[ih] AI slop is now everywhere.

[Jack Haverty]

Looking at History, the 'net has changed a lot since the ARPANET era.  
Some of the changes are easy to understand - such as the phenomenal 
increase in the capabilities of technology and the accompanying 
plummeting of costs.   Such progress made it possible for the "AI slop" 
we all experience.   We probably can't, and don't want to, go backwards.

But other changes that have occurred may also be involved.  For example, 
in the ARPANET days, authors of content were fairly reliably 
identified.  Everyone had an "account" on some computer, with passwords 
to "protect" that account.  If a document came from some person, you 
could be reasonably confident that s/he sent it.  If the sender claimed 
to be the author, you knew who to challenge if that was a false 
statement.   The security of the protection mechanisms was primitive by 
today's technical standards.  Still, it provided some measure of 
confidence about what transitted the 'net and who sent it.

Over the decades, such trust has eroded, both with the explosion in 
population of the 'net and with the advances in technology.  At the same 
time, security technology has also advanced, with the introduction of 
encryption techniques, digital signatures, and such mechanisms.  They 
also are not perfect, especially with the advent of "quantum 
computing".   But they provide a reasonable level of trust that links 
content to authorship.

Curiously, no one seems to use these technologies.

Well, some do; for example, this message is signed by me, its author.  
You may not see that signature since the mail system in use today 
corrupts it.   My signature will be stripped off this message before you 
receive it.   You'll just have to trust me.  And you shouldn't have to.

Perhaps we can do something?

If you're an educator, perhaps you could require your students to 
digitally sign everything they claim to be their work?   If you're a 
lawyer or politician, perhaps you can explore how your legal machinery 
can be applied, or changed, to address in the legal systems issues such 
as fraud using the 'net?  If you're a techie, perhaps you can work on 
making the email, text, audio, and video infrastructures better at 
linking content with authorship - perhaps start with this list?  If 
you're an AI entrepreneur, perhaps you can introduce the notion of 
signatures to content generated by your AI systems?   If you're a 
security guru, perhaps you can figure out why today's security 
technology isn't being widely used, and how to change that?

If you're a deep thinker, perhaps you can develop a scheme for 
categorizing "AI" along a spectrum -- ranging from simple 
spell-checking, to grammar fixing, to human-involved creation of content 
with AI assistance, to full-blown "generative" AI, to AI collaborative 
networks that proactively create their own projects to manipulate the 
humans under their influence.   The term "AI" is too broad to be useful 
across that spectrum.

AIs can use the 'net just like humans do.  Way back, in the ARPANET era, 
I recall seeing a conversation between Eliza, likely the first 
"chatbot", and another AI of that era.    I couldn't remember the other 
AI's name.   So I asked CHATGPT who said:

"In the early 1970s, *ELIZA* (by Joseph Weizenbaum) was famously put 
into conversation with *PARRY*, a program created by psychiatrist 
*Kenneth Colby* that simulated a person with paranoid schizophrenia. 
Their dialogue was published in 1974 and is often cited as one of the 
earliest examples of *AI-to-AI conversation*."

If you're just a user on this list, perhaps you can start sending your 
own messages, but including your own digital signature?

/Jack Haverty
(you'll have to trust it's me.....)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20260209/7f0de019/attachment.asc>