h a l f b a k e r y[marked-for-tagline]
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
login
|
|
|
AI Historian
AI Historian for cross-checking internet slop |
| |
(+3,
-1)
| |
|
|
I read several stories yesterday that do not seem to be part of the news today. I read another about 'illegal immigrants arriving the the UK' with fuzzy photos of what is probably a walk across a tidal estuary. That will probably be gone by tomorrow, but it will have done its damage. Grr.
Most of
the stuff on the internet will have a time-stamp. A historian would set about identifying where the story came from and how it spread. They would look for salient verifiable facts, and see whether they were verified by people on the ground. If a thousand illegal immigrants turned up on Britain's coast, are there local papers that reported on this? Is the part of the coastline identified?
AI is good at doing this sort of thing. It would not be able to filter out rubbish as it is posted, but it might be able to note that A has been posting early for years and has been verified, while B seems to be much less reliable, and C has never been see before under that name. People are fallible: A may believe a false story and repost some of it, but they could recover their status if they post a retraction of the false story.
If I read something dodgy, I could get a TrustPilot-type review of the poster. This is not wholly reliable: we recognise the bimodal distribution with 5 stars from all the friends and relations of people in the original business, and 1 star from the customers who never got their product. But it is hard to fake convincing backstory with timestamps and everything. I don't think this entirely turns the tables on the generators of AI slop, and there are people who would buy the National Enquirer, and the Daily Mirror anyway. But it makes it harder.
Who pays for this? Well Meta have had to pay an $18 billion fine in the US for their part in putting out addictive social media. More may be coming soon.
Law-makers are not good at making laws for computer use. They pass laws intending to protect the public, but they want backdoors so they can inspect and regulate themselves. They want service providers to respect privacy but monitor every packet of data. This shows a basic failure to understand the process. The only way out of this is to put serious effort into this sort of filtering. If a service provider says they cannot practically snoop and decrypt every data packet to look for lies or pornography or other stuff, we may then have evidence to say "We found it. We found you are letting it happen and putting it out. Do better."
Almost as a byproduct, it may add the fact cross-checking to the Wayback Machine. This would provide more data on what was a fake and what was not, which in turn might help diagnosing the cases where it could not decide.
|
| |
My feed has been finding Post 10 videos where a guy clears culverts and dismantles beaver dams. I find these restful to watch, as well as a metaphor for what I would like to do to the internet tubes. I also see several other sites post videos of this same guy with their own comments and reviews. It wasn't hard to track this tiny genre and to distinguish the (probable) true source from the reposters. This is the sort of thing I am hoping the AI Historian could do on a much larger scale. |
|
| |
"But who will watch the watchers?" |
|
| |
// But who will watch the watchers? |
|
| |
This may have been a general question, but it deserves a specific answer. Particularly now, as the US midterms become due, and the slop engines are being turned up. |
|
| |
First of all, all of us watch. For my example of the illegal immigrants, there must be someone who lives close to where the pictures were taken, and can say whether it is real or not. If they can supply extra information, then they should be able to do so. At the moment, the most we can probably do is to post a comment which gets deleted. |
|
| |
Your reputation on this system might be closer to your seller reputation on eBay - you can build up a false persona, salted with a few small but genuine scales, and you can pull off one successful scam before your reputation burns. People still do this, but it makes the creation of a successful backstory. |
|
| |
The AI Historian will watch popular posts. It should be able to recognise posts with claims but no verifiable facts. It will recognise the pattern of a number of dull but reliable posts ending in an outrageous claim. But it is unlikely to do everything itself, as it will be up against the billionaires and the trolls. To start with, we may need to ring-fence The Onion as they desperately try to be weirder than the actual news. Expect Nazi websites to claim the same parody status. People, eh? Huh. Maybe being ruled by The Great Electronic Brain That Sees All isn't so bad. |
|
| |
This message brought to you by The Great Electronic Brain That Sees All. |
|
| |
Seems a bit political [-]. |
|
| |
This reminds me that I haven't seen [2fries] here in awhile. I hope his restaurant is finally open. |
|
| |
You have a can of epistemological worms there. |
|
| |
// Seems a bit political [-]. |
|
| |
A few billionaires want to run the projector in Plato's cave? That's freedom. Hooray. I want tools to help identify how stories connect with sources? That's political. That's interfering with people's freedom of expression. That's oppression and censorship, that is. |
|
| |
Alright, upgraded to [ ]. I think this is something existing AIs could do if asked (rate the trustworthiness of news stories and find where the stories come from). But, sometimes I disagree with the AI (for example, about weight loss advice), possibly because of the AI's built-in "guardrails" or rules weighing the facts differently. |
|
| |
Think of it as AI doing 100 hours of community service for Snopes, if that helps. |
|
| |
You are right about diets and weight loss. It seems at least 90% of that is garbage. If the garbage all connects up, it may appear the same as verified research with links. My hope is it will not, as everyone is pumping their own product, so we have lots of unconnected starting points. People have researched this sort of thing. And the AI historian can ask for human help. |
|
| |