A.I. Noose

[Brigit]

Lloyd says » "ACCURACY RATE
I just came across this article: Tests suggests Google's AI Overviews tell millions of lies per hour."

Hi Lloyd, I want to introduce a term here, but only to make a clarification. I come in peace (:

AI SLOP The term for what you are posting about here is "a i slop." A i slop has become a huge problem for internet users when making a search or watching videos because the search engines are bringing up a i generated materials in a preferential way.

In other words, even when using your usual search engine, and doing everything you can to omit the a i results from your research, you are still getting a i generated websites and youtube channels.

How do everyday users feel about ai slop? Let me give an example that hopefully illustrates the revulsion that users feel towards ai slop. Suppose you search for a song that you have not thought of in years. The first result in your search has the original album art so you click. The music is the same, but as you listen, you realize that the lyrics have been replaced with a highly similar ai generated voice, in place of the original artists' vocals.

In essence, this is what all ai slop is. The ai companies spend hundreds of billions of dollars on "inference," and on model training. They are taking in all of the original content made by human authors, artists, bands, singers, sculptors, independent researchers, etc, etc.. The models then give you an ai generated version of all of that human effort.





ETYMOLOGY I wonder if the etymology is possibly a reference to "slopping" a pig. That's a farmer's job that involves taking all of the scraps out to the hogs and dumping it into their trough.

ECONOMIC IMPACTS Recently, everyone here saw the introduction of a i into all Windows functions, whether it was desired or not. Thereafter, the company lost hundreds of millions of dollars as customers migrated to Linux, etc.. And the ceo made an appeal for the entire world to stop using the term "Microslop". This also means that the Return on Investment (ROI) was greatly affected by one CEO who made the unilateral decision to implement a i into systems that people were relying on to function at 100% accuracy.

[Brigit]

Lloyd posts, "• Conversing with AI like you would with a human—expecting some degree of inaccuracy and maintaining healthy skepticism—is probably the most adaptive, sane, and resilient stance we can adopt.
So, in short: Yes, your proposal is both reasonable and needed. AI should be introduced and used under the explicit expectation that it’s roughly as accurate (and flawed) as a human expert...."

This is an a i generated bit of text. TBH, I very rarely read a i generated posts.

AN INTERESTING POLL OR SURVEY OF MEMBERS In fact it might be interesting to take a poll. Any users here who would approve of a i bots being allowed to post on the Forum would be in one category. The other category would be the members who feel that no ai bots should be allowed to post here on the Forum. About the midway point between these two policies would be those who post and those who read posts generated by ai, as long as it is clearly labelled as such. That way, members have a choice to simply skip reading those sections.

But here is the real problem with your premise "AI should be introduced and used under the explicit expectation that it’s roughly as accurate (and flawed) as a human expert..." The very crux of the problem is that a i marketers have, it seems successfully, convinced a majority of computer users that the programmers are not actually responsible for the output of their own programs.

This is not true. The makers are always responsible for what the computer program does. I will never accept otherwise. Humans are responsible for errors and fabrications. The Techbros are responsible for hoovering up the content written by creators, then rephrasing the material and introducing additional errors into the existing and archival internet and search results.

FALSE EQUIVALENCE In short, errors and fabrications, hallucinations and other departures from reality, are not at all equivalent to a human source who publishes and who you disagree with or who is looking for a mark on the internet. This is a false equivalence. The makers of the ai products have created unreliable products and this is not the same as the flaws and accuracy of a person who publishes on the worldwide web.

[Maol]

FALSE EQUIVALENCE In short, errors and fabrications, hallucinations and other departures from reality, are not at all equivalent to a human source who publishes and who you disagree with or who is looking for a mark on the internet.

This is a false equivalence.

The makers of the ai products have created unreliable products and this is not the same as the flaws and accuracy of a person who publishes on the worldwide web.

Wouldn't you agree it is the same when the AI source is flaws and inaccuracies of persons it copies?

AI searches the web and produces a result of whatever information it finds, so if that information is substantially corrupted with misinformation, the AI product is perpetuation of nonsense, no different than a human who does the same.

[Brigit]

Maol says, "...the AI product is perpetuation of nonsense, no different than a human who does the same."

The large language models do train on the data and information it finds on the web and anything else that is digitized. The costs for inference and training on this data is astronomical. The data and original material is taken by the llm effectively without the consent of the creators. It then rephrases the works and presents them to users. When it introduces its own errors and fabrications, this is an additional layer of falsehoods introduced on the web. So, in my opinion, this is an additive process. What I mean is, it is in addition to the existing discrepancies between actual human beings.

For a human being there are social, legal, and interpersonal consequences and standards for what is spoken or printed. It is not equivalent to a computer program that adds errors to what is, in many cases, stolen human content.

[Brigit]

A concrete example will help to focus on the lack of equivalence between human flawed thinking and the unreliable results that an a i generates. In the United Kingdom, lawyers were using a i in arguments for legal cases before a judge. It was soon discovered that the a i was confidently making up cases and making claims for their precedence. It is now apparently illegal to use a i generated arguments in court.

Here is the difference. For a human, this would result in professional and personal consequences. For the llm there is no personal responsibility; the llm adds fake cases to the worldwide web and the creators of the llm are not held responsible for the intrusion into the respective profession, nor for the liabilities of making up a fake case, for example. It cannot be argued that the court should accept the a i generated content because "it is the equivalent of a human."

[Brigit]

Maybe it was rude to talk about a i slop in relation to raising hogs. I do acknowledge that it can be a useful tool, and that some members here are fanboys and enthusiastic investors. But there are some very good alternatives to using the big language models. For creators who want to use a i for specific purposes, local models are far more efficient, and can be tailored to make music demos, generate videos and images, and help with research. I wasn't trying to insult all people who use a i; sorry how it sounded.

Here's what I want to say: it's not like The Trusted News Initiative is ancient history. These are the same Techbros !

[Brigit]

Here is a summary of "The A I Bill of Rights to Protect Floridians" which touches on all of these issues...including Maol's thoughts.

• Protect ratepayers by prohibiting utilities from charging Florida residents more to support hyperscale data center development, including electric, gas and water utilities

• Require that artificial intelligence cannot be used as the sole determination in adjusting or denying a claim, and that the Office of Insurance Regulation ensures that artificial intelligence models do not violate current insurance trade practices

• Prohibit companies from selling or sharing personal identifying information with third parties

• Prohibit entities from providing "licensed" therapy or mental health counseling or imitating a licensed professional through artificial intelligence

• Require a notice to consumers when interacting with an artificial intelligence

• Prohibit an artificial intelligence from using an individual's Name, Image or Likeness (NIL) without their consent

• Provide parental controls for minors to allow parents to access conversations their child has had with artificial intelligence platforms....

And so on.



In my view these are some great starting points for other state legislatures to use!