Fooling an AI Article Writer - Schneier on Security

Erdem Memisyazici • July 27, 2023 8:22 AM

Wow. I wonder when we’ll get so bored of A.I. made for A.I. that we’ll remember set it and forget it is not the best idea for most human things like news, raising children, enforcing law etc.

Not until a new generation is raised who knows ChatGPT doesn’t open your third eye and give you powers from the future. 😄

Winter • July 27, 2023 8:45 AM

There is more:
Scientists used ChatGPT to generate an entire paper from scratch — but is it any good?
‘https://www.nature.com/articles/d41586-023-02218-z

The researchers designed a software package that automatically fed prompts to ChatGPT and built on its responses to refine the paper over time. This autonomous data-to-paper system led the chatbot through a step-by-step process that mirrors the scientific process, from initial data exploration, through writing data analysis code and interpreting the results, to writing a polished manuscript.

To put their system to the test, Kishony and his student Tal Ifargan, a data scientist also based at Technion, downloaded a publicly available data set from the US Centers for Disease Control and Prevention’s Behavioral Risk Factor Surveillance System, a database of health-related telephone surveys. The data set includes information collected from more than 250,000 people about their diabetes status, fruit and vegetable consumption, and physical activity.

Jan Doggen • July 27, 2023 9:12 AM

That’s it folks: prepare for the so-often-announced singularity. AI feeding itself with AI generated drivel.

Morley • July 27, 2023 9:45 AM

It’ll be hilarious/sad if we actually fulfill the dead internet theory. Big ol’facepalm.

yet another bruce • July 27, 2023 10:15 AM

I recall regular human journalists have fallen for this sort of thing before. My personal favorite was when the Beijing Evening News ran a story repeating the Onion gag that the U.S. Congress had threatened to move out of Washington unless a fancy new Capitol was built.

https://www.latimes.com/archives/la-xpm-2002-jun-07-fg-whoops7-story.html

JonKnowsNothing • July 27, 2023 10:26 AM

MSM article (HAIL warning)

OpenAI discontinues its AI writing detector due to “low rate of accuracy”

On Thursday, OpenAI quietly pulled its AI Classifier, an experimental tool designed to detect AI-written text.

Prolly not a good idea to expect AI to own goal.

===

ht tps://arstechnica.c o m/information-technology/2023/07/openai-discontinues-its-ai-writing-detector-due-to-low-rate-of-accuracy/

(url fractured)

Ted • July 27, 2023 10:38 AM

From the article:

The only way this will ever be stopped is if Google steps in and dramatically deranks or bans AI-based sites like this

What say thee Google?

First – as the AI bot parroted in its article – was making Klikclac mandatory in Glorbo a disadvantage to casual players? (hehe🤭)

Second, where to even begin?

Steve • July 27, 2023 10:53 AM

As @yet another bruce points out, humans, of which journalists are a subset, get fooled by bogus stories all the time.

A cruise through a site such as Snopes[.]com will be sufficient evidence that bots have nothing on humans in terms of credulousness.

As an aside, I’ve come to the conclusion of the oft-cited Turing test is not so much a measure of machine intelligence as of human gullibility.

Clive Robinson • July 27, 2023 12:20 PM

@ Erdem Memisyazici, ALL,

Re : Can there be too many cat vids?

“Wow. I wonder when we’ll get so bored of A.I. made for A.I.”

The human race has a number of failings two of which are,

1, We keep cats as pets.
2, Those who can not have pet cats watch vids of others pet cats.

So I would say there will always be a “human subset that will never get so bored” of cat vids, so AI4AI is almost a foregone certainty…

Phillip • July 27, 2023 1:28 PM

All,

I agree with any feedback loop spiraling at great length into the abyss. OTOH, it should always be helpful to access to some reasonable distillation of what is on the internet, regardless of the correctness of the cumulative sources. For how long it has been the case: “I found it on the internet.” Does not make AI super-duper helpful, it might continue with mild to moderately helpful. Until loose mergers and acquisitions. One should hedge his or her bets favoring any real “Zero Point” soon coming.

Steve • July 27, 2023 4:40 PM

@Clive Robinson

1, We keep cats as pets.

And vice versa.

Clive Robinson • July 27, 2023 7:47 PM

@ Steve,

“And vice versa.”

+1 😉

A friend now long gone observed of cats, that under the cute looks and fur,

“they are nasty bu99ers”.

As are some other apex predators that kill.

Sadly of all the species that kill for food, all but a few spieces kills for sport/pleasure/tourture.

That exception species are generally us humans and the creatures we have as pets and working animals.

I guess the question is,

“Which way does it go?”

Do they learn from us, or do we learn from them…

Donald G • July 28, 2023 12:28 AM

Jan Doggen: Apparently there are already AI Article Readers, feeding into high-speed trading algorithms. So I doubt we’ll have to wait long.

Does it count as illegal market manipulation to post fake news that’s unlikely to fool any humans, but might fool an AI? (Blizzard is publically traded.) It’s definitely legal to trade based on obvious bullshit posted by unaffiliated parties; so, a Reddit lurker could potentially predict an AI-generated article, and the market reaction to it, for profit. I didn’t check whether there was any market reaction to this news.

anon • July 28, 2023 1:07 AM

It is a natural progression.

Humans have become so fond of generating BS, and then consuming it, that they figured out a way to outsource BS creation to an artificial “intelligence.”

All the better. Now we have even more BS to gobble up!

anonymous french • July 31, 2023 5:03 AM

Fooling other people is a sport for some. No wonder an AI became a victim too.

@Steve and @yet another bruce In France, there is a group of writers who created a fictional philosopher named Botul in the nineties.

Then they tried to fool other journalists and writers, so that they include fictional quotes from Botul in they articles or books. They got a few quite famous victims, for ex. prominent “philosopher” B-H Lévy.

See https://en.wikipedia.org/wiki/Jean-Baptiste_Botul , that’s quite funny indeed 😉

Carl • August 16, 2023 9:47 AM

Soon on Reddit: the AI will argue with another AI and wait for approval from the third AI.

Schneier on Security

Comments

Leave a comment Cancel reply