3 ms·
It's important to understand that large-language models like ChatGPT do not "think" about the prompt and do not in any meaningful sense "understand" the result
by zahlman 2y ago
It's important to understand that large-language models like ChatGPT do not "think" about the prompt and do not in any meaningful sense "understand" the result they give you. No matter how sophisticated they get, fundamentally they are still computer programs - systems for choosing words (and code fragments) that look right next to each other, according to some inferred rules. They are not driven by will or consciousness. You might be able to get ChatGPT to "tell" you (produce textual output) that it's intelligent and has feelings, but this is just as meaningless as if a book were titled "I am Intelligent and Have Feelings".
Or as Angela Collier puts it, AI does not exist (https://www.youtube.com/watch?v=EUrOxh_0leE https://www.youtube.com/watch?v=EUrOxh_0leE).
Such programs also often output things that are completely wrong and have no basis in reality, called "AI hallucinations", and as a beginner you are in no position to identify them. This is dangerous, because you can learn wrong things with no way to know that they're wrong. It could take a very long time for you to figure out the mistake, at which point a bad habit may already be ingrained. Responsible human teachers know what's technically correct, and will proactively warn that they're making a simplification - and will certainly not make something up just because it sounds right in context.
Learning Python is not fundamentally different from learning any other programming language, which in turn is not significantly different from learning anything at all. You must start with fundamentals and be guided through them in a reasonable order; at a certain point there are many other lower-difficulty things which you then learn in any order, and you should choose from those according to your own interests.
However, learning your first programming language (no matter what you choose) is fundamentally different from every other programming language you learn - because you must also learn how programming works. Unfortunately, there is really no good way to learn that "in a vacuum", independently from actually choosing and using a programming language.
The community of "regulars" for Python questions on Stack Overflow (of which I consider myself part) has some advice at https://sopython.com/wiki/What_tutorial_should_I_read%3F https://sopython.com/wiki/What_tutorial_should_I_read%3F . I wish to echo the last point: we strongly recommend against LPTHW ("learn python the hard way"). The material is presented in a very awkward order; the instruction greatly overestimates the difficulty of some concepts and underestimates the difficulty of others (compared to feedback we've received, in the form of the literal millions of questions that have been asked about Python on Stack Overflow); and the author was strongly resistant to the Python 3 migration for very ill-considered reasons disconnected from the facts of the situation (see e.g. https://eev.ee/blog/2016/11/23/a-rebuttal-for-python-3/ https://eev.ee/blog/2016/11/23/a-rebuttal-for-python-3/ - note that this link is from 2016; support for Python 2.x ended in 2020.). Further, the course is written with the assumption that you're comfortable with the command line, even though it's intended for people who haven't learned other programming languages before, and tries to use strained command-line metaphors that won't land properly for many beginners.
- meiraleal 2y ago> It's important to understand that large-language models like ChatGPT do not "think" about the prompt and do not in any meaningful sense "understand" the result they give you. LLM/MLMs are human intelligence as a service. With enough context memory (think terabytes, petabytes) and a well-engineered system, it can simulate artificial intelligence in a way that humans wouldn't be able to distinguish. And it could program itself to improve in ways that me, you and Angela Collier have no possible way to know. Dismissing AI makes some people feel smarter, I think this is dumb. Your reply, for example, looks generated by ChatGPT (similar structure, it is long and gets repetitive).