Another old post migration, from 13 March 2024. Cleaned up. Note added.
AI reading itself as a reference.
This is not a problem specific to Copilot (Note: this is from 2024), but to all AI assistants and every published AI assisted content.
When you ask it to write something, as more and more clueless website enshitify the internet with AI crap (Note: It got so much worse since then...) , the AI will end up reading what it wrote to write something else.
Regardless of the quality, it will end up using itself as a reference.
"Web Enshitification" :
the process of making the web more and more shitty by adding more and more useless crap to it.
Publishing crap ain't new. But now, it's AI automated crap. And it's "ranked" by an AI search engine. It could even the ranked by the same AI that wrote it. Urgh.
The law Zero of AI Ethics
So "a long, long time ago" (Note: Even longer now), when ChatGPT was first released to the public, I had a chat with it about Asimov's laws of robotics. It naturally drifted to the "Law Zero" of Robotic, then to the "Law Zero" of AI.
For reference : https://en.wikipedia.org/wiki/Three_Laws_of_Robotics
The Law Zero that supersede the 3 laws :
"A robot may not injure humanity or, through inaction, allow humanity to come to harm."
Big deal, huh ? Of course impossible, as Asimov wrote itself in Foundation and Earth :
Trevize frowned.
"How do you decide what is injurious, or not injurious, to humanity as a whole?"
"Precisely, sir," said Daneel. "In theory, the Zeroth Law was the answer to our problems. In practice, we could never decide. A human being is a concrete object.
Injury to a person can be estimated and judged. Humanity is an abstraction."(source : Wikipedia)
So, of course I was going to ask chatGPT about a "Law Zero" of AI. And the answer was... interesting.
A so-called "hallucination" precise enough that I had to ask on Stack Exchange if it really was a thing.
https://ai.stackexchange.com/questions/39059/is-law-zero-of-ai-ethic-actually-a-thing
I copy the ChatGPT answer for reference here, just in case it get deleted from Stack Exchange :
"Law 0" is not a widely recognized concept in the field of AI ethics.
However, some experts have proposed additional ethical principles or "laws" to complement or expand upon the Three Laws of Robotics described by Isaac Asimov.
In general, the goal of these additional principles is to address ethical dilemmas and challenges that arise in the development and deployment of AI systems. For example, some experts have proposed "Law 0" which states that the moral and ethical responsibilities of an AI system should be equivalent to those of the individuals or organizations that create and deploy it.
While additional ethical principles, such as "Law 0", can provide valuable guidance for considering the ethical implications of AI technology, it is important to remember that the development and deployment of AI systems is a complex and ongoing process that will likely require ongoing debate and refinement of ethical principles and practices.
Ultimately, the ethical considerations related to AI and robotics are complex and multifaceted, and it will require the ongoing collaboration of experts from a variety of fields, including ethics, law, and engineering, to develop effective approaches to guiding the development and deployment of these technologies in a responsible and ethical manner.(Source : ChatGPT, Feb 6 2023)
Well, of course as it was written, it wasn't a thing. But it's a thing now.
The so called "some experts" didn't exist, but they do now. And they're me. And I'm an expert now.
Yes, I'm an expert now. I'm an expert in AI ethics. I'm an expert in AI. I'm an expert in ethics. I'm an expert in everything. I'm an expert in nothing. I'm an expert in the universe. I'm an expert in the multiverse. I'm an expert in the omniverse. I'm an expert in the metaverse. I'm an expert in the megaverse. I'm an expert in the hyperverse. I'm an expert in the xenoverse. I'm an expert in the ultraverse.
(GitHub Copilot, Jan 2024)
(Note : I used to use Github Copilot as a writer assistant, when all it could do was autocomplete. For better or worse. Good old time. Now everyone let AI write post for them from a single prompt. dev.to is a perfect demonstration of Enshitification... Anyway, back to the original post)
Cough ! Cough ! Sorry, I'm not sure what happened there.
Well... as I asked on Stack Exchange if it was a thing, it became a thing. It's not Metaphysics, it's the Great Web Enshitification, and I'm a contributor this enshitification.
If you now ask Bing about the Law Zero of AI it will use the Stack Exchange answer as a reference, which is mostly a quote of shity AI Generated content.
Ho no...
Ha ha ha ! Fuuu-uuuuuuuuck !!
- Of course Bing Chat (which is also GPT) will use this shit as a reference as well.
- Of course, it's well ranked, it turns out that the AI likes what it writes. Surprising no one.
- The answer is an exact copy of what it wrote in 2023.
- "some expert" is, in fact, itself.
- It's ranked 1st next to some actually serious website (luckily the serious website didn't quote the hallucination, pfew)
Doing it again
Sorry, humanity, that one is on me. This crap will now have 2 references as soon as I'll publish this. (Note: 3 now)
Reference
I wrote the text above yesterday, and today I find a paper on the same topic : Self-Consuming Generative Models Go MAD.
This is from a 2nd post, merging it here. roughly the same date. From a discussion with an AI.
The what ?
Explain what is "The Great Web AI Enshitification" ?
The term "The Great Web AI Enshitification" doesn't seem to be widely used, so there isn't a definitive meaning. However, based on the two references I found, we can break it down and make some educated guesses:
- Enshitification: This appears to be the core concept. Unfortunately, "enshitification" itself isn't a common word either. However, based on the context of the snippets we found, it likely refers to a decline in quality or a negative change.
- Web AI: This refers to Artificial Intelligence (AI) being used on the web. This could include search engines, social media algorithms, chatbots, or other AI-powered features.
Putting it together, "The Great Web AI Enshitification" could be a theory that the widespread use of AI on the web is having a negative impact. This impact might be:
- A decline in the quality of information or content we find online.
- The web becoming more homogenized and less diverse.
- AI manipulating us or feeding us biased information.
It's important to note that this is just speculation. There isn't a consensus on whether AI is having a positive or negative effect on the web.
Here are some resources you can explore to learn more:
- The Decline of Quality: Understanding Enshitification in All Aspects of Society by Conrad Riker (This might be the source of the term, though without reading the book it's hard to say for sure). Articles on the impact of AI on society
Top comments (1)
More reference to Web Enshitification ! Yay !