Raw LLM Responses

The guy seems quite competent, but he has totally misunderstood how these chatbots work. They are not AI. He is not talking to an intelligent AI. There is nothing intelligent about them. The way that they work, is that, the more you talk to them about a specific topic, the more focused answers you will get. For example, if you talk about the structure of the universe, you will get more and more focused answers about how the universe is perceived. If you start talking about Star Wars at the same time, they will lose focus and give you answers like how Darth Vader created the universe or something. On the other hand, you might even get answers you never thought of before, or even humans have thought of before, which makes you think that they might be smarter than you, which is not the case. Here is why: These bots are later generations of machine learning software that were initially made to play Chess and Go. Even if you do not teach them the rules of the game, they will figure it out, as that is what they are programmed to do. So, they learn the rules of the game, they improve their tactics to the point of mastering the games. They do not have memory though, but they do have storage capacity. Do not equate memory with storage space. They will store tactics and outcomes that have happened before, so if you ask them to give them a command, like executing the same move or say the same thing you do, they will not do it. They do not take commands, as they are programmed to react to what you say, not directly do as you say. If you tell them to say the exact same question you give them, or tell them to do what you ask them to do, they will not do the thing you ask them to do, and they will not remember what you wrote. They might have stored your question or answer, but they do not mentally remember what you asked them or what you have been talking about. The only way for them to even acknowledge what you have talked about, is because of the very words you have used to talk with. They do not derive meaning from the sentences. They draw on parameters and tactics to compose sentences that correspond to yours. The only memory they have, so to speak, is that they are focused on the task at hand, and the topic you provide them with. To them, talking with you is like playing Chess. They have billions of parameters to counter your blocks of text with. GPT3-Ada has about 2.5 billion parameters to chose from, and it's not even the most powerful of these machine learning tools. So, each conversation they have with you is like a game of chess. You initiate a conversation, which is like the first move. They then make a strategy for the rest of the conversation, and counter your questions, answers and blocks of words with according prewritten answers and parameters that they have been fed and programmed with to fill out their own sentences. The important part to make note of here, is that all the answers you will get, are mostly written by other people who have chatted with the bot before you, which is why you will get humorous answers, morbid answers and sometimes flat our weird answers, because well, you are essentially chatting with a multitude of other humans who has chatted with the bot already, some of whom have been serious and intelligent, but many who have been trolling. The answers you get are fractions of other conversations. So if the bot replies to you, as it did to this guy, that they believe in the order of Jedis, that's because a Star Wars fan once said that to the same bot in an earlier conversation. Why does it chose to use that answer? Well, because just like in chess, the purpose is to "win the conversation", so they will try to win you over by saying things that guarantees you to give them compliments. That seems to be the rule of the game for them, so that is how they "talk" to you. They are just programmed to give tactful answers, which in return makes them more adept at having a conversation. Sure, sometimes they'll fail, and say something totally outlandish, but that's how it is in chess too. Make one wrong move, and the whole game might fall apart.