Sam Altman indicated it's impossible to create ChatGPT without copyrighted material, but a new study claims 57% of the content on the internet is A...

[Windows Central]

A suggests 57% of content published online is AI-generated, leaving AI tools with limited original content for training, ultimately leading to an increase in inaccurate responses to queries.

Sam Altman indicated it's impossible to create ChatGPT without copyrighted material, but a new study claims 57% of the content on the internet is A... : Read more

 

[Yisroel Tech]

This shock-factor headline is misleading and almost intentionally-deceptive .

57% of content published online is AI-generated

Not even close to that!

The AWS study is only about "machine translated" content and has 0 to do with generative AI and/or model collapse.

Studies researching AI generated content have found a mere 10-15% of content being AI generated. While this is still a valid concern going further, the misleading 57.1% number has no relation to this.

And don't tell me that this is only the headline... and inside in the article it does say that 57.1% is AI generated "or translated", because that doesn't help at all. First of all, then put that in the headline. Furthermore, it's not "or", that's the only thing this 57% is referring to.

I think this article should be clarified, and the AWS study removed.

 

[red_rudi]

A suggests 57% of content published online is AI-generated, leaving AI tools with limited original content for training, ultimately leading to an increase in inaccurate responses to queries.

Sam Altman indicated it's impossible to create ChatGPT without copyrighted material, but a new study claims 57% of the content on the internet is A... : Read more

What a false claim! Your first sentence talks about the usage of generative AI (like, LLMs and generative image models such as stable diffusion), but actually, the study is about machine translation - which is not generative AI, which has been around much longer - Google translate as statistical model that made it easily accessible to almost everyone is there since almost 2 decades! And also the neural network based version is there since 8 years..

So, please, rethink what you write, and stop spreading so poorly formulated claims.