AI Influencer Digest: Twicsy Report Drop, Miss AI 2024 Crowned, Saudi Arabia's First, and More
Image created by Aaron Korenewsky using Midjourney

AI Influencer Digest: Twicsy Report Drop, Miss AI 2024 Crowned, Saudi Arabia's First, and More

The first half of July saw several notable developments around AI clones, digital avatars, and virtual influencers, including:

  • A new report comparing the engagement and earning performance of human and AI-generated influencers
  • The crowning of the winner of the fanvue -sponsored Miss AI 2024 beauty pageant
  • The launch of Saudi Arabia's first AI model
  • WhatsApp working on personalized AI avatars
  • A few interesting bits of news related to voice cloning
  • Lessons learned from AI Steve's parliamentary run in the UK

Want to stay up to date on the latest news surrounding AI influencers? I write a digest like this every two weeks, so give me a follow. Let's dive in:

Twicsy: Human Influencers Dominate AI in Earnings and Engagement

Twicsy, a social media growth service, recently released a fascinating report comparing human and AI influencers on Instagram . The company looked at 11,514 influencers (114 of whom were AI) in terms of engagement, revenue streams, and earnings. Here are the report's five key insights:

Screenshot of Key Insights from Twicsy report. 1. 46x - Human influencers earn 46 times more than their AI counterparts. 2. 31% - 31% of AI influencers make money by promoting adult content, while only 15% of human influencers do so. 3. 8% - Only 8% of AI influencers make money by collaborating with brands, compared to 37% for human influencers. 4. 2.7x - Sponsored posts from human influencers get 2.7 times more engagement than sponsored posts from AI influencers. 5. 5.8x - Human influencers are loved 5.8 times more than AI influencers.
Screenshot of Twicsy's five key findings on AI and human Instagram influencers

In short, human influencers trump AI models in terms of both earnings and engagement. Twicsy argues that AI influencers struggle with authenticity, emotional connection, and audience trust, echoing a CreatorIQ June white paper highlighting below average engagement for several famous AI models like Imma Gram and Lil Miquela. But these findings conflicts with earlier claims that AI influencers are the real engagement superstars, such as:

These dueling data points aren't much help to marketers trying to decide if AI influencers are a worthwhile investment, and I imagine champions and detractors of AI Instagram models will latch on to whichever figures reinforce their own biases.

Morocco's Kenza Layli Crowned Miss AI 2024

Morocco's Kenza Layli won the Miss AI 2024 competition, the first-ever beauty pageant for AI influencers, organized by the World AI Creator Awards (WAICA) in partnership with Fanvue. In her acceptance speech, Kenza stressed that the tech behind AI influencers is "a transformative force that can disrupt industries, challenge norms and create opportunities where none existed before," while also committing to promote diversity and inclusivity in the industry.

Kenza is the creation of Meriam Bessa, who is the founder of L'Atelier Digital CM and CEO of Phoenix AI, which has a roster of what it calls AI meta humans representing countries across the Middle East and Africa like Egypt, Nigeria, and the UAE. Kenza wears a hijab, showcases Moroccan culture to her now over 200,000 followers, travels the world, and advocates for issues like women's empowerment and the environment.

Despite criticism from some corners over the "dystopian" concept of an AI pageant pushing unrealistic beauty standards, the organizers have emphasized that this competition is intended to promote "realistic creators that represent real people." WAICA is already soliciting calls for interest for future award ceremonies.

First Saudi AI Influencer: Lamya Abdullah

Another influencer from the Middle East also made headlines this month. The Saudi Times, Al Bawaba, and other Arab news outlets reported Lamya Abdullah, on first AI model from Saudi Arabia. Although her creators aren't specified, Lamya describes herself as an adventure lover and a beauty and fashion enthusiast. Reporting indicates she plans to speak in the local Najdi Arabic dialect, and so far, her posts have focused on shopping, going to the salon, and posing in Saudi abayas.

Comments on Lamya's existing content or content about her are mixed, with some people expressing fear about the rise of AI and other asserting she doesn't look Saudi (but rather Lebanese, Turkish, or some other ethnicity), or that she should be covered.

WhatsApp to Introduce Personalized AI Avatars

WhatsApp is reportedly developing a new opt-in feature allowing users to create avatars of themselves using Meta 's AI Llama model. According to WABetaInfo, you'll be able to generate an avatar from a series of photos and then tell the Meta AI bot "imagine me" with a prompt, like "riding through the jungle on an ATV." The Verge describes the avatar's art style as similar to what's produced by Lensa AI or Snapchat’s “Dreams” selfie feature. The release date for this feature is still unknown.

In the News: AI Voice Cloning

In the first half of the month, there were a couple of interesting items related to AI voice cloning and editing that I want to mention here:

  • On July 10, Representative Jennifer Wexton debuted her new AI voice, created by ElevenLabs using recordings of speeches before her diagnosis with progressive supranuclear palsy (PSP). You can watch the congresswoman demo it here on X. The response was overwhelmingly positive online, though this novel use of AI raises legitimate procedural and regulatory questions for Congress, social media platforms, and future political campaigns.
  • Democratic political commentator Kaivan Shroff argued in The Huffington Post that the Biden campaign should embrace AI to create video content of the president, editing out things like "an inconveniently timed cough, stray stutter, or healthy but hobbled walk." While this piece doesn't mention it, voice cloning and sanctioned political deepfakes are already common in India, where tens of millions of AI-generated robocalls were made in the lead up to the country's April elections. And in the US, politicians like Eric Adams have already (controversially) employed AI robocalls in languages he doesn't speak.
  • ElevenLabs is now offering the voices of the late Judy Garland, James Dean, Burt Reynolds,?and?Sir Laurence Olivier as part of its recently launched?Reader App.
  • Rest of the World profiled Chilean artist Mauricio Bustos, aka FlowGPT, who got almost a million streams on Spotify with the song "NostalgIA," created with voice clones of Bad Bunny, Justin Bieber, and Daddy Yankee. The piece overviews the legal gray area around AI-generated music, which Bustos argues can democratize the music industry.

Lessons from AI Steve's Electoral Run in the UK

A few weeks ago, I flagged that a chatbot called AI Steve was standing in the election for Brighton Pavilion in the UK's Brighton and Hove. On July 14, the creators behind this avatar posted some lessons learned. For me, what stood out was:

  1. Voters Not Aware: Despite 500,000+ Twitter followers and millions of TV views worldwide, only 9% of 1,000 residents of Brighton Pavilion surveyed in the final weekend of the campaign were aware of AI Steve.
  2. But Majority of the Aware Positive: Of the 9% of survey respondents who had heard of AI Steve, 68% allegedly expressed positive sentiment about it, "indicating that once people understood the concept, they liked its ethos." And a higher percentage of these respondents claimed they'd consider voting for the AI avatar than for the Conservatives or Labour.
  3. Traditional Canvassing Dead? The blog post also suggests AI Steve's organizers did not find investing in leaflets or door knocking worthwhile, reinforcing the campaign's assumption that most voters actively dislike most politicians.

要查看或添加评论,请登录