LLM Hype (Chatbots Hyped and Wrongly Characterised as "Artificial Intelligence") Cause Net Inflation
Net as in Internet, not limited to the Web*
If you drown out the signal with noise, then eventually all you'll end up with is noise (first a little noise that can be filtered out of view, then majority noise, then an impractical search for a needle in a noise-stack).
The US (to use an analogy) cannot just print its way out of crisis. Minting more dollar bills/coins (as long as it's a de facto universal currency) does not create real value and borrowing more and more money (usually from foreign lenders who then exercise control over the borrowers) is basically like kicking the can down the road to delay consequences. The consequences get worse over time. The borrower acts against itself (to appease lenders).
On the Web (or any other protocol), adding more and more words (or generating images by fusing/joining together existing work) does not create real value, it just lowers the quality of everything, on average. Brute force alone does not create value if done the 'cryptocurrency' way - basically a glorified NFT in which even heads of state now participate (because they receive bribes from 'cryptocurrency' bros).
Remember FTX? That didn't end well for savers, did it?
In the case of the Net (not just the Web, some people put LLM slop in E-mail), if there's a large deluge of machine-generated garbage, we'll lose our ability to speak, think, or process accurate information. That's partly why we worry about sites spewing out fake 'articles' about "Linux", "security", or both. We limit ourselves, scope-wise, to a quantity we can keep abreast of, but the problem is vastly bigger and it gets worse as time goes by or as hype goes on.
In order to avoid a collapse (of human knowledge) we all need to campaign against "chatbots" misused as something other than a toy. There's no "AI arms race", that's just an excuse for seeking more bailouts, as we saw last month (Scam Altman trying to play with other people's money).
Chatbots aren't new; image fusion isn't new either (I saw that 2 decades ago). Only the hype is new. It contributes to bad behaviour online. It ruins what used to be not perfect but at least acceptable. SEO gamers already ruin search engines, which became little but propaganda disguised as discoverability.
An associate reminds us that chatbots cannot produce news either, they cannot even summarise news, at that. Readers of ours can see last week's Daily Links on "summarize vs shorten". Things falsely marketed as LLM summaries (even by Mozilla now) are basically just culling information, without any comprehension of the text they parse and then shorten. Likewise, regarding things such as the "AI-generated" (not!) videos that we explained last month and the image fusion comment made above, those create nothing new, they just "borrow" from many places and then do some 'fusion' (like a surgical but digital embedment). This is a topic which could be expanded upon in the future, as we mostly focus on text today. But again, image fusion isn't novel, it's just some online 'cargo cults' playing with endless budgets (i.e. money to be lost). As Cory Doctorow explained it about a year ago: "Like Uber, the massive investor subsidies for AI have produced a sugar high of temporarily satisfied users. Fooling around feeding prompts to an image generator or a large language model can be fun, and playful communities have sprung up around these subsidized, free-to-use tools (less savory communities have also come together to produce nonconsensual pornography, fraud materials, and hoaxes)."
Don't even obsess over the edge cases, such as "nonconsensual pornography" (above). Even just seemingly innocent "AI" articles can do a lot of harm. BetaNews uses both LLM slop and image fusion (slop). It competes with real writers using pure junk that's often cleverly plagiarised (based on the original authors' texts). █
_____
* A lot of people conflate the two and cannot distinguish between a Web browser and "the Internet"; even E-mail is "webmail" to a lot of people, i.e. just "some Web site" with a "webapp". An associate said that "maybe [we can] list some of the non-WWW protocols by name and mention that there are many more, all of which help comprise the Internet." Gemini Protocol is one of these and as far as we can see it has not yet been targeted by LLM slop, maybe only LLM harvesting (someone wrote about this suspicion last week).