Per AxbomSep 2, 2023
AI won't steal your job.
A person using AI to fake your skills will.
Then the AI will cause them to make a devastating mistake.
And they won't know how fix it.
That's when you get your job back.
Remember to request a higher salary.
Show threadcͨ oͦ N ᴛⷮ rͬ aͣSep 2, 2023
@axbom in the hands of someone who knows how to use them as a collaborative agent can do some pretty astonishing things. ChatGPT and I learned about and we're creating polymorphic code within a couple hours of learning it existed. Anyone who uses it to generate and simply Ctrl+C Ctrl+V was doomed to fail anyway.
Show threadEyes Wide OpenSep 2, 2023

@c0n7ra @axbom yeah. It can do amazing things after harvesting the data of millions of people that put in labour and work and never consented to having their data used like that.

Happy you are on top of the pyramid and get to use that harvesting tool to keep your position. Good on you.

Show threadThe OwlChemist

@WideEyes @c0n7ra @axbom “never consented to having their data used like that.” I’d be really interested to see what the truth of that is. Maybe 1 in 1000 is very concerned with their data privacy, and 999 out of 1000 click through prompts and accept whatever terms pop up.

Show me exact evidence of a major AI using stolen data that was 100% not shared with or bought from a company like FB/Twitter/Google/Amazon that openly harvests and sells.

Sep 3, 2023 at 2:06amWeb
Show threadPer AxbomSep 3, 2023
@owlchemist @WideEyes @c0n7ra

Not sure anyone has to show evidence when OpenAI themselves openly admit to using copyrighted work?

What they are doing is claiming that this falls under the American Fair Use doctrine. The many ongoing court cases will tell us more, although I expect it can take 5-10 years for many of those cases to be finalised.

When it comes to image generators there have been many instances of copyright symbols being generated, as well as the Getty Images watermark. An image, or text, of course does not need the copyright symbol to be considered copyrighted, but it is the type of evidence that gives certainty to the supposition.

Researchers have also shown that chapters of many copyrighted books, such as Harry Potter, can be reproduced *almost verbatim using ChatGPT.

https://www.theregister.com/AMP/2023/05/03/openai_chatgpt_copyright/
OpenAI's ChatGPT may face a copyright quagmire after 'memorizing' these books

This top-drawer AI tech has a major science-fiction habit

The Register
Show threadcͨ oͦ N ᴛⷮ rͬ aͣSep 3, 2023
@axbom @owlchemist @WideEyes that brings a strange question to my mind. Is training an AI model akin to a human learning? Meaning that just bc something is copyrighted doesn't exclude it from taught to people. Obviously the issue is the $ being made by ChatGPT but the thought still intrigues me. Im leaning toward the idea that if your data is copyrighted and you wish to exercise complete control of it you must be obligated to somehow disable that data from be scraped from the web.
Show threadPer AxbomSep 3, 2023
@c0n7ra @owlchemist @WideEyes

Scraping from physical books is almost as easy now with immediate OCR. I take photos with my iPhone to copy text. By extension, the way to copyright a work then is to hide it. And nobody can buy your photos for use in their online articles....

But I agree, all of these things are historically social constructs by humans to protect something referrred to as intellectual property, and it's an interesting thought experiment to consider if we have to reinvent what author ownership even is anymore.

I've been blogging since 1997 and of course have always been aware that my content can pop up most anywhere. The way for me to demonstrate ownership has been to actually publish online to create a visible trail of references with dates. That also means I haven't expected monetary compensation for my writing though - which writers tend to do.

Still, as an advocate of ethics I am fond of the idea of consent, and am less impresssed when others make money by using my content without my prior approval. These days I use a creative commons license. If OpenAI were to follow that, their tool would have to apply the same license when ingesting my content. But all this mostly goes to show that licensing carries very little weight in the real world. Power and social standing does.
Show threademberquill Sep 3, 2023
@owlchemist @WideEyes @c0n7ra @axbom "1 in 1000" is still millions of people, by the way.