Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Things are about to get worse for generative AI

garymarcus.substack.com

Post reply on HN
Print view View original

691–700 of 769 posts

Re: Things are about to get worse for generative AI

#691

Post by idopmstuff »

SergeAx wrote: post #672
idopmstuff wrote: post #663

Earlier quoted context omitted.

This has always felt like the important-but-ignored distinction to me. You can definitely draw Mario! Copyright doesn't protect against you doing so. You can also use tools to recreate copyrighted materials. For example, you can use Word to type out the text of a copyrighted book. Perhaps more relevant to the AI discussion, you can use a scanner and printer to reprint copyrighted text. What you can't do is use those…

But... OpenAI is clearly profiting their $20/m on drawing Mario pictures and word-by-word reproduction of NYT articles?

On the latter example, my question is whether anyone is actually using ChatGPT to read NYT articles. My understanding is that to produce the examples of word-for-word text in their lawsuit, they had to feed GPT the first half dozen paragraphs of the article and ask it to reproduce the rest. If you can produce the first half dozen paragraphs, you already have access to the article's text. Given that, is this theoretical ability to reproduce the text actually causing financial harm?

Re: Things are about to get worse for generative AI

#692

Post by stubish »

regularfry wrote: post #468
AlphaWeaver wrote: post #288

Earlier quoted context omitted.

No legal precedent has been set as of yet. The "precedent" you describe is the argument AI companies have been using (that training their models on information available on the Internet should be considered "fair use") but whether AI training actually satisfies the four-factor test for fair use remains to be seen.

It's a null question. Training itself is neither publication nor distribution, so copyright can't be relevant at that point. "Fair use" just isn't a concept applicable to training.

Training stores a variation of the source material, which is arguably distribution. And selling the result or selling access to it certainly is. So fair use applies, and hoping a court thinks the process is transformative to count as fair use. Given original material can be spat out, my money is on a court thinking this is about as transformative as a compression algorithm.

Re: Things are about to get worse for generative AI

#693

Post by CatWChainsaw »

edgyquant wrote: post #373
block_dagger wrote: post #176

Earlier quoted context omitted.

ChatGPT is not an intelligent entity? What’s been comprehending and rewriting all my crappy code for several months? An auto-complete? There’s obviously emergent behavior there that is actually defined by the maker and most users as “intelligence.” Edit: typo

It’s not a human and so the entire argument comparing it to one is moot. It’s a program on a machine and doesn’t have rights, this anti-human way of thinking is seriously fucking scary.

As Max Tegmark said in his recent interview with Lex Fridman, a lot of the technology being developed now, and how it's being talked about and how it's used, is anti-life.

Re: Things are about to get worse for generative AI

#694

Post by tsumnia »

schneems wrote: post #587
tsumnia wrote: post #408

Earlier quoted context omitted.

> are you teaching the class or taking it? I teach it, my background is located in my profile and my research focuses on CS education. Scale and impact do matter, I wholeheartedly agree. However, I stand by my point that genAI is mirroring how humans learn - repetition of previously observed actions. As part of my dissertation, I argued that humans operate using 'templates', or previously established frameworks / sys…

I took an AI in ethics course from a state backed school (Georgia Tech) and the answer to questions that weren’t “that’s illegal based on protected status” were “well, it depends.” Which, sure, that’s true, but maybe not helpful. In my view it encouraged nihilism and apathy instead of developing ethical frameworks. From that lens, I feel teaching a course might be more limiting in the range of heuristics you’re willi…

Apologies for the delayed response, but on the bright side it's faster than I respond to some emails XD. I should preface the course I was referring to was "Intro to AI", not "Ethics in AI". I only have a single lecture dedicated to ethics, but do try to pepper it in as we cover topics. My original comments were more addressing "how humans learn" rather than any higher level ethical concerns. Your last section on "deeper intent" is correct, there wasn't any.

I have a pretty neutral stance to GenAI, mostly due to personality however it also stems from my background as well as recognizing students' interests. Prior to CS Education, my master thesis involved computer vision for catching "high valued targets", but was also funded to help minimize human trafficking. I have students in my classes that are very interested in going to work for defense companies like Lockheed and Raytheon, and I have others that are really interested in using AI for "social good" areas like healthcare and education. I try to have a neutral stance because: A) I hated the professors that I took that would use their lecture time to express their political opinions, B) opinions that are opposite to a student may otherwise discourage them from learning the material, and C) my primary focus is to make sure they learn the material and do it "right".

When I started teaching, I used the analogy that if they go on to write the software for the life support machine I'm hooked up it, it WORKS. If someone wants to go on to use AI to create weapons, I can't stop them anymore than I can force them to read a chapter or convincing the person beside on the highway to slow down. I just work to ensure they do it correctly (which includes being mindful of the ethical ramifications of using algorithm X for task Y).

What would an ethical framework for designing AI for a drone even look like? I have no idea, nor is it something I'm interested in delving into. I got out of face recognition for those reasons. Does an ethical framework for GenAI require the same elements, a fraction of them, or a completely different set of guidelines? Who gets to decide them - the 'experts' in AI, the government, society as a whole?

Personally, I've made the comment that the current opinions on regulating AI are like "everyone trying to be AI's parent". We're never going to agree because everyone has a different opinion on the "right" way to handle AI. Plus, human cognition is so unknown and illogical that we may never figure out a way to perfectly replicate human intelligence. I instead try to stay somewhat optimistic and marvel at the math we've used to create "AI".

Re: Things are about to get worse for generative AI

#695

Post by stubish »

aimor wrote: post #467

I did an interesting thing and looked at how well the Llama2 models could compress text. For example, I took the first chapter of the first Harry Potter book and recorded the index of the 'correct' predicted token. The original text, compressed with 7zip (LZMA?) to about 14kB. The Llama2 encoded indexes compressed to less than 1kB. Then, of course, I can send that 1kB file around and decode the original text. (Unless…

I wonder what the loss would be for 'translated into Finnish'? Translations between just about any human languages will contain less than 100% of the original.

Re: Things are about to get worse for generative AI

#696

Post by Teever »

Nevermark wrote: post #679
lovecg wrote: post #677

Earlier quoted context omitted.

You make a good point, there’s a difference in copyrighting exact works vs. the characters or the story that the work is made up of. I’m not arguing for removing literal copyright (though the terms should be shorter). But I think it’s fine if other people rushed to make their own Star Wars movies after it came out. Hardcore fans are pretty good at deciding what’s “canon” vs. not anyway, and the rest of us don’t care…

The only problem with this view, is: > "the rest of us don’t care as long as the work is of good quality" Copyright protects Disney. But it also protects every creative author, no matter how disadvantaged, from mass shareholder driven behemoths. Today "Disney likes your work" is ear music. Without copyright it would be a death nell.

So how can we modify copyright so that it protects the little guy more than it protects Disney?

Re: Things are about to get worse for generative AI

#697

Post by bkudria »

lovecg wrote: post #589
jerf wrote: post #561

Earlier quoted context omitted.

"Who truly owns the tales of Snow White and Cinderella?" If you want to make your point, you need to choose something that isn't already public domain. Disney already only owns their own interpretations, and, arguably, whatever penumbric emanation they can convince a court is stealing from them, but it still certainly isn't the entire space of Snow White and Cinderella stories. There is some fairly recent stuff being…

I think the world would be completely fine without a copyrighted C3PO or Robocop. George Lucas didn’t have billions of merchandising revenue in mind when working on his wild and thought unlikely to be successful science fiction movie in the 70s. Robocop was also a labor of love. We don’t really need Nth Star Wars sequel powered by those extra profits. The art form could be healthier overall.

> George Lucas didn’t have billions of merchandising revenue in mind

From https://www.newyorker.com/magazine/1997/01/06/why-is-the-for...

> Lucas’s most significant business decision—one that seemed laughable to the Fox executives at the time—was to forgo his option to receive an additional five-hundred-thousand-dollar fee from Fox for directing “Star Wars” and to take the merchandising and sequel rights instead.

Re: Things are about to get worse for generative AI

#698

Post by SergeAx »

idopmstuff wrote: post #691
SergeAx wrote: post #672

Earlier quoted context omitted.

But... OpenAI is clearly profiting their $20/m on drawing Mario pictures and word-by-word reproduction of NYT articles?

On the latter example, my question is whether anyone is actually using ChatGPT to read NYT articles. My understanding is that to produce the examples of word-for-word text in their lawsuit, they had to feed GPT the first half dozen paragraphs of the article and ask it to reproduce the rest. If you can produce the first half dozen paragraphs, you already have access to the article's text. Given that, is this theoretic…

I think it would be quite enough to prompt OpenAI with article title and author name. This is how LLMs are working.

Re: Things are about to get worse for generative AI

#700

Post by kayodelycaon »

Snow_Falls wrote: post #681
kayodelycaon wrote: post #674

Earlier quoted context omitted.

The current length is indefensible to me. There is zero justification for it beyond pure profit. I’d prefer it be lifetime of original creator, only for the original creator. All other cases (transfer of copyright, death of the creator, etc) would have a maximum of 20 years. Alternatively, I’d prefer 30 or 40 years, but I would grudgingly accept 20. :)

See, I actuall disagree, o think it should be a set time, regardless of authors death and inheritable. I think there would be a lot of things not made near the end of the creators life if they didn't think their family would be supported by it. Hence, why I think it should be a fixed length. Also I think it could he shorter than 20 years, I picked that just because its the length of patents. IMO, if you haven't profi…

I may not have explained my first idea correctly, it’s lifetime or 20 years, whichever is longer.

As for 40 years, it allows time for a given work to solidify in history. Multiple generations will have grown. If the work is relevant to multiple generations 40 years after creation, then it should belong to everyone. Twenty years is too short. It would be available to people who grew up with the first printing and to whom it is still relevant.

Basically copyright should be long enough for most things to become firmly irrelevant, outlive their usefulness, or made their changes to society before entering public domain.

A lot of people who oppose copyright would not like this because it intentionally keeps the stuff they want to use out of their reach until they would no longer want it.

For me, the purpose of public domain is allowing what has been important in our cultural history to be for everyone, not locked behind a rent-seeking corporation. But I also want authors to be able to keep control over their own works as long as it remains relevant to them.

Post reply on HN