Thousands of authors urge AI companies to stop using work without permission

OpenAI has maintained since this whole fiasco started that their datasets were not “scraped”. Too, I think I’ll trust his 30+ years experience with AI and his degrees, rather than trust mutuals that have zero experience working with LLMs on a daily basis.

 

https://news.bloomberglaw.com/ip-law/openai-facing-another-copyright-suit-over-ai-training-on-novels

 

He works with these kinds of data sets every day, he knows how they’re acquired and does some of the acquisitions.

 

Does your husband work with OpenAI? If not, he has no say about the specifics of how OpenAI acquired the data, even if he knows the technicals of how it's done.

OpenAI is a corporation. And it's a rule at this point that if it's a corporation, they have done dirty somehow. That's how you survive in the business world.

AI and the human brain are two entirely different things. It takes years and years for a human to read hundreds and hundreds of books. It takes years to then decide to try and learn how to write. And then the craft itself can take more years to learn, possibly even a lifetime.

There is an entire journey of hardship, uncertainty and learning that brought me where I am today. All the short stories and books I have written were done so with tears (excuse the cliche).

But AI can learn and improve in a matter of days once training is initiated. Yet you're telling me that it's the same thing?

Therefore, no, AI is not the same as the human brain at all. There's a similarity but that's it. You as a human reading books and gaining influence by them isn't the same.

 

A professor, not having students purchase the book that goes with the class? I don't think so. If that were the case, there would not be a market for used text books. Please refrain from straw-man arguments in this debate.

 

He has and does data collection for both the contracts that he’s a team member of, as part of his employment and he also has smaller scale LLMs that he’s working/playing with privately (he’s trying to create a procedurally generated roguelike and wants to incorporate AI to make each interaction different with each play through). I could care less what bullshit you’re attempting to call. I know what he does everyday, I also know what he’s taught me.

This bandwagon bullshit is just that, bullshit. Most people who have adopted opinions around this situation have no true knowledge as to what LLMs do and what they’re capable of doing.

 

But it actually is the same in a lot of ways (never mind you danced and skipped over my other points). I'm talking about how the output is generated, not that it's different, and for that matter it does not appear to be convincingly distinct in many cases.

It's cobbling large chunks and unique particulars for the sake of mimicry; I asked chat GTP to write a detective story in the fashion of various authors. Guess what? Ayn Rand's style featured a character named John Galt. Such a coincidence.

 

Personally, I'm struggling to understand the difference between a machine reading a book or books, and producing a work based on them, and a human doing the same thing, except that the machine can process much more data than the human can and probably can't be sued.

I mean, if a human writes a Lovecraftian story involving Yog-Sothoth, and a human does the same, where's the difference?

I can buy moose's book, read it and think "I like what this guy's doing, I'm gonna write a story in the same style", and as long as I don't lift elements (characters, setting, etc.) directly from it, what's wrong?

 

Theoretically at least, the end product can be just as well-written as a human. In practice, AI doesn't yet come close to human level, but that's besides the point right now. Let's assume that is is just a good as a human.

It's a question of underlying value. If you have an author who generated a book using a machine in seconds, and you have an author that wrote the book manually over a year (and that's not even counting the years it took to learn the skills in order to write it), then which author will you respect more?

Historically, mass-produced goods using automated processes have never been as good as goods produced using manual, traditional methods. And we value this accordingly in our society. A loaf of factory bread is cheap, for example, but a sourdough loaf made by an artisan will not only be much better than the factory bread, it will be much more expensive.

Same with jewelry. You can produce fake jewelry that looks just like the real thing, but the real thing will always be valued more.

And it's for good reason. Even though you can automate the sourdough loaf making task of an artisan by replicating it perfectly using machines in theory, we haven't been able to do it.

Even if you do, the question of authenticity will always come up. Authentic has never replaced replicas historically, and there's no reason to believe it will this time, either.

 

Well, one is a machine and one is a human being. This gets to the heart of what creativity is. The human being analyzes, synthesizes, interprets, combines, and reflects. The human being comes up with new thoughts and ideas based on what it takes in.

Machines do not create in the way humans do.

 

Maybe not yet, but who's to say they won't in the future?

 

That's not necessarily true. The reason for the price differential is supply and demand. There are fewer of the hand-made items, compared to the mass produced ones. That doesn't make them intrinsically better. People value them more for their rarity, hence why many companies produce limited edition or unique items.

I agree, hand made items are often better, but just because Fred the baker made a loaf by hand, it doesn't necessarily mean it will be better than a mass-produced loaf. Fred might be a terrible baker.

But the mass produced items makes those things available to those who would otherwise not be able to have them. Anyone who wants one can buy a Bible nowadays, but before Gutenburg, they were the preserve of some very wealthy institutions and individuals. Whether that's a good or a bad thing is very much a matter of opinion.

 

Well, I asked ChatGPT. Doesn't look promising.

My prompt: Will ChatGPT ever think like a human?

ChatGPT's response:

ChatGPT and similar AI models are designed to simulate human-like conversation and generate responses that are contextually relevant and coherent. However, it's important to understand that AI models like GPT-3.5 do not have consciousness, emotions, or subjective experiences. They don't "think" in the way humans do.

These models operate based on patterns and information present in the data they were trained on. They generate responses by predicting what comes next in a given text based on the patterns they've learned. While they can produce responses that may appear human-like in some contexts, their "thoughts" are essentially the result of statistical and computational processes.

The goal of AI research is to improve the capabilities of these models to make them more useful and human-like in conversation, but they do not possess true consciousness, self-awareness, or the ability to reason and reflect in the same way humans do.

In the future, AI may become more sophisticated and capable of simulating human-like thinking processes to a greater extent, but true human-like thought remains a complex and largely uncharted frontier in the field of AI. It's important to manage expectations and be aware of the limitations of current AI systems.

 

This is the important passage. *Current* AI systems. We're only a few years into the advent of AI systems accessible to the great unwashed. Five years ago, there wasn't anything like it. I'm not at the forefront of AI research (although I was involved in a little bit of it 30 years ago, when I was a post-graduate), so I wouldn't want to predict what AI will be like in five or ten years' time. It may go nowhere, or Skynet might be about to go live.

 

Well, that's going to require something very different from what AI is now. Not just a machine that gives the predicted patterned response, but one that can give a considered response.

A brain is billions of microscopic nano-machines working in parallel. Nothing exists even close to it.

 

But it could. The computing power is there.

https://www.lesswrong.com/posts/KsKfvLx7nFBZnWtEu/no-human-brains-are-not-much-more-efficient-than-computers

 

Well, the same could be said for anything. No, we don't have the technology for a Start Trek transporter - but we could. There's not a lot of weight in such a statement.

It's not a matter of computing power, but the way in which the information is processed.

 

I just checked. No, the computing power is not there.

https://www.scienceabc.com/humans/the-human-brain-vs-supercomputers-which-one-wins.html

 

That article is a bit dated. One thing to consider, and I don't have the time right now to pull the exact figures, so I'm paraphrasing and unknown source, but mainly for the concept: humans have maybe 100 trillion connections in our brain. The current computing power and models have maybe 10 trillion 'connections' and can already do some things as well as humans can. So the concern, or promise, depending on your disposition, is what happens when the computing power and other logisitic bottlenecks allow for AIs with close to or as many connections as humans?

They are already capable of doing some things as well as us with only 10% of the connections.