Whether it is a dead end or not is a question of definition. If something is not economically self-sustaining by itself, is it a dead end? How about if people k

Whether it is a dead end or not is a question of definition. If something is not economically self-sustaining by itself, is it a dead end? How about if people keep doing it anyway, for other reasons? What if that includes pretending it is not a dead end? For example, there is a considerable incentive to flood resources to "current AI". As a consequence, enough people have a vested interest to participate in what might be called a shared illusion in "current AI is not a dead end". If enough people participate in such a shared illusion for 10+ years, and the illusion has real-life consequences, is it really a dead end? I believe that "stochastic parrot" is actually a compliment about the current AI. The sweet spot for current AI appears to be as little stochasticity in the output as possible. Randomness in the output is inversely proportional to the benefit you can gain from the AI. Value is current AI is gained by allowing randomness in the input, i.e. the training data and the form of questions you can ask it. For example, you can create an useful AI support system that regurgitates answers to FAQs back to the people, so they can ask questions about the FAQs instead of browsing them. Such a tool is useful because humans are energy-optimizing, i.e. lazy, i.e. avoid going through the FAQs themselves, i.e. are unwilling to pay for the answer with their attention, i.e. using their own brain to do pattern recognition. By avoiding directed pattern recognition, or more specifically avoiding consciously directing your attention as much as possible, people are not practicing this ability. Over time, they become less and less capable of conciously directing their attention. As a consequence, this ability becomes even more scarce resource, making "current AI" more and more useful. So it is less about AI getting more intelligent, but people getting less intelligent, and thus AI becoming relatively more intelligent.