B&H Photo Published an AI-Generated Guide Written by a Fake Person | Hacker News Reader

B&H Photo Published an AI-Generated Guide Written by a Fake Person

petapixel.com

83 pointsPaulHoule47 comments

>Multiple companies have toyed with the idea of using AI to generate stories, all of whom have been widely panned for the decision. G/O Media (which operates Gizmodo and Kotaku) attempted it earlier this year, the result of which was error-riddled pieces and widespread pushback from its staff. The company plans to move forward with AI authors despite this. Gannett, which operates USA Today, also attempted it but pulled back after it resulted in “botched” coverage of high school sports. More recently, Sports Illustrated was accused of the practice, a decision that resulted in widespread derision and the firing of its CEO.

all that matters is attracting $USER to $PLATFORM at a rate which minimizes $COST_OF_CONTENT_PRODUCTION to a point where it is below $INCOME_PER_USER. none of these "news outlets" have any integrity, they might back off of AI but they've already demonstrated that they'll publish anything that drives viewership. This also calls into question the legitimacy of human-written content published by these "news outlets" because whatever ideals journalism may have aspired to in the past no longer exist (if they ever even existed in the past).

If websites do this too much they'll just disappear. Why go to the website if I can get the answer from an LLM.

The same way most simple apps will probably not be needed if the Operating System can just conjure them on the fly.

"Give me a calculator app that includes a button for conversion to ternary"

Is the same as "Give me an app that every day will send me an expert article about Photography topics, keeping it varied so I learn about many aspects".

If something can be produced very cheaply by a website owner it can also be produced by the user for their own consumption.

> whatever ideals journalism may have aspired to in the past no longer exist (if they ever even existed in the past)

This is when I like to remind people that if you do a good enough job at journalism, you get an award named after a pioneer of sensationalistic “yellow journalism” who propagandized to start the Spanish-American War: https://en.m.wikipedia.org/wiki/Joseph_Pulitzer

Don't forget about the Nobel Peace Prize named after Alfred Nobel: The inventor of dynamite.

[1]:https://en.wikipedia.org/wiki/Alfred_Nobel

From the link:

"There is a well known story about the origin of the Nobel Prize, although historians have been unable to verify it and some dismiss the story as a myth. In 1888, the death of his brother Ludvig supposedly caused several newspapers to publish obituaries of Alfred in error. One French newspaper condemned him for his invention of military explosives—in many versions of the story, dynamite is quoted, although this was mainly used for civilian applications—and this is said to have brought about his decision to leave a better legacy after his death. The obituary stated, Le marchand de la mort est mort ("The merchant of death is dead"), and went on to say, "Dr. Alfred Nobel, who became rich by finding ways to kill more people faster than ever before, died yesterday." Nobel read the obituary and was appalled at the idea that he would be remembered in this way. His decision to posthumously donate the majority of his wealth to found the Nobel Prize has been credited to him wanting to leave behind a better legacy. However, it has been questioned whether or not the obituary in question actually existed."

To be fair to Nobel, dynamite has plenty of peaceful engineering applications, and its invention did represent an important innovation. If this story is true, it probably says more about the nature of journalism than the nature of dynamite.

> like to remind people

Is this phrase used figuratively, or are there actual times when you've exercised this preference upon someone?

I think if you said that to me, in a tone which implied you were enjoying saying it, I'd probably ask you why it's relevant. For example, can you demonstrate that Pulitzer Prizes are generally awarded to "yellow journalists"? Or can you show how the recipients of the prize are in any way awarded for behaviour which you're saying Pulitzer the person exhibited. Because it just sounds like a quirk of history, rather than revealing any important facts about journalism or how prizes are awarded

The Pulitzer Prize is supposed to be awarded to journalists whose work is actually good for society in some way, rather than simply churning out sensationalist content for crass commercial or ulterior political motives, and yet it's named after someone whose entire career was driven by the exact opposite motivations. It's funny, much like it would be funny to name a human rights award after Idi Amin. And it answers the implied question I quoted at the beginning of the post--that the "ideals" that journalism "may have aspired to in the past" did not, in fact, exist.

Thank you! I was just watching Oppenheimer yesterday, and cringed at the part where he points out that Alfred Nobel invited dynamite as a reason he might get the award for the A bomb. It's a silly idea. As you put it, it's a quirk of history. Interesting, but not relevant.

I had no idea "yellow journalism" and the Pulitzer Prize were linked in that way, thanks for that piece of information.

> When reached for comment, a B&H representative claimed the company’s policy was not to use AI-generated content and that it would “investigate this matter with our content teams.” Shortly thereafter, the article and profile were deleted. But for a month, the story remained online and positioned itself as an expert guide written by a real person and visitors to the blog would have been none the wiser.

Seems like a slight non-story to me: a parsimonious explanation here is that a staff (or contract) editor was given an assignment, and chose to fabricate it instead.

(A more interesting tidbit about B&H: their sales tax hijinks are legendary[1], and significantly predate Amazon's.)

[1]: https://www.dpreview.com/news/4007164682/bh-photo-accused-of...

At first, I figured that's the explanation, but you're ignoring a more puzzling statement in the article:

> To explain how an AI-generated story written by an AI-generated person was published despite this policy, B&H says it was part of a “test” to identify and protect against such content.

> “Our editors wrote a test article in mid-November as part of their efforts to identify and protect against AI-generated content. They overlooked adding disclosure to the article indicating that it was AI-generated, which was a mistake, and did not delete the article after the test.”

This sounds... rather unconvincing.

Yeah, that's pretty suspicious; I missed that.

Real Principal Skinner saying he only went to the bordello to get directions away from the bordello energy

So, someone had an assignment to write an article, found this on Google Drive (or some other document management system), and hit publish?

IDK, it all seems fairly plausible to me.

Also they tried to report it but their dog ate the form.

Regarding the sales tax “hijinks”, B&H actually won that lawsuit because the judge agreed that the law did not require them to pay sales tax on those payments as the NY AG claimed. It’s not hijinks if they are interpreting the law more correctly than the government.

I think it’s fair to call creative tax dodges “hijinks,” even if they ultimately end up being legal. This is the standard that I’d apply to Apple as well (with their Irish Maneuver).

(Another example of this in B&H’s case is their store-branded credit card, which gives you a NY tax rebate. This appears to be fine legally, but it matches the general trend.)

I wouldn't consider that a tax dodge at all given that the New York State Tax Department had previously issued advisories stating that the way that B&H was accounting for their sales taxes was the correct way to do so[1], and what they were doing was no different from the manufacturer supporting a sale with a drop in wholesale price on the item, which they sometimes do instead of manufacturer coupons. The attorney general was just plain wrong.

[1] https://www.mondaq.com/unitedstates/tax-authorities/1121150/...

There's no virtue in ignoring tax loopholes, it makes you into a schmuck if you pay more taxes than you're legally required.

Sure; these are non-exclusive properties!

Regarding the B&H complaint, that was in 2019, you can update it with "B&H Photo Prevails in NY Sales Tax Dispute" in 2021 [1].

[1] https://advertisinglaw.fkks.com/post/102h8g1/bh-photo-prevai...

> …a parsimonious explanation here is that a staff (or contract) editor was given an assignment, and chose to fabricate it instead.

The fabricated author biography and photo suggest that it's not just an "oopsie!", especially considering that they also said, "We typically do not create a writer profile unless we use someone more than occasionally". B&H understood what they were doing.

> and chose to fabricate it instead

In large companies usually there are even simpler mundane explanations. Most likely in this case there was no previous company policy, and some local department was running an experiment.

Then it got discovered, policy was established, and article was deleted. Nothing really malicious, nothing really interesting.

> their sales tax hijinks are legendary

Courts ruled in their favor. Don't blame B&H, blame the tax code

Note that it was a single article, B&H claims it was an accident and they deleted it since then.

"Sorry we got caught."

I'd use Hanlon's razor[1] instead.

[1]: https://en.wikipedia.org/wiki/Hanlon%27s_razor

Why do I care about this?

Commercial blogs have been posting stuff under fake bylines forever. I remember sites in the 90s that had non-existent people would write out random promotional blargh.

Am I supposed to expect that B&H has artistic integrity or something? They’re just a gear vendor. I don’t care if their guides are written by AI or JK Rowling as long as they help me pick out a lense or whatever.

My rule of 100: for every scam or ruse that's discovered, there are 99 others like it that fly under the radar.

Applied to medicine, my field: for every error that's discovered and made public, 99 others just like it go unremarked.

For every air industry near miss, 99 others are never reported.

The only foolproof method to detect AI-generated material is to examine the initials of the author. The author of this article is Alexandra Ibarra.

I worked at another retail site. We were also looking at doing this with SEO purposes. Whatever they were using to generate the articles was picked up by Google I assume because those pages were never indexed. It was also a secret from the ten person team they had turning out copy for seo blog articles.

The crazy thing was that whole section of the site was getting maybe a thousand views per hour on a site that got seven figure hits a day. Most employees didn't even realize it existed. The amount of money burned on it attempting to chase SEO was crazy.

AI is heralding a new era of enshittification. Was this a "Oops, we're sorry! Our mistake!" Or was it more a case of, "We're sorry we got caught!" I don't trust them.

The last time I tried ordering something from B&H Photo their inscrutable fraud prevention algorithms put a hold on shipping my order after they charged my card. No idea how much "AI" was involved, but I'm fairly confident it's either currently or in the future going to be more than "none at all." They tried calling me to verify, but calls from numbers not in my contact list go straight to voicemail because of all the phone spam I get. I called them back and talked to someone in their verification department, thinking that would be the end of it and the item I ordered would promptly be on its way.

Apparently that didn't work. A couple of days went by without it shipping, and then they tried calling me again. I canceled my order and let them know they'd permanently lost a customer.

Eventually someone from their customer service department reached out and said, "We can't tell you what the criteria is for triggering order verification," as if explaining that to me would somehow change my mind about doing business with them in the future. "Oh well yeah, your super-smart fraud prevention algorithms clearly make mistakes, and I'll happily just play phone tag with you for a week or so to get everything cleared up." Nope.

For this and many other reasons, I'm now buying everything I can at physical retail stores with cash, in part to help make sure it's still going to be possible to buy things in physical stores with cash in the future. I'm hoping it might yet be a while before real-life meatspace commerce is going to become thoroughly AI-enshittified.

Assuming good intent...

If you were testing that a 3rd party detector could protect the system from an insider (writers submitting ai generated articles), you would create a profile.

Since it is the holidays, it's _really_ easy to forget something, particularly when services have "freeze" dates when no further changes to production systems are allowed.

I am pretty sure CNN is also doing this, using some content from real people, then having the AI write the article. Check out their bylines and who "contributed" when you real an article.

It is not about B&H but how [Google] SEO works: you can write the best posts and will be hiddent in the long long tail even if they have specific content within a known audience.

Internet Search is broken.

Remember kids, AI-generated content isn't protected by copyright so you can do whatever you want with it.

And of course the article wasn’t archived anywhere. Wanted to see if it was actually a good article.

Who cares? This is both my initial reaction and, on reflection, something I’m genuinely curious about. If one can’t tell the difference without deep investigation, the consumer likely doesn’t care (I don’t). Hopefully / presumably it would still go through whatever editorial controls are in place. So who does care?

If this is rooted in fears of our jobs being taken, the train has left the station and we need to deal with this reality sooner rather than later. If anything, we’re incredibly fortunate ai is taking white collar jobs first (like mine). Hopefully these people have enough social and political capital to ensure laws change in healthy ways (ubi at some point?).

Maybe that works for pure entertainment.

I buy a lot of books, including a lot of nonfiction. I care quite a bit about the author. There's a lot of garbage out there with just flat wrong information. One of the few things you have to go on is an author's reputation.

Which, if you're making that up, becomes a garbage signal, and something approaching fraud. So for now at least, any outfit that publishes something that's supposed to be informative that's demonstrated to be "genai" derived means they've utterly lost my trust, and I won't be buying from them anymore.

There's also an ancillary issue I've never heard a good answer to. If you are publishing "genai" output, why would I pay you for it? I'll just ask GPT4 myself and not pay your cut.

If the answer is, "because we larded it up with all the signals that make it look like a human wrote it", well, see above about fraud.

GenAi is a tool like any other. If an author has been able to build earn my trust, I’d trust them to use the genai tool responsibly as well. This includes being ultimately responsible for the veracity of any claims made. If the facts end up being wrong, they’d lose my trust regardless of whether genai is in the loop or not. This abstracts to a publisher or company level as well.

For now, gen ai has enough issues that a human should probably be in the loop, either as a collaborating author or editor. At some point, that will probably change and that is okay with me.

You also make a good point about why buy books. Perhaps at some point gen ai will be good enough to generate high quality, true, and personalized info such that we won’t have to!

The fraud issue is interesting. I personally disclose when I use genai. But I wouldn’t be offended if someone else didn’t. Again, if I trust someone, I trust them to use the tool responsible. If I don’t, I’d take what they say with a grain of salt anyway.

As soon as I get the feeling that an article was written by a LLM, I stop reading. Usually it's repetitive phrasing, or grade school essay style summaries that turn me off. In short, a poor ratio of text to actual information. Even bad writing can have useful information in it, and I will plow through it. But vapid text that meanders without saying anything isn't worth anyone's time.

On the other hand, if LLM-generated text is actually good, and I can't tell... I guess then I don't care?

I was always under the impression that the grade school style summaries started before LLMs. Search engines recommends that kind of writing as condition to get into the answer box. It is bad writing never the less.

The LLMs had to be trained on something.

I basically agree with wolpoli and I think you as well. If the content ends up good, who cares.

Right now, it takes work and human collaboration to get quality content out of an llm. In the future it probably won’t.

> Who cares? This is both my initial reaction and, on reflection, something I’m genuinely curious about. If one can’t tell the difference without deep investigation, the consumer likely doesn’t care (I don’t).

"I was reading something to learn about some topic I don't know about, and it might've all been a bunch of bullshit, but since I didn't know it was bullshit, who cares!" What the . . . .

If I’m learning about a new topic, I do assume it might be bullshit and have a variety of sources to help give me confidence on truthiness. The llm really has nothing to do with this! I don’t know if I’d trust a random blog, or a legit looking website, or gpt4 more but I’d read them all and make my own judgement.

If I can tell it's AI generated, I stop reading immediately because if the creator didn't care enough to at least edit the article to not be obviously AI generated, why should I care enough to spend my precious time reading it?

Reply on news.ycombinator.com