HN Explorer

Show HN: Factiverse AI editor – Fact-checking text made smarter and simpler (editor.factiverse.ai)

65 points by vinni2 ↗ HN
Heya HN, after 7 years of dedicated work, we're thrilled to unveil Factiverse AI Editor - a revolutionary tool to validate or debunk factual claims in any text, including AI-generated content.

Here's how it works: Our cutting-edge machine learning models analyze your text and identify check-worthy claims. We then scour search engines like Google, Bing, and Wikipedia, alongside manual fact-checks, to retrieve supporting and disputing evidence. The credibility of each source is carefully assessed using another machine learning model trained on expert fact-checks.

Try out the Factiverse AI Editor at https://editor.factiverse.ai/ and be sure to sign up and provide feedback on our Product Hunt page at https://www.producthunt.com/posts/factiverse-ai-editor or directly within the app.

To get started, check out our tutorial video at https://youtu.be/rMBHHfn6mk0 and hear a special message from the founders at https://youtu.be/Ri5rR_clpxg. Visit our homepage at http://factiverse.ai for more information.

Join us in revolutionizing fact-checking with AI!

29 comments

[ 5.2 ms ] story [ 71.5 ms ] thread
gojomo
While there are many, many reasons to worry about a flood of fake & false autogenerated AI content, in media & forums not yet armored against the new reality, I am also hopeful that various kinds of LLM writing-prostheses can help in places that do adapt.

This seems like a good example, automating what a speedy & diligent stickler-for-what's-supportable could discover, & recommend, about drafts from users (or other AI-writers).

Other opportunities might include drafting for clarity, & against common misreadings. Might services like Twitter improve if a user-loyal & advisory-not-censoring writing coach was there to warn when your wording might be misunderstood? When your reply might've assumed-the-worst about the root message whose actual or likely intent was more agreeable to you?

Sympathetic reading & sympathetic reading can be hard for those under various online stresses, or having just been fed a heavily-spun/decontextualized narrative, or just learning a new space's realities. It's possible specialized AIs could help improve online conversations, if deployed and tuned by those seeking understanding rather than division.

unshavedyak
Yea, it's probably true that as much as LLMs could flood us with bullshit, they might also give us the tools to deal with the floods.

It's just that it's much easier to generate floods of bullshit, especially in the beginning (where we are).

JimtheCoder
Overall, this is pretty cool. Good Job.

One issue I have with it, though...

It would be nice if I could filter what count's as a valid "source".

Meaning, having the BBC or The Guardian as a supporting source doesn't really make me feel much better.

Just a thought, but good stuff overall.

JPLeRouzic
I just tried and I had the same thought.
CharlesW
Did you try it? Each claim results in Supporting, Mixed, and Disputing citations. You decide what it means if [Media Source You Hate] supports or disputes a claim.
JimtheCoder
Yes, I did try it...that's how I found out that sites like BBC or The Guardian are used as sources.

All I am saying is that it would be nice to have the ability to choose categories of "sources" so I don't have to sort through a ton of clutter.

Using the El Nino example that was auto filled, a large portion of the sources look like they come from sites that more than likely will be using things like ChatGPT to generate their content in the near future, thus negating any real benefit...

CharlesW
Ah, I see. For me, it's just as helpful to know whether [Media Source I Hate] supports or disputes a claim as it is whether [Trusted Media Source] supports or disputes a claim. I'm generally always curious about the spin, especially when I know it's not objective.
codingdave
The whole point of checking facts is that it is objective. How those facts are spun into a story is subjective, but if you are checking anything that does not have an objective true/false answer, then that is not a fact.

So if you are just intending to report on different spins and leaving it to individual to choose which media to listen to... I'm honestly not seeing how that is any different than where we stand today.

miahwilde
What is a fact - fundamentally?
masswerk
An observation in the context of a specific frame of reference (and only in this context). As a frame of reference is mostly a social construct (or an idea anyways), it mostly relies on social consensus, which is both fundamental but also limited to the particular frame of reference.
masswerk
Just adding: The above is actually basic epistemology. Anytime you design a study, you are supposed to ask, what is considered a fact, in what framework is it considered a fact, how is it defined in that framework, what is the claim and reach of the framework (hence its legitimacy), what are contenders or alternatives, how does this conform with the study design and the overall framework it fits in.
elemos
Why is it an editor? Often times I want to fact check other things, not my own writing.
abeppu
Or rather, if you're actually writing, you probably already have sources in mind. Some people probably are willing to write stuff before they know whether it's true or false (astrologists, fiction authors), but who among those writers cares to fact check whatever they just put down?
vinni2
OP here, it is an editor because our first market segment is content creators. A browser plugin is on the way!
CharlesW
IMHO the pricing model isn't quite right — you could charge professional/enterprise users a lot more, but you're too expensive for filthy casuals like me. If there was a tier in the $20-30 paid yearly range with reasonable limits for people who might use it a handful of times per week, I'd be very interested.
vinni2
OP here, that's great feedback. We are still tinkering the pricing model.
abeppu
I found it annoying to be led through a multi-step tutorial modal that includes an "Add your own text" pane, but of course when you try to change the text away from an off-the-shelf example it nags you to sign up. The 3-preloaded examples aren't great, but I have to expect that you chose examples to demo on which the system works well (and for which there's a cached response) ... but I'm not going to make an account to see how well it does (or how long it takes) on an arbitrary new article.
dylan604
>alongside manual fact-checks,

what does this entail, or is it exactly what i think it means? does it list the names and credentials of those manual biobots?

ncr100
Positive feedback:

* Yes, more of this kind of technology, yes please. Actual trust is different from digital trust, and this tech moves digital trustworthiness closer to Actual. This is Human-centered technology, IMO.

Critical feedback:

* [minor] Red/Green colors for people with Deuteranopia MIGHT be indistinguishable: https://www.color-blindness.com/coblis-color-blindness-simul... https://www.allaboutvision.com/conditions/color-blindness/re... I know the factiverse Editor Results website is NOT the primary product, and I have "full-color" vision, but hey whenever I see Red and Green used for critical data, I criticize!

Idea:

* Build this into an assistant, an always-listening device, as a lie-detector, for use in conversations, or for video streaming services as a "truth track". Could be entertaining. I watch a lot of MSNBC and I want to know what is BS.

Zardoz84
A fact check tool that asks to Google and Bing about if something it's true... it's a joke, not ?
throwaway14356
I would love to see someone use the tool on a diversity of topics. Having me imagine how it is used is probably not what you want.

I see that one can look at the tool as a terrible way to search google a search engine full of bullshit without the most basic search filters...

..it can also be seen as a fast useful tool to at least check thing's superficially while you write.

Perhaps it can help cure some of the proverbial "eating fat is bad for you" doctors.

thumbuddy
I had this same idea about four years ago, never acted on it. Really glad to see someone else going for it!
YeGoblynQueenne
Congratulations on your launch and good luck with your product!

I have a question. You say that you "scour search engines like Google, Bing, and Wikipedia, alongside manual fact-checks, to retrieve supporting and disputing evidence". I presume that's done automatically? In that case, there is some concern that such search engines will soon be retrieving content created by Large Language Models that will have little chance of being factual. What is your plan to deal with that possibility?