Big Sister is watching thinking for you.
On the other hand… It won’t be too hard to train an algorithm with different (there’s no such thing as “unbiased”) sources. The other half of the population is lazy, too, so someone is going to take advantage of it.
I agree with you completely. I also think it was a mistake to call the new technology “artificial intelligence”. “Machine learning” or “Statistical machine” would have been better, because even if ChatGPT gives amazing results, the AI has neither a consciousness nor a creative intelligence that can think outside its “box”. The “box” is the statistical values that have been trained into it.
Therefore, I consider ChatGPT as a tool for very specific tasks, for some of which there are existing tools, but which are cumbersome to work with or time-consuming to use.
The advantage of ChatGPT4 as a tool is the fact that you can explain the task to the AI and train it. For example, I wrote a training block that allows ChatGPT to output optimized prompts for MidJourney, with very satisfying results, for creating illustrations and cover images.
Another application is a translation aid for fiction words for which there is no equivalent in the dictionary in the respective language. Fiction novels often create words that are difficult to translate, but with ChatGPT you can find an equivalent that will be understood by the readers in the translation (screenshot as example).
I agree with you, you have to know exactly what you are asking and what you want to achieve by using the tool.
As long as you don’t publish any of that, nothing’s wrong with it. Otherwise you’re asking for trouble. Because you have no control over (or even knowledge of) the training data. See: https://www.reuters.com/legal/transactional/lawsuits-accuse-ai-content-creators-misusing-copyrighted-work-2023-01-17/
I am frankly relieved that it comes to a judicial clarification, but do not think that a copyright problem is derived from it, because it is neither copied nor alienated the work of an artist. The AI training works in such a way that for example 10 million cat pictures are looked at and statistically seized, in all colors, poses and appearances, so that the AI determines a statistical value, how a cat looks. Within 170 billion training sessions, the single image becomes irrelevant.
Incidentally, the training algorithm follows the same principle as a draftsman who makes cat pictures hundreds of times as a concept drawing until cat drawing becomes second nature to him. The AI routine does basically the same, only faster and more comprehensive. I don’t see any copyright infringement in this, but hopefully the courts will rule competently on this.
The question here is how do you depict that cat in writing.
Surely you are not gonna satisfy yourself with “Fur, tail, pointy ears, hunts mice.”
The description of a cat or any other image or object can be as comprehensive as you like. Since version 5 MidJourney also understands linguistic descriptions much better than before. I have ChatGPT translate my description into English and additionally suggest a number of keywords that I may not have thought of. This gives a better result for my application, while meeting my target.
But in writing, GPT is not looking at millions of pictures…
It is mimicking writers.
It doesn’t even understand what it is saying.
And now with the question you likely won’t like:
If I understand right, that tiger (nice looking, that, I won’t argue) is the result of your instructions.
How proud of having produced it are you ?
GPT has no intelligence or consciousness, but has strengths in linking statistical values and pattern recognition. It does not understand what it outputs, because it has no “understanding” of what it is doing, although it sometimes appears that way.
But to get back to the utility for texters: ChatGPT is a text-based AI, so it can be extremely useful for research, or for correcting errors in manuscripts.
Of course, an author can sit down and do it all himself, surf the net for days for a historical novel, use Wikipedia, but why do that to himself, when ChatGPT can give an appropriate summary in seconds?
This is the added value for authors, because they can then spend more time on the actual task.
You are deviating from the initial subject.
Research ? I have no issue with a smarter Google.
Composition ? That is something else. Completely.
Ultimately, with enough circumlocution, I got ChatGPT to write a satire about undermining Feminism. I got it to invent a religion, too, as I mentioned before.
I don’t find it Woke, I find it full of legalistic protections for the developers. Woke can be what it abhors and ChatGPT seems to “desire” to be as vanilla as possible.
This is one of the dangers of the thing. Could we get Naked Lunch or Howl from it? Nope.
That’s okay: It’s a tool. But I can use a tool to write anything I desire and it shouldn’t get in my way.
Garbage in, garbage out, as far as the LLM goes. It should be trained well. It should let me be as dirty or subversive as I want.
A screwdriver would never say “I won’t screw that kind of screw!”
You’re asking the wrong question. For me, it’s not a question of whether I’m proud of it, but that AI allows me to realize my idea of a picture, even though I’m not an artist who can possibly draw it without help - even if not in seconds. For me, it’s the freedom to express creativity even more, whether I want to draw a cat or an alien with pointy ears.
I’m excited about the possibility and see the benefits for my work.
I have no further questions.
As to the main thread here, this thing can help you edit in a broad way, but it cannot help you write–which is a reflective, integrated process ChatGPT cannot do. It doesn’t remember between sessions and sometimes forgets information during sessions (Ask it to remember a word, eventually it forgets it) and, importantly, is not self-reflective.
Reflection and self-reflection, synthesis and evaluation of the same are hallmarks of intelligence and self-awarness under several definitions. The bot can’t do these things. It does not write.
As you said, it uses pattern recognition without understanding. Any story it makes falls short of being written therefore, and as Vincent Vincent first said, the writer didn’t write what was generated by it.
It is a screwdriver, not a human editor. (I say all this in a friendly way) 
But it does write, unfortunately.
(Or patch parts together that it fetched here and there → ©)
So you think only artists have the right to visualize their ideas? Maybe you are also of the opinion that only published authors have the right to publish, or only professional athletes are allowed to engage in sports? The art of dealing with AI is another. Apart from that, many graphic designers who make covers use prefabricated stock photos, as you can often see from the appearance of the same motifs on different books. AI at least allows me to generate something unique, because the same image doesn’t exist a second time.
Which already makes it more advanced than some US presidents, to be fair. (It outputs intelligible sentences.)
Nope.
Nope nope nope.
I am 100% for people personal growth.
Selling the result. Drowning the market in real-fakes…
Starving those who actually can…
A tiny bit less.
Wrong again …
However, it is capable of recognizing certain patterns in the text and linking them with a high probability. By the way, this does not happen on the basis of words, but in smaller units, in word fragments.
Since the number of word combinations in each language is limited, the AI knows them all and can create new ones from this pool.
Once again, with over 100 billion training units, the individual text becomes irrelevant. There is no copyright infringement. If you keep insisting on it, prove it.
There are amazon books written by GPT.
You are saying the exact same thing I am, but you somehow see it as a good thing.
I am actually speechless at this point.
Patching copyrighted material together is a copyright infringement.
