OpenAI’s voice scandal sounds larger AI alarm

OpenAI CEO announced GPT-4o, a new AI model, resembling Scarlett Johansson’s voice, sparking a dispute on AI’s use of copyrighted material without consent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
On May 13th, Sam Altman posted one word through his X account: “her”. The word was tweeted by the OpenAI CEO to announce the release of the company’s new multimodal AI model, GPT-4o, that could interact with humans through voice, camera, and text. In the company’s online live-telecast of the announcement the same day, one of the AI voices named ‘Sky’ was charming and flirty in its responses, ready to please the company’s executives with whom it talked.

Sky’s intonation and personality also sounded like the sultry AI voice in a film called Her, portrayed by actress Scarlett Johansson. A 2013 iconic movie by director Spike Jonze, Her starred Johansson’s voice as an AI personal assistant who forms a relationship with actor Joaquin Phoenix’s lonely character. It’s a tragic story about loneliness, love, and human emotion. It’s also a film that Altman has constantly mentioned as one of his favourite renditions of how AI can interact with humans.

Witness the final chapter of the Lok Sabha Elections unfold on HT with live vote count and results. Explore now! Explore now!
A week after the tweet, Johansson posted a public statement accusing OpenAI of consciously, without permission, copying her voice to make the AI model Sky sound like her. She mentioned that Altman had approached her in September 2023 and then again two days before the launch of the multimodal AI model to license her voice, something she had refused.

“The voice of Sky is not Scarlett Johansson’s,” said Altman in a counter statement on May 20th, adding that the company cast a voice actor for Sky before any outreach to Johansson. Though OpenAI paused the use of Sky, the damage was done.

Can AI Give You Compensation or Credit?

For more than a decade, Big Tech has been developing generative AI by feeding these models data from the World Wide Web, all of it gathered without permission or consent. It then analyses it and regurgitates logical, contextual answers. Most of the technology leaders, like Altman, want to see the culmination of this technology: An idea called Artificial General Intelligence or AGI, a machine that can respond, interact, learn, and think like humans, trained on data created by humans.

According to estimates by MarketsandMarkets, a market research publisher, the projected AI market size is expected to reach a staggering $407 billion by 2027.

But today, these models regurgitate synthetic content without compensation, consent, or credit to any of us. For the profit of a few corporations.

The question of compensation is something top media companies are already working on. Last year, media companies like CNN, The New York Times, and Reuters coded their platforms to stop AI bots from OpenAI or Google from crawling their websites and learning from the content they create. The muscle-flexing helped as the ever data-hungry AI companies are now partnering with media companies to license their data for AI training. Recently, OpenAI and NewsCorp signed a content deal valued at over $250 million. Media organisations across the world might be able to adopt this method to generate revenue for their content.

What the Johansson-versus-OpenAI dispute has done is highlight how AI can use intellectual property, copyrighted material, and even imitate individuals without permission, credit, or payment. If a popular beloved actress like Johansson’s cadence can be so readily copied by tech companies, what chance do creative individuals stand? The independent bloggers who run websites, the influencers, the small mom-and-pop stores of digital content who cannot get SEO leads anymore but aren’t big enough to sign a deal with a tech company?
tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys tys
Synthetic Media Requires Consent

It doesn’t help that AI technology is advancing at a much faster scale than policymakers can understand. We’re already in the new era of multimodal AI. These tools can recreate anyone’s voice or video likeness in minutes. With two minutes of Prime Minister Narendra Modi’s audio, for example, you can recreate a synthetic voice of him in any language that says anything. While the potential of digital identity theft, financial fraud, and scams has just increased, synthetic media also threatens personal privacy and consent.

The post OpenAI’s voice scandal sounds larger AI alarm appeared first on The Muslim News.