BuzzFeed discloses use of AI to write stories

I still choose to believe that Buzzfeed articles are written by a room filled with chimpanzees and typewriters.

Monkey Keyboard GIF

10 Likes

Funny. I wonder how an AI trained on corpus of your actual comments here would have fared in generating the same comment.

Also, while we joke about AI- vs human-driven comment forums, bots are a very real problem. I wish more sites took the example of BB BBS and treated them all as what they are: spam.

15 Likes

The article was definitely weird. It almost felt like a random walk, moving from one topic to a peripherally related one by the luck of the dice— a series of somewhat related sentences, but without a well-defined path through the forest of ideas, so in the end you don’t really know what the point was. But I wonder if it might actually act like a confusion induction to some fraction of the readers, and put them in a suggestible state.

5 Likes

It’s pretty clear the internet and search results are rapidly going to devolve into AI-generated nonsense, often trained on text that includes content that wasn’t created by humans to begin with and makes no sense. There are already examples of plausible-appearing but completely wrong “answers” being provided by AI to simple factual questions that were derived from website SEO gibberish text - that’s just the tip of the onrushing iceberg. (Especially as search engines include more AI responses to queries.) The argument has been made - probably correctly, that the economic model for the big AI gold rush will largely consist of AI companies generating enormous amounts of AI-generated information pollution and selling tools to detect the garbage their AIs are producing, just in order to navigate around it. Which most people won’t. Search engines, which are already struggling (and often failing) to provide relevant results, will just collapse.

And that’s not even getting into the impact AI bots will have while clogging up social media and email, etc. I notice that post-Elno, Twitter replies are getting spammed with pretty obvious bots running the latest scam. I suspect something like Twitter, a mass social media environment where anyone can talk to anyone, will be impossible soon, replaced by highly curated, heavily gated, and much smaller, online social spaces.

Which makes it a bad (or, depending on intent, relatively entertaining) result. What the AI seems to be getting really good at is creating bland, plausible text that doesn’t really say much of anything - which means when it includes falsehoods, they’re harder to spot.

10 Likes

If any legislation emerges from the current swirl of fear and confusion and ignorance swirling around politicians over AI, the simplest and most effective one should be that all editorial content not generated by humans must be clearly labeled as such. Put it under the well-understood rubrics of truth in advertising or accurate product labelling.

11 Likes

So much “content” would necessarily fall outside the law that it would still have huge impacts, though, even if all articles were required to be labeled. Also, I could see such a law getting snarled up in 1st Amendment issues as well… (plus, it would potential just force content into non-US-based sites).

4 Likes

That’s why I specified editorial content. It has the potential to do the most harm and is easier to regulate than user-generated content. If the U.S. regulates on this issue in any way, this should be the initial approach.

It’s far more likely that the EU will once again take the lead here, as it did with the GDPR that’s now an aspect of most Americans’ on-line lives. I can see them coming up with a standard mark or marks that have to be slapped on any content initially generated by AI.

5 Likes

It will also dilute the reliability of sites like Wikipedia that may be crowd sourced and human edited but where third party citations from places like BuzzFeed are being utilized. A fitting word I’ve been hearing a lot lately to describe what is going on is “enshitification”.

9 Likes

This type of AI generated text interestingly has the same sort of feel as a typical AI generated image - if you squint your eyes at it and take the overall view, it has the appearance of being a certain type of picture, but then if you look more closely at any of the specific details you see that none of them really turn out to be what they were supposed to be.

I think @beschizza has commented on this effect in his old AI generated ‘Dune’ images. It forces your eye/attention to wander around frustratingly- unable to properly latch on to the finer details.

I have a hunch that this effect, while annoying in the current crop of AI works, will likely be harnessed in the near future to make a new type of art image that makes this never ending attempt of the brain to properly analysis the image somehow pleasurable. Imagine people staring at these images enraptured for hours on end, uneaten food dishes piling up around them as their families struggle in vain to break their attention…

5 Likes

They love eyes and are overly sensitive about hands, which they can’t seem to get right.

Wait a minute…overly sensitive about hands, rapidly veers off-topic. Trump is an AI!

5 Likes

I feel like editorial content - leaving aside 1st Amendment issues - is a drop in the bucket, but also I suspect we’re going to get into a “California Prop 65” situation (which required labeling everything with carcinogens, but resulted in labeling… everything). That is, we’re going to see the adoption of “AI-powered” tools for writing (and art, etc.), so there’ll be a very blurry line between AI-created (and human-edited) and AI-assisted work that will just result in pretty much everything getting the label, making it useless.

4 Likes

The first and last sentences of this post were written by a human

For no good reason, let’s do a little ChatGPT experiment.

Request: rewrite this post in the style of Dr Seuss.

Actual output:

Oh, the things that BuzzFeed said,
Using AI to make stories for their bed,
Their stock price soared, investors so glad,
To replace human writers, it’s not that bad!

But wait, said CEO Jonah Peretti,
Creating fake news could be quite messy,
False stories also from big outlets flow,
Like the AP’s report, “Jailed and Fired For Being Pro-Fascism: A Review”

The New York Times also had a say,
On a case where the government went astray,
Violating the Constitution, they did it with glee,
Creating a body where employers could fire those who disagree

Vox also chimed in, a new tool to see,
Bias in job decisions, oh my, oh me,
And more than 10% of Fortune 500,
Changed their policies, all thanks to data done right

But the online community, they had a hoot,
Saying the algorithms were too snazzy and cute,
They’d have to change them, toot sweet,
When BuzzFeed will deploy, still incomplete.

1 Like

I think that even more than this Buzzfeed example the recent CNET stuff points to how this is going to destroy the internet. At CNET Red Ventures is using AI to generate money management articles that they are using to insert credit card affiliate links into. The articles are intended to be generic and essentially useless, all that matters is getting that link in front of you.

What I think makes AI so dangerous is the ease and scale at which anyone can set themselves up generating huge amounts of web presence for affiliate sales while providing no value to the internet from a content standpoint - just shooting out lots of chaff (which will eventually also get used to train machine learning algorithms, making them worse…)

If I wanted to I could probably overnight write something that crawls Amazon for popular items, takes the name of the item, pipes that to GPTChat with the prompt “Write a positive review of [product name]].”, takes the output, makes every mention of the product name an affiliate link and publishes it to a review website filled with nothing but positive reviews my AI has generated.

I could generate, let’s say, 1 every two minutes (which is actually pretty slow…I could probably generate a more). At the end of my first day that’s 720 reviews. That never slows down.

And this is easy to do. Really easy. So eventually the internet is overwhelmed by AI generated cruft and useless. This is what I am really worried about with all this stuff, more so even that the copyright issues.

5 Likes

i posted over in follow up about how apparently cnet’s bot was plagiarizing existing articles. so currently at least humans are still needed for the seed. ( eta: link to the cnet article )

what will be really weird, i guess, is when - like with @MononymousSean’s dr seuss - ai is plagiarizing ai

separate thought: i didn’t realize till reading the generated posts above just how repetitious gpt can be. space filling variations around what would normally just be a sentence or two. like reading a grade school essay going for word count

8 Likes

The Futurism article really gets it wrong as to what the machine learning algorithm is doing and how it is doing it, but it does get it right in that the end result amounts to about the same.

It is probably especially noticeable because the kind of article they are generating is super generic to begin with so all of the input to the model is garbage “I need to write a 300 word article about a boring thing that I don’t care about to put an affiliate link into it” things that were written by humans doing the absolute minimum amount of work.

The algorithm is definitely not taking what is out there on the internet and changing a few words - it’s creating a database that includes that kind of article, generating word frequency and proximity tables from that and using that data to generate content statistically from that corpus. So it is going to be similar to all of it, but it will also make up stuff too since it doesn’t understand any of it.

I love that I can tell GPTChat to generate a negative product review and it will just make up believable stuff about a product not working as it should - that shows how dangerous this stuff is

3 Likes

That’s what I was thinking - by someone who doesn’t understand the subject and can never get to the point.

5 Likes

That’s fascinating. I actually can’t read the GPT-3 content without drifting off at the anodyne emptiness. It’s not really saying anything, and the language feels like it was chosen not to upset anyone while it’s achieving this nothing. If it’s a fake Chat GPT, well done! :slight_smile:

4 Likes

what you described is how gpt works. cnet does not say they use gpt. they lead with a mention of gpt and “other automated technology”, then call theirs variously an “ai assist” and “an ai engine”

the futurism article mentioned that microsoft has had a non gpt algorithm that did spit large chunks of copied text with no modifications. so it’s possible to design algorithms that way.

it feels entirely possible they licensed an algorithm which works via transformation rather than generation, or that they trained a relatively unsophisticated algorithm using such a small data set that the result was the same

3 Likes

Oof, yeah. There’s just layers and layers of terribleness to the coming AI content wave - the more you think about it, the more problems arise. Meanwhile the advantages are… pretty limited.

3 Likes

This topic was automatically closed after 5 days. New replies are no longer allowed.