(precursor: imma be saying “I” a lot in this post. yes I mean a lot of these observations from a personal perspective, but I (haha) hope that it is also clear that I don’t mean them from only that)

ha ha, only serious. What if none of this is new?

indeed I agree a lot of it is not. the method may vary but the motivation/philosophy does not

however I do believe (and, hell, this is why I’m posting) that there is value in differentiating (ed: non-mathematic reference) in the details

What if none of this is new?

If this is a dealbreaker today, then it should have been a dealbreaker

in concrete terms, I agree. it also largely squares with when google started being offensively useless/less-good (ime it varied by domain (which post-hoc I think also got impacted by search-eng product dev decisions? supposition, never tried to trace. don’t think I ever will), affecting different ones at different timepoints)

then it shouldn’t be acceptable from DuckDuckGo, which has the same page-one results including an AI summary and panels

every time I see this shit pop up from DDG (or, similarly, in other contexts (e.g. AWS)) I every so “give it a test” and when it fucks up I send feedback of “please for the love of god stop forcing this shit on people” (<-- actual quote (sometimes more detail is added))

If summaries are unacceptable from Gemini, which has handily topped the leaderboards for weeks, then it’s not acceptable using models from any other vendor, including Alibaba, High-Flyer, Meta, Microsoft, or Twitter

exactly correct, and an entirely succinct explanation of a lot of at least some discomfort/rejection of these systems. there is a lot of detail and nuance into when/where/why people reject things built/relying on those systems, and I don’t want to get sidetracked on those here (not least because lemmy’s probably abysmal at margins), but they exist and I think it’s well worth engaging with all those communities wrt the substantive parts of their nope.gifs

If machine learning isn’t acceptable in collating search results today

this is almost a false equivalence imo (and I’m somewhat surprised to see you make the statement). 1) (speaking broadly) at the risk of being of being presumptuous (wrt the diverse viewpoints held by many others in the community here), I don’t really think a lot of people (here) would be ones going “ML==AI”? in fact, I feel like a number of the people would be ones (like myself) specifically trying to delineate between these. 2) “in collating” is a very specific subphrasing (and again I’m somewhat surprised to see you use it)

if fake, hallucinated, confabulated, or synthetic search results are ruining the Web today, then they were ruining the Web over two decades ago and have not lessened since

“yesssssssss… but”

there’s a very, very, very long conversation that is to be had here. and, hell, one of my perpetually-promised posts (yes I know) is something that touches on this

remind me later to get into a full rant about point-by-point examples of how continually-encroaching synthetic-media situations have dovetailed with a coinciding devolution in critical thought and detailed coverage. (def later tho: it features at least 3 side rants, and it takes a lot out of me)

The issue is sheer data; ever since about 1991, before the Web existed, there has been too much data available on the Internet to search exhaustively and quickly

“yesssssssss… but”

again, I think a notable substantive point of differentiation (still not math) here is the particulars of the endeavour. the “how” and the “why” of responding to user queries is, under LLM world, substantively notably different to what it was under “the previous mode” (and yes I know it’s progressive and there’s detail here too, but I hope you can see “2010 goog” vs “2025 gemini goog” easily enough without elucidation)

The problem is recursive: when a user queries a popular search engine their results are populated by multiple different searchers using different techniques to learn what is relevant, because no one search strategy works at scale for most users asking most things.

(okay I never actually dug into the SE biz, but you’ve given me a thing to read about ty)

I’m not saying this to defend Google but to steer y’all away from uncanny-valley reactionism

I also ponder this myself sometimes, and I appreciate that part of it

The search-engine business model was always odious, but we were willing to tolerate it because it was very inaccurate and easy to game, like a silly automaton which obeys simple rules

y’know, I fucking hate the “we” here. (not directed at you and it’s a whole thing but:) it’s another false equivalence, brought on by abusive extractive fuckers. igwym but… gah. rage.

(that touches on another post I’ve been trying to write for 3 years (this one I have not yet succeeded in clarifying (parts of it exists in some voicenotes to friends etc)))

This is specifically about the current shitty “AI Overview” that is currently and specifically pissing people off and Google essaying that as the only interface to search. (Sure they might not, but OTOH predicting future behaviour from past behaviour.)

That they did fucking terrible overviews in the past is not some sort of excuse for the current and specific issue with the current and specific AI Overviews.

wtf with calling hating this extremely hateworthy shit “uncanny-valley reactionism”

nah, not yet, imo. reasons: you overestimate how much & how many people care, and the complexities of doing such blocks well

I run the hosting of some small (by global standards) media setups and I’ve floated crawler defense there and had fairly little bite so far despite the issues the crawlers have caused. engagement brain is still extremely fucking prevalent

using it to source data in the meantime. I can tell you with absolute certainty that openai lies about its sourcing

if one presumes the money is gold and thus the scrolls were thin-film heavy metals, aerosolisation of that would explain quite a bit of the current situation