> What Factually does is different. It takes a question typed by a user and hands it to a Large Language Model, or LLM, to generate some query strings. It performs up to three Internet search queries, then feeds the top nine web pages it found to a pair of LLMs ...
So it selects its sources according to their SEO-gaming proficiency?
In the UK, we can rely on fact checkers from the BBC, who are impartial and would never be caught doctoring videos of presidents of ally countries [2]. The UK government would never send 100 current/past members of their party to interfere in a foreign election [3].
In the US, free speech protections are very selective (depending on what you planning to say). The rest of the Western world does not even have the laws protecting free speech. No need to worry.
Regardless of the legal protections of free speech, the general notion worldwide is that corporations are allowed to create and post whatever nonsense they feel like with utter impunity. It's one thing for a single person to write the timecube, it's another entirely to promulgate fact-checker robots which completely obscure the real facts with utter bullshit. We could all see this coming from miles away. I see definite need to worry.
It almost seems naive for the author of this article to expect every organization labeling itself as a "fact-checker" to practice thorough journalistic standards, or for them to be unbiased. It seems inevitable that just like opinion/propaganda networks claiming to be "News," the same concept exists for "fact checkers," despite many other fact checking organizations being a lot more legitimate.
While I like how the concept that fact checking has tried to respond to the social media age's flood of inaccuracies and disinformation, encouraging the idea that we should try to more thoroughly verify news stories and sources to make sure they are accurate, in practice I'm not sure if they've been a major net positive.
For the target audience, who is presumably someone who has fallen for some misinformation or propaganda, fact-checking often seems to come across as condescending to that person. "Actually, that thing you believed isn't real, here's some smarty-pants reasons why you were wrong."
Either that, or it's like the community notes system where it's an endless war of clever comebacks.
I don't really have a solution in mind, just these thoughts on the present state of things.
What an odd comment considering the related fact check on Snopes is actually completely right about the incident in question [0].
The original article by Aphyr is specifically about factually.co which uses LLMs to synthesize "fact check" pages. The issue here is that LLMs (which do not formally reason and do not have a concept of truth) are being used to mass produce factually incorrect summaries of events masquerading as fact checks.
I used to use slopes as my primary "fact authority", then post 2016 something changes. They now often answer with a very different context than they start out with. They do state that context they're actually answering, but sometimes it's very subtle that it's nearly unrelated. In these cases, if you just read the title and first paragraph and answer/summary, you will be misled. Being misled several times, seeing others misled many times, and to often see that they completely avoided responding to the actual topic, I don't trust them anymore. I'd rather just spend a few extra minutes looking at original sources than be tricked like that.
There's also seems to be a heavy bias in what they cover, which is fine (I believe biased/adversarial news is the best way to dig out truth), but I doesn't help my trust.
> What Factually does is different. It takes a question typed by a user and hands it to a Large Language Model, or LLM, to generate some query strings. It performs up to three Internet search queries, then feeds the top nine web pages it found to a pair of LLMs ...
So it selects its sources according to their SEO-gaming proficiency?
Yes. That is the narrative the industry I is selling: AIs running our societies. Apparently AI runs Albania now too?
Thet want to make sure you do not have any choice and at that point You Will Like It.
> Unavailable Due to the UK Online Safety Act
Cute [1].
In the UK, we can rely on fact checkers from the BBC, who are impartial and would never be caught doctoring videos of presidents of ally countries [2]. The UK government would never send 100 current/past members of their party to interfere in a foreign election [3].
[1] https://web.archive.org/web/20251111010701/https://aphyr.com...
[2] https://www.telegraph.co.uk/news/2025/11/03/bbc-report-revea...
[3] https://www.telegraph.co.uk/politics/2024/10/17/labour-sends...
Maybe AI is the catalyst for newspaper and news companies to start thriving again as the last remaining credible sources.
Doubt they'll be able to transition. They are too deep in the slop.
Quis custodiet ipsos custodes?
Seems like a defamation suit waiting to happen.
How can we possibly stop this madness? Will it require draconian legislation and enforcement?
Increasingly I think that "free speech" should apply to humans only, not to humans armed with a gas-powered bullshit spewer.
In the US, free speech protections are very selective (depending on what you planning to say). The rest of the Western world does not even have the laws protecting free speech. No need to worry.
Regardless of the legal protections of free speech, the general notion worldwide is that corporations are allowed to create and post whatever nonsense they feel like with utter impunity. It's one thing for a single person to write the timecube, it's another entirely to promulgate fact-checker robots which completely obscure the real facts with utter bullshit. We could all see this coming from miles away. I see definite need to worry.
It almost seems naive for the author of this article to expect every organization labeling itself as a "fact-checker" to practice thorough journalistic standards, or for them to be unbiased. It seems inevitable that just like opinion/propaganda networks claiming to be "News," the same concept exists for "fact checkers," despite many other fact checking organizations being a lot more legitimate.
While I like how the concept that fact checking has tried to respond to the social media age's flood of inaccuracies and disinformation, encouraging the idea that we should try to more thoroughly verify news stories and sources to make sure they are accurate, in practice I'm not sure if they've been a major net positive.
For the target audience, who is presumably someone who has fallen for some misinformation or propaganda, fact-checking often seems to come across as condescending to that person. "Actually, that thing you believed isn't real, here's some smarty-pants reasons why you were wrong."
Either that, or it's like the community notes system where it's an endless war of clever comebacks.
I don't really have a solution in mind, just these thoughts on the present state of things.
[dead]
[flagged]
What an odd comment considering the related fact check on Snopes is actually completely right about the incident in question [0].
The original article by Aphyr is specifically about factually.co which uses LLMs to synthesize "fact check" pages. The issue here is that LLMs (which do not formally reason and do not have a concept of truth) are being used to mass produce factually incorrect summaries of events masquerading as fact checks.
[0] https://www.snopes.com/fact-check/ice-chicago-pastor/
Could you point to a snopes page that is demonstrably wrong or misleading?
I used to use slopes as my primary "fact authority", then post 2016 something changes. They now often answer with a very different context than they start out with. They do state that context they're actually answering, but sometimes it's very subtle that it's nearly unrelated. In these cases, if you just read the title and first paragraph and answer/summary, you will be misled. Being misled several times, seeing others misled many times, and to often see that they completely avoided responding to the actual topic, I don't trust them anymore. I'd rather just spend a few extra minutes looking at original sources than be tricked like that.
There's also seems to be a heavy bias in what they cover, which is fine (I believe biased/adversarial news is the best way to dig out truth), but I doesn't help my trust.
> Could you point to a snopes page that is demonstrably wrong or misleading?
I have the same question for you as the parent comment.
As if past or current fact checking wasn't already that. I guess rewriting history is part of the course.