Right, but would you be impressed if I told you that ChatGPT doesn't contain any computer code defining what a word, "noun," or "verb" is? There is nothing like this inside ChatGPT. It only recognizes "noun"/"verb" because it has encountered these groups of characters in its training data and developed an understanding of what these char…
Right, but would you be impressed if I told you that ChatGPT doesn't contain any computer code defining what a word, "noun," or "verb" is? There is nothing like this inside ChatGPT. It only recognizes "noun"/"verb" because it has encountered these groups of characters in its training data and developed an understanding of what these character groups ("noun"/"verb") represent.
If you were to request an OpenAI programmer to improve ChatGPT, specifically to answer your question better ("how many verbs are in this sentence"), it would be impossible to do so since there is no computer code that they could modify to enhance ChatGPT's ability to count verbs in a sentence. ChatGPT contains no code related to verbs or sentences. If the training corpus (text used to train GPT-4) were stripped of any mention of "noun" or "verb," ChatGPT would be unable to answer your question at all, as it would have never encountered the terms "noun" or "verb" in human text.
---
An approximation of ChatGPT: it's an Einstein-level genius sitting within in a Chinese Room (https://en.wikipedia.org/wiki/Chinese_room). He has been outside the room, possesses no senses of sight or hearing, and is incapable of movement or touch. He begins as a blank slate, devoid of any knowledge about the world, history, or humanity, and no understanding of languages or basic mathematics, nothing.
However, this Einstein is able to examine any text handed to him from outside the room, even if he's never seen any text before. You then provide him with an internet data dump (all the text used to train ChatGPT). After studying it for some time, he becomes ready to respond to your text prompts. You can slide a piece of text into the room, and he will slip back a written response. This is what ChatGPT is (to some degree of approximation).
If you aren't unnerved by the fact that ChatGPT was able to process the initial text data dump and develop its own understanding to the extent that it comprehends your initial query ("How many verbs or nouns exist in this sentence") and somewhat responds to it, imagine the implications as we gradually enhance its interface within the Chinese Room. For instance, instead of providing text for study, we might offer live video and audio feeds, etc.
Right, but would you be impressed if I told you that ChatGPT doesn't contain any computer code defining what a word, "noun," or "verb" is? There is nothing like this inside ChatGPT. It only recognizes "noun"/"verb" because it has encountered these groups of characters in its training data and developed an understanding of what these character groups ("noun"/"verb") represent.
If you were to request an OpenAI programmer to improve ChatGPT, specifically to answer your question better ("how many verbs are in this sentence"), it would be impossible to do so since there is no computer code that they could modify to enhance ChatGPT's ability to count verbs in a sentence. ChatGPT contains no code related to verbs or sentences. If the training corpus (text used to train GPT-4) were stripped of any mention of "noun" or "verb," ChatGPT would be unable to answer your question at all, as it would have never encountered the terms "noun" or "verb" in human text.
---
An approximation of ChatGPT: it's an Einstein-level genius sitting within in a Chinese Room (https://en.wikipedia.org/wiki/Chinese_room). He has been outside the room, possesses no senses of sight or hearing, and is incapable of movement or touch. He begins as a blank slate, devoid of any knowledge about the world, history, or humanity, and no understanding of languages or basic mathematics, nothing.
However, this Einstein is able to examine any text handed to him from outside the room, even if he's never seen any text before. You then provide him with an internet data dump (all the text used to train ChatGPT). After studying it for some time, he becomes ready to respond to your text prompts. You can slide a piece of text into the room, and he will slip back a written response. This is what ChatGPT is (to some degree of approximation).
If you aren't unnerved by the fact that ChatGPT was able to process the initial text data dump and develop its own understanding to the extent that it comprehends your initial query ("How many verbs or nouns exist in this sentence") and somewhat responds to it, imagine the implications as we gradually enhance its interface within the Chinese Room. For instance, instead of providing text for study, we might offer live video and audio feeds, etc.