Pangram Image 1.0

https://www.pangram.com/blog/introducing-pangram-image-1-0

137•gmays•21h ago

Comments

my002•18h ago

People often shit on AI detectors but Pangram's text AI detection has been very good, IME. Not perfect, but good enough to be useful in most circumstances.

jszymborski•16h ago

My trouble with Pangram is that the calibration is totally off. It can tell you if there is a drop or an ocean of AI in the text you submit. The "X% of AI" message doesn't really mean much. You can test this yourself: Slowly add generated text to a human write paragraph and you'll see it hit "100% AI" quite quickly.

jrhizor•15h ago

Good?

Matticus_Rex•15h ago

It analyzes in segments, and the percentage is of the segments, not of the words.

jszymborski•12h ago

Yes but the segment sizes are huge and any amount of LLm generated sequences within that block cause all of it to be flagged as AI.

Alpha3031•14h ago

I don't really know why you would expect a "X% chance of containing AI" number to rise slowly or smoothly alongside the percentage of text that has a 100% chance of containing AI.

kingofmen•12h ago

Because it isn't an "X% chance". It's an "X% of words".

Alpha3031•9h ago

Surely not at the sub-paragragh level? OP said their test passage was a paragraph.

xnx•17h ago

I'm very skeptical of AI detection and watermarking for text (especially short passages), but AI detection for images and videos seems much more plausible (for now at least).

I would love if Meta licensed this tech to use for everything on Facebook, Instagram, and Whatsapp. Including ads!

mediaman•17h ago

Why are you so skeptical of text detection?

Do you believe it generally doesn't work or do you believe that if you work at it you can evade them and therefore there's no point to it?

JacobAsmuth•16h ago

You should look at the data. Look at the actual measurements on the accuracy of Pangram's V4 detection model. Try to fool it. It's not easy at all.

zeroonetwothree•15h ago

It’s often inaccurate with short text. For example it has flagged text I wrote by hand with no AI input or editing.

But I agree it’s good for longer text of 1000+ words.

fxwin•8h ago

> For example it has flagged text I wrote by hand with no AI input or editing.

Do you mind providing that example?

74948494i4•17h ago

What’s the use case for these things?

poszlem•17h ago

What's the use case for knowing when you are looking at a fake AI image and not at a real photo? Is that seriously your question?

tantalor•17h ago

A few likely use cases,

1. Social media moderation. You want to limit the proliferation of this stuff because it annoys users.

2. Trades where authenticity matters, like journalism, law, education. You have to weed out the crap and people will try to deceive you.

80fj0348i•13h ago

eCommerce - you may want to reduce fraudulent listings, e.g. used goods, housing, collectibles/auctions...

alkonaut•16h ago

Can’t these detectors be used in the training loop of models?

JacobAsmuth•16h ago

People have tried it, it doesn't work.

bobajeff•16h ago

Which people? I'm no AI researcher but I believe basically all GANs employ a discriminator which basically to me it sounds like this is.

dwohnitmok•15h ago

I don't know about Pangram's viability as a discriminator but I can tell you as someone who has recreationally trained GANs that they're really really finicky to train and training fails in all sorts of inscrutable ways. It wouldn't surprise me if even though in theory you could do a GAN-style form of training with Pangram, in practice it might actually be very difficult.

treyd•12h ago

Modern diffusion models aren't trained like GANs. GANs failed to scale and generalize. Diffusion models do sometimes use Variational Auto-encoders (VAEs), which are a similar structure but which functions differently.

XTXinverseXTY•15h ago

In theory yes, but likely not economical for the slop arbitrageur

It is not a lucrative business and they are not clever people to begin with

paulpauper•16h ago

just compress it? lossy should fool detection

quixoticelixer-•16h ago

It doesn't

JacobAsmuth•16h ago

Didn't read the article I see...

hex4def6•16h ago

At least on gemini generated images vs real ones, it works surprisingly well.

Taking a photo of an AI generated image on my grubby laptop screen with my phone, moire and all, still got detected. So the signal it's using is pretty robust.

halfcat•2h ago

An actual photo you took is not AI-generated though.

Where’s the line? If you take a selfie in Times Square, and there’s an AI-generated image in an ad behind you, that photo you took isn’t AI-generated.

d0100•2h ago

There is an image showing literally your case, it highlights the AI generated image in a portrait on the wall

leothetechguy•16h ago

Awesome research like this restores my faith in humanity.

sean_pedersen•15h ago

An ill advised undertaking IMO: Any attempt to build AI generated content detection systems is flawed, since the outputs of such a system may be used to train an even better fake data generator. This leads to an equilibrium state of digital uncertainty: nothing in the digital realm can be deemed as real anymore - only as digital.

zeroonetwothree•15h ago

How is that a downside exactly? If we can’t tell then we are no worse off.

sean_pedersen•15h ago

It will create false trust in their score - giving people a false sense of certainty.

karahime•14h ago

Adding to this, the most motivated actors to get around it or find an edge case, e.g. in politics, will use it to say "Look, of course the scene depicted in this compromising image actually happened, the detector said so".

brookst•2h ago

Same way weather forecasts are a bad thing?

80fj0348i•13h ago

> the outputs of such a system may be used to train an even better fake data generator.

[citation needed] - my understanding of how entropy works in real photos vs. synthetic images (AI or not) is that this is extremely difficult without severe degradation of image quality.

Trung0246•14h ago

Interesting, it detected NovelAI stuff now. Looks like one can easily defeat it by having pixel-level modifications to make AI images to have characteristics of digital brush strokes and reducing "noise" in output. At least in my case for anime stuff.

nwhnwh•11h ago

The cookie modal is disgusting... It takes a few moments to realize that if you only want the necessary ones, you need to close the modal as the existing action button is for allowing everything.

raincole•11h ago

From my test it works surprisingly well. I have a digital painting that is only posted very recently (therefore not in the training set). I asked GPT, Gemini and Riverflow to generate images in the same style, and use SD to add noises to remove SynthID.

Pangram gave correct results for all of them (the original one is No AI; all the others are Likely AI).

VerifiedReports•8h ago

Is what?

sashank_1509•7h ago

We can do this for code too I’m guessing?

I think for small snippets and well defined codebases, we probably can’t,

But if we say a fully vibecoded project vs handmade, I suspect there are AI tells outside of basic stuff like bar names etc that a model can learn, and it will be fun to know.

We could probably also do these for maths.

I guess what I’m getting at, is even in these days extremely verifiable fields, I guess AI has diverged enough that I think we can train a detector to separate AI from how a human thinks?

sashank_1509•7h ago

https://www.pangram.com/history/images/50f285f3-66db-4484-b3...

I’m honestly surprised pangram got this right, with my human eyes it feels indistinguishable but apparently not !

gosen•5h ago

that's really nice, I was so impressed by pangram AI text detector so I hope image detector is also that good

modeless•12h ago

AI detectors can't work in theory, but Pangram works well in practice. This is because the AI companies are not trying to fool Pangram. They have no incentive to; in fact the incentive goes the opposite way. They want to watermark their outputs to leave no doubt as to whether something was generated by their AI.

eranation•8h ago

I was a skeptic. Couldn’t trick pangram yet.

buf•6h ago

removed

cheshire_cat•6h ago

Seems not so simple after all. How would one even go about training a local model on their own writing voice?

vintermann•5h ago

You can't trick it with AI generated text, but can you trick it into calling something AI generated which isn't? From what I've seen these AI detection tools are very good at detecting AI generation, they're also very good at calling anything AI generated.

mjdv•4h ago

Pangram boasts an extremely low false positive rate, in contrast to other AI detection tools. This has matched my limited experience.

If you have text that you can prove is not AI-generated (e.g. it is demonstrably older than about 5 years) that gets an AI flag, you could try tweeting at the Pangram founder, I've seen screenshots of him offering bounties (admittedly mostly as a dare to claims he doesn't believe).

Hoasi•3h ago

Really? Pangram is very easy to bypass in written form.

gunalx•2h ago

Yes. Even as little as a "humanize and rephrast this to be more natural" prompt with high temperature on a uncommon model reduces the confidence greatly.

orbital-decay•2h ago

AI companies might have no incentive but determined actors do, by definition. Anything Pangram could possibly do (even semantic variation detection) can be trivially bypassed by doing img2img on a custom trained local model, which is in itself done trivially in practice by many people. Same for text, train it on anything and use distribution-aware sampling. From your words, you already have false sense of reliable detection, so a work of a determined actor will trick you into believing false negatives.

Deus Ex

Dutch 'idealist' bank ASN blocks non-Google Android, including GrapheneOS

FAA Begins Retiring Decades Old Floppy Disks from Air Traffic Control

Gemini Omni 1.2 Flash

Blender's UI and Runtime: Much Faster Than GPUI

Claude Code sends your email address to websites

RIP Cybertruck

Zero Hour: Telling time was a matter of survival and strategy during WWI (2014)

Living beyond the limits of survival: wood ants trapped in a gigantic pitfall (2016)

Kugelblitz – Mac-Native Google Earth Alternative

Multiplayer tanks that demolish real buildings from OpenStreetMap

MS Paint Remote Code Execution Vulnerability

We've built an open-source feature flag server for 11 years

From Linux to Wikipedia: How Volunteers Build the Internet

Mbodi AI (YC P25) Is Hiring Research and Robotic Engineers

Ask HN: What in the world is taking Ubuntu so long?

University libraries are being swarmed by AI bots, putting public access at risk

Lippitt-Knoster in Practice (Field Notes for Managing Complex Projects)

Pwn.college

Where Each U.S. State Gets Its Tax Revenue From

AI for scientific discovery: How researchers use Claude

sshole

MTG-I2 First Images

Chomsky and the Two Cultures of Statistical Learning (2011)

Show HN: Agentic Identity Broker, OAuth Delegation for AI Agents

The Hackintosh Reborn

Britain to make salary disclosure in job ads mandatory

Nvidia Kubernetes Engine (NVKE)

Inkling 2 Small

Window Sweaters

Deus Ex

Dutch 'idealist' bank ASN blocks non-Google Android, including GrapheneOS

FAA Begins Retiring Decades Old Floppy Disks from Air Traffic Control

Gemini Omni 1.2 Flash

Blender's UI and Runtime: Much Faster Than GPUI

Claude Code sends your email address to websites

RIP Cybertruck

Zero Hour: Telling time was a matter of survival and strategy during WWI (2014)

Living beyond the limits of survival: wood ants trapped in a gigantic pitfall (2016)

Kugelblitz – Mac-Native Google Earth Alternative

Multiplayer tanks that demolish real buildings from OpenStreetMap

MS Paint Remote Code Execution Vulnerability

We've built an open-source feature flag server for 11 years

From Linux to Wikipedia: How Volunteers Build the Internet

Mbodi AI (YC P25) Is Hiring Research and Robotic Engineers

Ask HN: What in the world is taking Ubuntu so long?

University libraries are being swarmed by AI bots, putting public access at risk

Lippitt-Knoster in Practice (Field Notes for Managing Complex Projects)

Pwn.college

Where Each U.S. State Gets Its Tax Revenue From

AI for scientific discovery: How researchers use Claude

sshole

MTG-I2 First Images

Chomsky and the Two Cultures of Statistical Learning (2011)

Show HN: Agentic Identity Broker, OAuth Delegation for AI Agents

The Hackintosh Reborn

Britain to make salary disclosure in job ads mandatory

Nvidia Kubernetes Engine (NVKE)

Inkling 2 Small

Window Sweaters

Pangram Image 1.0

Comments