The Daily Wrong: AI-Generated Lies Every Day | Hacker News Reader

The Daily Wrong: AI-Generated Lies Every Day

dailywrong.com

140 pointsmemorable28 comments

This must be one of the best/most entertaining posts I've seen on HN.

Especially the bit about Gillette making razors for babies. Literally something that could become true in a parallel universe.

Incredible work!

Well the five blades with one on the handle thing was originally an onion article. Who knows.

I remember sitting at work in 2004 laughing at the original Onion gag:

https://www.theonion.com/fuck-everything-were-doing-five-bla...

Laughed again a few years later when 5 blades became real. These days I use a good quality twin blade, no need for 5 blades.

The babies one is at the bottom of this page: https://archive.ph/Y1oNG

Full "article": https://web.archive.org/web/20220705095136/https://dailywron...

tl;dr If you're on Firefox you can temporarily disable mixed http/https protection using the padlock icon near the URL. Of course it goes without saying that you should never provide anything sensitive with this protection disabled as MITM attacks are possible. [1]

The setup of this page is strange. http forwards to https but all the assets (JS, images, CSS) are full domain http only inside a served https page. There also some references that use the protocol free // prefix so the authors knew this was possible if they wanted both http and https but only used it in some parts of the page.

[1]: https://developer.mozilla.org/en-US/docs/Web/Security/Mixed_...

So how does this work? Does someone ask DALL-E to generate images of pregnant dinosaurs with text (if asked for text, DALL-E is prone to generate nonsense words which however strangely fit the general tone of the image), and then writes a clever article about it?

See the discussion by the site's author here: https://news.ycombinator.com/item?id=32168903

Only the images are generated by AI. The prompts seem to be written by humans. Impressive still.

I was flabbergasted for a moment thinking everything was AI generated. Even a bit scared, honestly.

The articles are GPT-3 generated, albeit heavily cherry-picked.

https://dailywrong.com/sample-page/

This is fantastic! Bluebird Sues Cat is a masterpiece.

This is hilarious, kind of like the onion but without pesky humans, would be interested in how these are generated, perhaps the title is used to generate the image with dalle.

Note when generating text try https://text-generator.io which will save you some when generating text or code Vs open ai

If the images were smaller, I think this might render decently on mobile -- on my iphone I either have to zoom out to see the image or zoom in to read the text.

Something is wrong with your ssl cert authority

Seems on brand

Images are currently broken for me.

Archive link is working: https://archive.ph/xNHUf

"image cannot be downloaded securely", I think I only can override it in Android's Chrome settings

Great satire, but sadly too late. The people who worked on deep-fakes, DALL-E and the likes of it won't stop and delete their products. Instead, evidence will be harder and harder to trust, until we've got a zero-trust society, in which the most manipulative party wins. In today's world, that would be Putin. Isn't it wonderful when AI developers have no sense of ethics?

That's not how technology works. Unless you want the world to hide & censor the information on how to make similar models, the data exists for anybody with sufficiently powerful computer resources to make similar models. The only thing that hiding away this tech would do is restrict access to well-funded attackers, which seems worse than the current situation.

As I am accused of having no sense of ethics.

Evidence has always been hard to trust, we should have had a zero-trust society for a lot longer. But because of the Gell-Mann amnesia effect and our trust in the unreasonable effectiveness of journalism, we find the USA fighting wars disheveling millions over weapons of mass destruction that never existed.

Nothing has really changed, it only became easier to automate what was already there before.

A “zero-trust society” seems like a dangerous phrase to advocate. Society exists because of trust in institutions.

Actually its a spectrum, there are high trust and low trust societies.

https://www.youtube.com/watch?v=404IeUzGNZ4

I agree with you, but no society functions at the “0” end of that spectrum. That’s anarchy. Even Mad Max / Book of Eli level dystopia still had nonzero trust in local warlords as institutions.

The example given of a high trust society is Finland. The example given of a low trust society is Afghanistan. [1] Which would you rather live in?

https://theconversation.com/trusting-societies-are-overall-h...

Finland is not one high trust society. Its two. They almost had a civil war back then when the sovjets nearly invaded and society split in two high trust societies (left - stalin collaborateurs) and right (finland patriots). There are still towns were the sides do not shop in the same super market.

> A "zero-trust society" seems like a dangerous phrase to advocate.

Thatcher and Reagan were a long way ahead there. They trusted institutions so little, particularly government, that they declared society doesn't exist. We've been trying to rebuild "society" since the 1980s.

Sadly the restoration our very human institutions, if it can be done at all, may only occur after all the painful lessons have been learned from thinking we can install crypto-cybernetic overlords as our servants.

> ... we find the USA fighting wars disheveling millions over weapons of mass destruction that never existed.

This violates all forms of acceptable reasoning.

> Nothing has really changed, it only became easier to automate what was already there before.

But even if we would accept that kind of reasoning, your excuse is that it will cause the same problems, but faster?

I don't get your position. I really don't get it.

> Isn't it wonderful when AI developers have no sense of ethics?

I'm not sure that's true. It's a subject I am curiously investigating, along with what really constitutes ethics in our technological milieu.

Look at it like this; if a technology exists, people having exposure to it, thus adapting their world-view, would be better than a powerful few having secret access to it.

Proliferation versus strategic limitation of technologies is a hard problem. Think about gun control. Don't forget, computers and the internet were developed primarily by the military. What did we expect to happen when we put the fruits of billion dollar defence budget into the hands of every man, woman and child on Earth?

Perhaps the outcome is a levelled playing field that massively subtracts from the power and importance of mass and social media. A "zero trust" world may restore focus to human relationships and attachment needs, and ultimately cause a total cultural rejection of "AI" in all its manifest forms.

To a large degree we can solve this with a data provenance infrastructure. It should be possible to have every step in the chain cryptographically sign images, video, audio and documents in a way that it can be proven no AI meddling has occurred. For an iphone it would not be a big leap to have the hardware sign video directly through the secure enclave.

How you combine that with data privacy though, I have no clue.

Reply on news.ycombinator.com