Midjourney considers concrete actions to prevent the spread of false images of candidates in the US presidential elections


Mathilde Rochefort

February 9, 2024 at 4:30 p.m.

4

Trump false arrest © © Midjourney

A Midjourney-generated image of Trump being arrested. © Midjourney

Political images could soon be banned on Midjourney. It was David Holz, CEO of the platform, who announced it during a live session on Discord.

As Americans head to the polls at the end of the year, more and more experts are concerned about the impact of generative artificial intelligence (AI) on misinformation. Last year, with Midjourney updates in particular, more and more photorealistic images surfaced online, some featuring Joe Biden and Donald Trump. As technology continues to improve, David Holz is considering banning the creation of visuals of presidential candidates.

AI executives concerned

I know it’s fun to take pictures of Trump, I take pictures of Trump. It is aesthetically very interesting. However, it’s probably best to step back a bit during this election. We will see “, declared the leader in particular.

Midjourney already has community standards aimed at limiting the creation of misleading or harmful images, but concerns persist about users circumventing these restrictions. The publication of false compromising images of one of the candidates could have catastrophic effects and influence the vote of some. While Holtz has not formally announced the ban on this type of content, it is likely that Midjourney will implement it quietly, he suggested.

He’s not the only prominent generative AI player concerned about the election year. Sam Altman, CEO of OpenAI, was concerned about the realism of deepfakes, which are much more sophisticated than in 2020. Mustafa Suleyman, director of Inflection AI, for his part indicated that his company’s chatbot would not be not authorized to defend a political candidate.

election poll united states © Andy.LIU / Shutterstock

This year, Americans will elect their new president. © Andy.LIU / Shutterstock

A very vast problem

This week, several AI companies announced measures also in this direction. Meta, Google and OpenAI will, for example, integrate a watermark into all images generated by their AI models.

Experts believe that these new regulations respond to temporary problems, but do not solve the broader problem of political disinformation, the cause of dramatic incidents in the United States in 2020. Last month, a deepfake audio of Joe Biden urged voters not to vote in the Democratic primaries. An event which highlighted the plurality of methods used to affect the elections.

Midjourney

Download

Midjourney

  • Commercial rights included in all paid plans
  • Relevance in understanding prompts
  • Accessible prices

Midjourney stands out as a pioneer in the field of illustration generation using artificial intelligence. Accessible to everyone, regardless of the level of artistic sensitivity, the user simply has to enter a text prompt describing the desired image. Midjourney then uses its artificial intelligence algorithms to analyze this textual input and generate an image that aligns with the descriptions provided. This mechanism for analyzing and generating art provides a rich user experience, opening the doors of creativity to a wide range of users.

Midjourney stands out as a pioneer in the field of illustration generation using artificial intelligence. Accessible to everyone, regardless of the level of artistic sensitivity, the user simply has to enter a text prompt describing the desired image. Midjourney then uses its artificial intelligence algorithms to analyze this textual input and generate an image that aligns with the descriptions provided. This mechanism for analyzing and generating art provides a rich user experience, opening the doors of creativity to a wide range of users.

Source : Bloomberg



Source link -99