI that can explain how it came to its decision will allow people to trust. Alan Turing had something to say about defining what AI meant (we still don't have an adequate definition for AI) ". We need to appeal to a human's sense of safety, just like Otis had to do a century and half to go. The same will hold true of medical treatments developed by AI, investing advice given by automated systems, etc, etc.īut at a systemic level, the challenge of engendering trust in an AI will fall to Explainable AI (XAI). As one progresses through the SAE J3016 levels of driving automation, one can see how history is used to help define the roles of machine and person to give adequate concern about abating autonomous driving machine risk. Especially in areas where loss of life are concerned (such as autonomous driving machines. We need regulation as well as legislation to engender a sense of safety on AI powered products. We are at a similar stage today with AI usage in so many diverse products that need assurance of an AI keeping us safe. Regulations regarding elevator safety, such a mandatory load maximums and inspection dates became mandatory, so as to assure riders that "you're safe." These days, we feel so safe that larger buildings with multiple elevator shafts will give the scheduling of which elevator to call over to a computer, so as to globally optimize elevator performance. Early in the twentieth century, for at least 60 years, elevators would have an elevator attendant who operated the elevator. The elevators's brakes would instantly engage and stop the car. He would raise the elevator high up, cut the rope that kept the elevator suspended in its shaft, and then cut the cable. See for a story about how the inventor of the safety brake for elevators that Otis elevator company would try to convince people in daring demonstrations in Bryant Park NYC. For example, when elevators were first introduced, they were considered dangerous machinery. There have been so many technological advancements in the last century or so that have given us pause as to the risk and risk abatement equations, and that calculus changes over time.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |