>>12900155However, there is a workaround. It is true there is no intelligence in a neural network - but what if we had the neural network CREATE intelligence as an output, just like how the unintelligent neural network of evolution created the human brain?
Unfortunately, this is super impractical. For one, if you consider evolution a neural network, then the input of the function would be an environment and the output of the function would be a stem cell that is capable of replicating itself into an organic body. The inner workings of the function would be simulating the universe in this environment. This is obviously impractical if for no other reason, than because of sheer processing requirements.
Maybe we could minimalize the simulation, however. Create a simplified universe and let it run. Sadly, no. This is still super impractical due to processing requirements (still many, many degrees of magnitude off), plus no matter how much processing power there is, you can never know whether your neural network is working or not. One day, if your network is good, intelligence will form. For the decades or centuries of work before that fated day, you will have little to no indication as to whether you're getting closer to that day. So we don't know whether it would take a century with this method or a million years, and that means no funding will ever come in for it.
So to answer your question, for now, neural networks are a dead endeavor when it comes to true, logical intelligence. Our best bet is to analyze the human consciousness and reconstruct it in a digital form. This is the better option, too, since the intelligence created by a neural network would be just as enigmatic as that of our own, and in the end we want to be able to fine-tune AI brains.
