A developer constructed a ‘propaganda machine’ utilizing OpenAI tech to spotlight the risks of mass-produced AI disinformation

• A developer constructed an AI-powered propaganda machine in lower than two months.
• Utilizing OpenAI instruments like ChatGPT, the mannequin value lower than $400 a month to function.

A developer has constructed an AI “disinformation machine” utilizing OpenAI-powered know-how like ChatGPT.

The undertaking, named CounterCloud, took its creator two months to finish and value lower than $400 a month to function, highlighting how low-cost and easy it may be to create mass propaganda.

The creator, who selected to stay nameless, used the title Nea Paw and claimed to be a cybersecurity skilled, Wired reported.

Rework expertise with studying that works

Functionality improvement is important for companies who wish to push the envelope of innovation.Uncover how enterprise leaders are strategizing round constructing expertise capabilities and empowering worker transformation.Know Extra

In a video posted on YouTube, Paw described how the undertaking was carried out and why they have been doing it.

They mentioned they needed to see AI disinformation “work in the actual world,” including that “sturdy language competencies of huge language fashions are completely suited to studying and writing faux information articles.”

Paw mentioned they began by inputting “opposing” articles into ChatGPT and writing prompts instructing it to write down a counter article.

The AI bot would then provide completely different examples of the identical article, written in numerous kinds and with completely different viewpoints.

They mentioned it will do that by “creating faux tales, faux historic occasions, and creating doubt within the accuracy of the unique article.”

A gatekeeper module was then added in order that the AI would solely reply to probably the most related content material, whereas audio clips of newsreaders studying the faux AI-generated articles and pictures have been added to create authenticity.

Additionally they created faux journalist profiles and faux feedback beneath some articles.

The system was then informed to love and repost messages that aligned with its narrative or write “counter” tweets to people who did not.

In two months that they had a totally autonomous AI-powered system that generated “convincing content material 90% of the time, 24 hours a day, seven days every week.”

The creators have not but set the mannequin reside on the web, as “it will imply actively pushing out disinformation and propaganda. As soon as the genie is out on the web, there is no such thing as a realizing the place it will find yourself,” Paw mentioned.

However they added that they thought extra good would come from publishing CounterCloud and educating the general public about how these methods work from the within.

The risk AI poses to democracy via its skill to speed up the creation and unfold of on-line disinformation has been one of many greatest issues within the AI debate.

i’m nervous in regards to the impression AI goes to have on future elections (at the least till everybody will get used to it). customized 1:1 persuasion, mixed with high-quality generated media, goes to be a robust power.

— Sam Altman (@sama) August 3, 2023

The CEO of OpenAI, Sam Altman, just lately voiced his concern in regards to the problem, saying that “customized 1:1 persuasion, mixed with high-quality generated media, goes to be a robust power.”