Just a reminder for those that care and have access to articles, only days left to get echo logprobs from davinci – and, for example, compare logprob of the text of 100 NY times stories from 2010-2019 to stories from 2022 the AI can’t yet have been trained on or to similar pastes from other sources.
Otherwise, you’d have to put the AI in a state to complete their articles, that ends up being from a wide swath of knowledge and not rooted in continuing anything.
I think it’s against copy right law, though it’s hard to enforce. So, the word could is proper here 
Long time ago, I was a college student, I was copying several pages from a book, a librarian came up, gently reminded me that I could copy portion of it, in order to get verbatim for my analytical school paper. But I should not copy the entire book, because of copy-right.
It absolutely is. The broader point being that combining copyrighted works and tools which can faithfully reproduce those works verbatim does not automatically constitute a violation of copyright.
The courts may decide this case differently.
¯\_(ツ)_/¯
If NYT (and legacy media in general) were smart, they would be leveraging the large amounts of content/data they own and start creating their own models… or working with OpenAI to get custom ones…
Exactly. Why was suing (probably preceded by cease and desist threats) their first reaction, instead of “How can we use this technology to our benefit?”
Part of their legal fiduciarily duty is to maximise return for their shareholders, this leads to cases like this being made. It is a meritless hail marry case which they would of been in trouble for not trying.
The argument goes something like this: The future of America as a dominant super power lies in the successful execution of world class artificial intelligence systems vs a newspaper that uses the internet to gather their stories. I wonder who will win.
It is sad to see old media industries die. I’m not sure who will pay for generating content and investigative reporting that was once created by NYT. We have had the same problem with the production of top quality music and other media. A new model for supporting such things is being created, organically, right now and it still has some years to go before maturity.
IOW, I think the NYT has as much chance for survival as a canoe in a tidal wave.
Nice metaphor. I think it fits quite well. They spent so much time sticking to what worked and not innovating they were caught by the OpenAI tidal wave with their pants down and full of advertising stickers.
Going to be interesting to see who will be riding that tidal wave with success 
. My bet is independents. Reminds me of nightcrawlers.
Independent journalist who capture things live and sell them to the publishers. Except this time it will be API services that provide the content to content creators.
There is a official statement from OpenAI regarding the topic:
Summary by GPT-4:
OpenAI’s Commitment to Journalism: OpenAI expresses support for journalism, partnering with news organizations, and contests the legitimacy of The New York Times’ lawsuit.
Collaboration with News Organizations:
Fair Use and Opt-Out Option:
Addressing “Regurgitation” Issues:
Dispute with The New York Times:
Looking Forward:
For all the people wishing “old media” dies off. Remember democracy is held within boundaries by a free media. Once one group of powerful companies or government control the media you are living in Russia, n Korea or China. If you want freedom and democracy, value the free press and it’s control of the democratic process. Open ai can quote all the tenets, company policies and marketing it likes, telling us they are for the greater good, but look at Google and privacy, and “don’t be evil”, it’s all in the power of the CEO, if the CEO wants to bend or change the rules in the future, they will. Be careful what you wish for
Whether it’s through an API or the web interface of ChatGPT, I can’t imagine LLM being useful without information from traditional media.
In the international community, the truth as perceived by OpenAI is, in reality, just the opinion of a single company.
During the GPT4 release of the standalone feature update, I had the view that ChatGPT would be an AI that would never fail but would be more influential than any other AI. For now, this is still the case, partly because of its own limitations.
The emergence of ChatGPT has had a greater impact on the world than previously hypothesized. or any imagination Nowadays, there are more and more complaints about Ethic AI, but it is not pure ethics. There are many reasons hidden beneath Ethic. Even if people who oppose AI do not come from profit, Most of them come from misunderstandings and biases that do not recognize LLM, but some social problems lack the movement to deal with other problems that will arise in the long term. Adhering to the same problems, especially labor issues and producing works that violate copyright
More than a year ago, AI in the top 10 publicly available It has billions of users, but ChatGPT is frequently in the news, both directly and indirectly. Mixed with other news from the AI world because some people make a living from selling content called ChatGPT, it can be sold whether it’s good or bad. But it was found that there is content that is intentionally negative in behavior, whether it be news, including biased AI research, distortion, writing combined with other damaging content. Even major news media have produced negative content covering everything from its creation to its use in crimes. This is no different from teaching people to use it to commit crimes themselves. (If presented for benefit, there is no need to tell how to create the prompt level and some usage steps.)
And in the end that is just a thought. Enemies that must be eliminated because GPT causes loss of benefits. including power in society As long as users of GPTs are free to use the website, human control over information in society will be destroyed. such as news control Product advertising and distortion of information.
These benefits are enough to be stopped. It’s not about any data rights violations at all. Even without ChatGPT, data violations still exist. Moreover, it is not just a topic of trade and business. but also the impact on political power. To get rid of ChatGPT with Ethic AI is the right way. People in society still lack knowledge and understanding of problems and management. OpenAI itself is not very good at managing other matters. such as the occurrence of output that violates the law. It doesn’t come from practicing development alone. And it doesn’t come from specific input only. But instead, it’s just used to develop and cope.
For example, in theory, ChatGPT is currently in the Social context stage. It is considered to be the last step in human learning, with the Rewarding Model being important. Even if the input is not stored, But learning this way To add to the weight of the award action. Resulting in learning behavior Think about whether blocking external data would help. If one day there is a hot new character, GPT-DALLe, it can be created. How should I protect against prohibited input from the back of the house? In practice? In that way, having to request permission to use the data has a cost that is no different from having to pay in some way. And must do it continuously, while others have the freedom to learn and grow in this process more independently. For example, the top 10 most popular AI. ChatGPT is the only one that checks input before it is entered. This method is management. Weaknesses of the system that must be acknowledged before knowing
In the NY case, the outcome of the decision was no more important than the payment for the data. But their actions in response to this will reveal what else they can do with OpenAI.
Another solution would be to wait for bing to grow more but I don’t really want to hope for it.
