AI companies seems to have violated many copyright laws, at least the lawsuits claim. The silicon valley tech companies liked IP when it is beneficial to them and not when they violate it.
If you remove the IP laws, suddenly the pharma companies valuation nosedives. Many business models are build on those IP laws. It will be interesting to see what happens to them…
IP laws are part of the exceptionalism of our country. Without them, there is no incentive to work hard. What has Cuba invented? Russia? any socialist or communist country in history. In 2000, I had a friend move to San Franciso, to work for a startup. He wanted to get rich, worked 60-70 weeks, plowed money back into shares. In this case, it did not work out. But without IP, no one would take that chance. Jack Dorkey’s company would not exist. Think he does not have an patents.
The US Constitution grants Congress the power to enact laws concerning intellectual property (IP) under the Intellectual Property Clause, also known as the Patent and Copyright Clause. This clause, Article I, Section 8, Clause 8, specifically empowers Congress to “promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries”.
I understand where you are coming from, to an extent agree. Now, India sent a mission to mars for $73 million, very cheap price compared to the cost incurred in the west. The scientists who worked on that project are government employees, who worked for a government employee paycheck and the scientists didn’t even receive bonus!!!
Having said that, I understand if you are an artist, you want to be economically benefit from your creative work, or a company that invests lots of capital wants to have protection to receive economical benefits of their investments.
Do not confuse protecting economic interests with “exceptionalism”. Separately, it is not even clear what is “US exceptionalism is”.
When I first heard this I scratched my head, it did not make sense. Copyright does not protect content, it only protects the form, the expression of that content. Trademarks, patents, and industrial secrets protect content qua content.
Large language models read papers, like humans read papers, and by predicting the next word large language make sense of the text. When humans extract learning from texts they are not prohibited from using that learning, they are prohibited from regurgitating the text in its original form.
There are degrees of infringing. In writing entire sentences get copied occasionally by AI.
Fragments and who gets paid is the jist of it.
Mostly AI is not infringing as far as the output goes.
At some point if I use the color blue in art I am not in any way infringing on anyone else. If you use the words “the goods” to write something you are not infringing.
You have to come a lot closer with the output.
Much of it is poor artists or writers looking for a payday and possibly getting minor royalties. That might be fair.
That the AI companies are profiting from their protected works w/o compensation and w/o permission. They have a case and I am in agreement with them. What companies have done with training their LLMs is unethical and almost certainly illegal.
Back when I was an amateur semi-pro motorsports photographer I could restrict people from using my images for commercial purposes. That is what is happening here. It is no different.
That don’t wash. Either legal or illegal, no in between. Copyright has clear rules.
Suppose I read a book, learn something, make a million bucks, do I owe the author, the editor, or the publisher anything? Or just the price of the book?
When you post on the internet with free access you have put the work in the public domain. You still have copyright protection.
What I see in the near future is an update of copyright law to take into account AI LLMs.
Consider stock market prices, you are allowed to see the current price but you are not allowed to download historic prices. Or option pricing, you are allowed by the CBOE to download options chains manually but not with robots. AI could be considered robots.
I had to come back to examine the logic of this claim.
What is the purpose of copyright law, of all IP laws? To protect the economic rights of the owner/creator of the property in question. These laws do not exist in nature, if they did life would be impossible, hunting would be illegal, harvesting would be illegal, both cause harm to some life forms to benefit other life forms. IP laws belong to the realm of morality.
How does progress happen? Newton put it succinctly, using the giants metaphor, “By improving on what others before us accomplished.” If IP laws block the transfer of knowledge progress at human level ends, civilization ends. The only progress mechanism that remains is natural selection, the survival of useful random mutations.
IP laws have a conundrum, how to protect creators without endangering the future of civilization. Nature decided that pray have no absolute rights. Morals decided that nature is wrong, humans have god given rights. Quite a conundrum!
In conclusion, the only obligation of the IP consumer is not to harm the IP creator, to give the creator what is contractually specified and no more. If progress requires it, we can update our IP laws.
OK, but the purpose of IP laws is to compensate the originator of the information/invention for some limited period of time for that innovation.
AI has been trained on copywritten material since the beginning (hence the lawsuits by the Times and others for appropriation without compensation). If the AI were trained solely on, say, books on which the copyright has expired (“David Copperfield”) that would be one thing. When it is trained on the output of the Times, Science Journals, etc. merely because they are available on the web, that is per se “infringement”.
They are asking for either exclusion from, or compensation for the use of their IP. That seems entirely reasonable.
Likewise, if I create a superhero, say “Spider-Man” (and it is still under protection) then AI should not be allowed to use it to create memes or stories or videos without compensation to the originator. Do we really want amateur AI movies of James Bond or Jason Bourne flooding the market?
IP laws exist for a reason. You could argue about “length of protection” or other details within those rules, but to say “just do anything you want, merchandise and sell it and give nothing back” seems a recipe for confusion, dilution, and muddle.
Most of the books I have read had valid copyright. Did I break the law by learning from them, by quoting from them, by benefitting from them? There are standards of proof, none of which have appeared in this thread to support the claim. Show me proof and I’ll become a believer.
I did mention that copyright law might need reform.
I commented on that extensively above. Did you miss it?
No, but somebody somewhere PAID for those books. Even if you read them free in the library, the library paid to have a copy in its stacks. That is unlike AI scraping the text from the web without compensation.
Without irony, I asked Google if AI uses copyrighted material without compensation. It’s AI responded:
Yes, AI models are often trained on vast datasets that include copyrighted material, and this use typically occurs without permission or compensation to copyright holders. This raises significant legal and ethical questions about whether this constitutes copyright infringement. While some AI companies argue that their use is "fair use," copyright holders are increasingly suing, claiming unauthorized use and seeking compensation.
Technically, by buying the book, you paid a license to use the information in the book. Some years ago, publishers tried to shut down the used book trade in a lawsuit contending that only the original purchaser had paid for the right to use the information in the book. The claim was that reselling a used book violated the copyright by the second owner not paying his tithe to the publisher, to access the information in the book.
Some years ago, stereo system component CD recorders were a thing. Those recorders could only record on a specific type of CD-R, that was very expensive, as it included a royalty charge, that would be divided among the record producing companies. Jack Valenti tried to have the same sort of charge applied to all video cassettes, contending that the cassettes would be used to copy copyrighted material. But, the large numbers of video cameras being sold to private users, to produce their own recordings, blew a hole in that argument.
But why? As a copyright holder I get to dictate what can and cannot be done with my protected work. And people cannot just use those works w/o proper permission for any use they want.
Actually the copyright law states the LLM products are not patentable. Because they are machine made not human made. The human being only chooses to broadcast the product. There is no process for the human being to create anything make all products public domain by law.
Last I checked the AI guys were claiming the machines were know it alls. You’d have to deny that to give a human being a patent on something s/he did not create.
If I go to my refrigerator and pull the milk carton out of it, I did not create a milk carton.