A reverse Turing test:
an AI so convincing that another AI can’t tell it’s an AI.
8 Likes
Really, we don’t need the web at all. With the increasing power of hand-held devices, it should be possible to generate exactly the same bullshit that the generative content SEO spammers are creating, right there on-device. It won’t be any more (or less) meaningless or misleading. It doesn’t even count against your data plan (but your phone may get a trifle warm from time to time). And it’s personalizable; your private generated web can say exactly what you want it to, whether it’s scientific “facts” from your own personalized generative Wikipedia, to customized news that supports all your prejudices.
Ready when you are, capitalists. Let’s get this civilization-ending monster on the road!
6 Likes
I posted this elsewhere here a day or two ago, but it seems very apposite here…
Read it and despair.
15 Likes
Google’s generative content, at least, is less prone to hallucinations, than ChatGPT.
I tried asking ChatGPT for 10 “interesting or weird facts” about topic X which I know pretty well, and most of the results that came back are either so wrong it’s hilariously bad (much like that infamous court case) or so obvious it’s neither interesting nor weird.
ChatGPT: Yes, Ross Alley in San Francisco Chinatown is home to a hidden alley called “Waverly Place,” featuring vibrant street art, including murals, graffiti, and colorful installations. It’s a hidden gem for art enthusiasts.
Those of you who can check Google Maps will realize Ross Alley and Waverly Place are two separate alleys in Chinatown.
Yet this sort of content will be indexed, and if there’s enough of it to become authoritative, then Google’s own AI can be affected. And that is when we truly get buried in garbage content, generated by AI, summarized by AI, and consumed by AI, because we are all too busy doing nothing.
10 Likes
I don’t necessarily feel like this is an existential threat to the internet although it is going to be shitty for a while.
Tracking down sources for things basically cuts through LLM bullshit pretty fast, and do think it is interesting to consider a world where sources and provenance become more and more important over time.
As much as Google is feeding the bullshit cycle, if it gets bad enough and another engine pops up with a focus on producing higher quality results and cutting out garbage, we could also see improvements. User-submitted content is always going to be suspect, but at least for publications/news/etc you can start to build a level of ‘trust’ for lack of a better word, and if sites start pushing content farming crap they lose ranking for their whole domain.
None of that is simple; these are complex interconnected systems, but LLMs aren’t simple either and they exist, and search engines aren’t simple either.
And as others have mentioned, the worst case is that the world returns to webrings and smaller aggregators, leaving the AIs to converse among themselves, infinitely eating and regurgitating the same slop in an infinite cycle.
2 Likes
So are all views equally valid always? Even those who believe some are “subhuman”? Or that some are not capable of full citizenship because of their gender? Or that not everyone should get a living wage? Or because of one’s place of birth? Or because of one’s sexual orientation or gender?
At what point is it okay to ignore views that are literally genocidal? Or is listening to genocidal talk part of understanding “both sides”?
Is it okay to let nazis in a synagogue? Or a fascist in a lesbian bar? 
Just trying to determine the limits of “both sides” having valid points about whether or not violent genocide is okay to debate…
6 Likes
It will be interesting to see whether the open web (the bits outside the social media walled gardens) can revert to its state before a few large search engines monopolised the flow of traffic. The web was designed to work without search engines, because search engines hadn’t been invented yet.
If you maintained a web site before 1995, your site probably included a page of links to related sites that you regarded as reliable. If someone found your site by following a link from another site, that other site would appear in your log file. We’d check our log files to find what sites were sending us readers (not “customers”) and check those sites to see whether they were worth adding to our list of useful sites. The word “web” was chosen to reflect the idea that people would follow links from site to site, rather than there being some central site that everyone started from.
In 1995, Alta Vista was a gamechanger. It was the first search engine to index a fairly large proportion of the web. Like many, I stopped actively maintaining my page of useful related sites because Alta Vista did a better job. In retrospect, that was a bad decision, because it was also the start of the commercialisation of the web that made it much less reliable.
I could never see the point of webrings. If they were open webrings, where any site could add itself to the ring, they’d fill up with junk sites. If they were curated webrings, where the curator decided what sites to add, then they seemed less flexible than having a page of useful links on your site. Given that the webring widget included a link to a page listing all the sites in the ring, I don’t think many users used them as a ring anyway.
10 Likes
This remind me of a somewhat prescient post I wrote way way (ha) back in 2017:
“I’ve invented a movie that creates itself, watches itself, then deletes itself, all within a nanosecond. A few billion masterpieces were crafted, thoroughly enjoyed, and destroyed as you read this post.”
Oh, and this one from the same time period:
“This account is now under the control of an AI that creates posts in the style of walkswithdave as a labor saving endeavor to allow the original account holder additional personal time.
While this AI makes every attempt to recreate the experience of a genuine post by user walkswithdave, you may at times notice a slight deviation from expected output.
We assure you that as time progresses your individual responses will be monitored and incorporated into the process of generating an authentic user experience.”
These were both ridiculous and fanciful at the time but now…
7 Likes
And if a human (or an AI masquerading as one) publishes something about it on a site that’s not obviously trash, then someone else will cite it in a Wikipedia article.
Hm. Someone is probably working on their cool idea to use AI to fix all the {{cn}} tags on Wikipedia.
3 Likes
Generated content to boost your search engine ranking is one thing, but what worries me more is the use of AI-generated content to spread propaganda and disinformation more effectively. Another use is to generate phishing messages and the like. We already have deceptive, slanted and misleading content on the Web of course, a lot of it, but there is the potential to have much more and to make it less easily detectable.
5 Likes
Perhaps it’s time to dus off my DROLS search skills…
@STR@
*ARTIFICIAL INTELLIGENCE
AND
NETWORK(COMPUTER)
AND
ADVERTISING
COMERCIAL
[transmit]
—Do not cite the deep magic to me, witch. I was there when it was written.
edited to add: Note the all caps because back in the day, nobody wanted to waste bits on having separate upper and lower case characters in their character set.
2 Likes
At least when AI turns the internet into a swamp of shit we can turn them off instantly.
I think this is a huge step forward. Think of it as computers advancing beyond the A2 level, where they typically have to be spoon fed by an experienced and patient teacher, and into the B1 stage where they can learn independently. Soon, computers will not only be able to understand humanity, but will be able to learn from that understanding in order to anticipate and counteract any threats that humanity might offer up.
This might be a good opportunity to sell an SEO course with lifetime access. We’ll need to adapt ourselves to the model that best fits the computer’s image of ourselves. And SEO promises to be an important component of what it means to “human”.
1 Like
This search engine equivalent to the nanotech “grey goo” problem.
4 Likes
Their AIs will write it, their AIs will do the SEO on it, my AI will read it, my AI will summarize it and present it to me to decide if I want to read it.
4 Likes
Google has been doing this for 20 years (since 2003 when they deprecated pagerank) so I don’t know what is different now.
Yes there is so much VC money about that someone could gather a war chest big enough to spike Google.
Problem is that the only way you are getting a giant war chest these days is if you have an AI startup that promises mass, I mean utterly mahoosive, redundancies in the industries you are gunning for.
That’s all they care about. Providing good social media experiences and good search have solutions that involve less AI but that is not attractive to VC,
3 Likes
This topic was automatically closed after 5 days. New replies are no longer allowed.
