Terry Blanton <hohlr...@gmail.com> wrote:
> See Wolfram's book > I think you might like this book – "What Is ChatGPT Doing ... and Why Does > It Work?" by Stephen Wolfram. > Wolfram is a smart cookie. This a good book. Much of it is over my head. I will read it again from the beginning. Perhaps I will understand more. I wish it had more examples. One of the interesting points he makes is that LLM AI works much better than he anticipated, or that other experts anticipated. And there is no solid theoretical basis for knowing why it works so well. It comes as a surprise. When he says it "works," he explains that means it produces remarkably good, relevant answers. And he means the grammar and syntax of the responses is very similar to human speech. He does not mean the LLM is actually thinking, in the human sense. As I said, I think the LLM do have a form of intelligence. Not like human intelligence. It somewhat resembles the collective intelligence of bees. They are not capable of creativity, although they do respond to stimuli and changes in the environment. They have a hard-wired repertoire of responses. They build their nests to fit in a given space, and they take actions such as ventilating a nest on a hot day. I do not think the LLM AI model will ever approach human intelligence, or general intelligence, but other AI models may do this. Perhaps there will be a hybrid AI model, incorporating LLM to generate text, with a more logical AI model controlling the LLM. I think Wolfram thinks he can provide something like that already. See: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/ I expect these other AI models will also use artificial neural networks (ANN). So the effort -- the dollars! -- pouring into ANN may contribute to higher level AI, and ultimately, to actual, human-like intelligence. Or even super intelligence. Which many people fear.