AI Learning: Copyright Infringement or Fair Use?

[@joeparkhurst57]

I am a singer. I don't live in the past, but I remember the past. I turned down two small record deals (shop deals where a successful small label produces you and shops you to the BGM's or SONY) waiting for the right fit. Then Napster happened. Then sampling happened. Now, A.I. can write in the style of specific writers. Oh well. LoL - this is the parable of "It's okay, they're taking the neighbors, then one day - they take you." lol lol lol

[@MrGrumblier]

I think the reality of the AI writing will be a sharp decrease in new authors getting published. As with the film industry, publishers will find it far more profitable to get AI to write in the style and universes of established writers with an established following than to gamble on new writers. With AI, it will no longer matter if a popular author retires or dies, AI will be able to carry on in their stead. It would be interesting to see what an AI would have done with the Wheel of Time series, for example, and how that would compare with Brandon Sanderson's completion.

[@Prof_C]

Yes, I wonder what the incentive is to have writers create. I have heard the term "minimum viable content" used for ad-supported TV, in a new context of what is the minimum content that will keep people watching. Will we have "minimum readable content"?

[@drdubious2432]

Once AGI is fully operational, few paying jobs will be left, so copyright issues will become irrelevant. No one will have money to pay for anything.

[@AdamWeatherall]

Money doesn’t exist in the form if modern currency, only gold and like substances are actual “money”, they will print whatever (debt notes) they need to keep the facade going. The kinetic value of money and that value in terms of population has been insignificant for more than a decade, corporations and governments are 80% of all kinetic movement of the flow of money and therefore of value since 2012.

[@guycwilson]

Why is it an either/or situation? We should be asking questions about how and why this is being done, whether it constitutes plagiarism or other illegal or unethical use of the work of others. We should also be asking serious questions about our entire intellectual property framework, the nature of creativity, and about plagiarism. I would contend that the present system serves the interests of corporations and a few highly successful creators, is damaging to society, stymies our understanding of the evolution of creativity in an age of continuous change, and is not fit for this century.

[@Prof_C]

Agreed.

[@cloudedmund8444]

simple. AI should not have rights to it's work. therefore no one should have the rights to publish AI's work as well.

[@Prof_C]

Interesting. So far, the copyright office has said that you can't copyright a work that is AI generated.

[@rrmackay]

All data is the same to an LLM, there is no good or bad data there is only data.

[@jpaulnatale]

Not a grey area. It is plagiarism. AI should at least site sources

[@MrGrumblier]

Citing sources for AI is easy: everything.

[@Prof_C]

@@MrGrumblier 🤣

[@kokbasE]

This video doesn't cover anything about the dark web at all. Very misleading title. People whining about authors whining about LLMs maybe plagiarizing a little bit is getting played out. Came here because I thought you were going to discuss the possibility that LLMs have been trained with data not only containing copyrighted works, but also using more nefarious/illicit text from the actual dark web. That would have been a far more interesting video.

[@Prof_C]

Point taken. I used a GPT for generating youtube titles to make this title and it did make it very clickbaity. I revised the title and thumbnail by using Google Bard's KZbin Extension and asking it "@KZbin Create a new title for this video: kzbin.info/www/bejne/mYepp6dupLB3jZo" . If you are interested in Dark Web AIs, check out FraudGPT and WormGPT

[@ashtonal]

Copyright is dead!?

[@MrGrumblier]

It has been dead for quite a while. The people making money from it are just coming to realize just how dead it is.

[@Prof_C]

I don't think so, but I think the scale and speed by which LLMs can generate art that is similar or (perhaps) derivative, is causing alarm

[@CapnSnackbeard]

You are here first and foremost to legitimize the Capitalist system by which art is commodified. It is axiomatic to your position, which means to accept that AI is stealing, one must first first accept that art can be owned.

[@Prof_C]

Well, the writers of the US constitution copy and pasted the copyright clause from the Statue of Anne (England), and author's have enjoyed the rights to their works for a limited amount of time since (I think it started at 12 years, but now is an insane 75 years plus the lifetime of the author). As for myself, I release about everything I do under a creative common's license so people can take and reuse/remixe as they like.

[@Rivanni1]

If people dont want their work used they should not publish it.. what's wrong with people painting or writing like thier favorites or use an ai to do the same? People should be honored that thier works were used to train thier LLMs.

[@TychonAchae2]

If people don't want their cars stolen, they shouldn't park it on the street. You should be honored if your car is stolen or used by other people without your permission.

[@MrGrumblier]

@@TychonAchae2 Bad analogy. It is more like if people don't want others to have the same custom paint job, they shouldn't paint their cars. Works that are used as the basis for another work don't deprive the original author of the original work.

[@saturn45092]

​@@MrGrumblier That an interesting take. But AI output do outshadow the original work by the inf output of the neural network and negates its value and, therefore, destroys incentive to continue creating. I do not want to see future where people are working for free to train models of some random owner and he will make profit from them for free. The core of copyright protection was to support the generation of new ideas. I think the only way to make it fair - give copyright protection of the neural network output to the owners of training data. That is how AI models will not parasite on human labour but coexist. Even if you will dispute my statement that current state of ML is parasiting, anyway the "novelty" of neural network output (books, pictures, documents etc) has no value if anyone can generate equal variation of it without any skill involved and, thus, avoid copyright. But in case of custom neural networks, which trained on original materials of a specific person and its output belongs to the same specific person - we will move away from AI anarchy to a more orderly reward for labor. We need regulation and lycensing.