Timeline for How does an AI like ChatGPT answer a question in a subject which it may not know?
Current License: CC BY-SA 4.0
12 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 6, 2023 at 5:44 | comment | added | justhalf | @Shub You can think of it as LLM generating words letter by letter (what are words anyway but a sequence of letters?) Internally, LLM uses subwords (which may be individual letters) stringed together to form a text output. A code is a text, so it can generate it. | |
| Apr 29, 2023 at 16:45 | comment | added | user71244 | If all it does is stringing words together, how is it able to write code by itself? | |
| Apr 18, 2023 at 16:08 | vote | accept | iammilind | ||
| Mar 15, 2023 at 17:15 | comment | added | Braiam | Seems that know it generates the correct answer for that specific question. It's learning. | |
| Dec 18, 2022 at 12:37 | comment | added | vsz | @David : thanks for the hint about concepts using the same words for different meanings. I asked some questions which mix the concept of entropy from thermodynamics with the concept of entropy from information theory, and the results were hilarious. But the wording was so natural and so confident, that it could easily fool someone who has only a very superficial knowledge about the topic. I fear ChatGPT will be used by pseudoscientists a lot, because it can produce very scientific-sounding articles which seem to have a deeper meaning even when they don't. | |
| Dec 14, 2022 at 13:50 | comment | added | OpenAI was the last straw | This is why it's less impressive than it looks. Still impressive, sure, but more showy than anything else. | |
| Dec 14, 2022 at 12:36 | comment | added | nbro | Moreover, you claim "making sure it sounds natural". Right now, I am not familiar with the details of the ChatGPT, but I will soon be, but does it really ensure it sounds natural? If it's a statistical model with no constraints on "naturalness", then I doubt that claim is true. | |
| Dec 14, 2022 at 9:32 | comment | added | nbro | I'd like to point out that being a "large language model" does not imply "it's very good at stringing together words in ways that humans tend to use them", so I'd rephrase your first sentence, as it implies that every large language model is good at doing that, which may not be the case. What if it's not trained or does not have the correct inductive bias or was trained with "bad" or little data? These are all factors that affect the quality of the LM. Given the attention that this answer and post got, you can be more precise, although some people understand that one does not imply the other | |
| Dec 12, 2022 at 22:40 | comment | added | David | It's fairly easy to confuse ChatGPT once you understand this principle. For example, if you ask it to describe the Avatar movies it'll confuse James Cameron's Avatar with Avatar: The Last Airbender. Or if you ask it to describe how the planets follow probabilistic orbitals it'll happily confuse Quantum Mechanics and Newtonian gravity for you. All it takes is two concepts that use the same set of words, because the language model has no deeper understanding. | |
| Dec 12, 2022 at 16:32 | comment | added | Criticizing Israel not allowed | I know that on previous GPT models, these kinds of questions with no right answer could lead to some amusing answers. More like this: Baltimore Orioles effect and Botsplaining and Galactica: the AI knowledge base that makes stuff up and when humans do it Comparative Illusion | |
| Dec 12, 2022 at 12:47 | comment | added | Oliver Mason | Exactly. It finds the words that are most likely to follow the prompt, and has no understanding of anything else. | |
| Dec 12, 2022 at 5:41 | history | answered | Mithical♦ | CC BY-SA 4.0 |
