There is a place near me that makes sourdough donuts and they are the best thing I have ever had.

But also sourdough helps a lot of people who have sensitivity to modern yeast, so I may not like it all the time but I support it.

You beat me to it!

I read that food you cook is less appealing because you have been smelling it while you cook and so it's not as interesting to your taste buds.

This could be bullshit but it also makes sense.

They are getting more views on their propaganda by stealing and upsetting artists and their fans. That's the strategy.

GenAI is just continuing what ipads, short form video and teaching to the test already started.

I recommend this post series on balloon juice for a more nuanced and informative non rant.

balloon-juice.com/2025/10/29/g...

Other major issues I have are
- it encourages skill atrophy
- it's outputs will always necessarily be a little bit wrong because it is a model
- there is no clear water marking or way to ensure generated content can be seperated from real content

Substack is a great example and probably one of the main sources used to train these algorithms how to generate code.

Now the website has limited traffic or engagement.

Supporters argue that this data is fair game but when this data was shared nothing like generative AI existed and few of us could have imagined that our words would permanently become part of something that might replace us.

The owners of the data used to create these models are us.

So even if we did ask them to compensate contributors to responses - they can't.

These companies are now building what will hopefully be profitable businesses off of decades of writing images and other data that's been shared on the internet - for free, without anyone's informed consent.

When's the algorithm has been trained using a set of copyrighted works those copyrighted works don't even need to be stored as part of the data set there's also no way for us to tell which works are being used.

We can tell sometimes because a style is very distinct (Gibli for example)

GenAI, requires a huge amount of data not only for the tasks allowing you to understand requests but also for the task flowing to generate responses.

In order to get that data the companies that built GenAI scraped the internet for text and images as well as whole books and copyrighted works.

Hi, I am a Data Scientist who works with it daily.

The issue that I have with AI is not that it exists or that in and of itself is immoral and it is not even related to the large economic and environmental impacts though those are a concern.

My issue is the stolen labor of humankind.

I thnk the problem is when the mission or vision is too broad.

As I have moved up in responsibility at work I find that having a clear and specific vision from leadership helps with decision making and resolving disagreements.

Also the data is more targeted so the training would be significantly less than generalized llms or multi modal models.

Training data to do one task is on a different scale from training it to do every task.

Nothing in that paper indicated to me large scale data scraping. Please cite.