- Creative AI
- Posts
- OpenAI in trouble with Scarlett Johansson
OpenAI in trouble with Scarlett Johansson
Once again, OpenAI finds itself in trouble.
Erik Knobl
May 26, 2024
Good day, humans. Generative AI is not perfect, and its massive use generates problems. This week, three different issues occurred that are worth discussing.
In today's Newsletter, we include:
Scarlett Johansson. Image generated in Midjourney
OpenAI angers Scarlett Johansson
Once again, OpenAI finds itself in trouble.
According to Scarlett Johansson's account, the story goes as follows: About nine months ago, OpenAI CEO Sam Altman approached the actress with a request to license her voice for a new ChatGPT conversation feature; Johansson declined. Just two days before the company's main event last week, in which GPT-4o was launched, Altman contacted Johansson's team, urging the actress to reconsider. Allegedly, Johansson and Altman never spoke directly, and Johansson never permitted OpenAI to use her voice. However, the company debuted GPT-4o days later, drawing attention to the "Sky" voice, which many believed sounded alarmingly similar to Johansson's in the movie "Her."
And then this post on X:
It's a clear reference to the 2013 film of the same name, in which Johansson is the voice of an AI assistant that a man falls in love with.
In the statement, Johansson claims she was "shocked, angered, and in disbelief that Mr. Altman would pursue a voice that sounded so eerily similar to mine." In response, Altman issued a statement denying that the company had cloned her voice, saying they had already hired a different voice actress before contacting Johansson.
OpenAI has since removed Sky's voice from its platform "out of respect" for Johansson.
Finally, according to a Washington Post report, records show that OpenAI did not intentionally seek to copy Scarlett Johansson's voice for ChatGPT.
Why is this important?
This controversy not only questions the ethical practices of OpenAI but also opens a broader debate on the limits of artificial intelligence and consent in the digital age. How far can technology go without infringing on individual rights?
Microsoft's Copilot Plus PCs are going to record everything
Microsoft has taken another step towards integrating artificial intelligence into laptops. The company has introduced new "Copilot Plus PCs."
However, one of the main features it will enable is "Recall," which will use AI to create a "photographic memory" that allows you to search for everything you've done and seen on your PC. Although this feature promises enhanced productivity, it raises significant privacy concerns.
Absurd recommendations generate concerns about the future of generative AI.
Google's "AI Overview" feature makes things up
A large number of users have been sharing on social media how Google's new "AI Overview" feature makes absurd claims:
AI Overview from Google recommends looking at the Sun for 5 to 15 minutes, or up to 30 if you have dark skin.
According to AI Overview from Google, doctors recommend smoking 2 to 3 cigarettes a day during pregnancy.
AI Overview from Google claims it is possible to train 8 days a week.
It would be funny if it weren't so dangerous.
A similar situation has already happened with Google's Gemini image generator in the past, where the company eventually withdrew it and adjusted it before relaunching it.
Despite promises of innovation and productivity, these systems are still far from perfect and can produce unexpected and dangerous results.
Let's move on to more creative news:
Krea.ai's available tools
Krea launches its video tool in Open Beta
Krea.ai has announced that its Krea Video tool is now available for all users. Krea Video uses keyframes and text prompts, intending to provide more control and flexibility in video creation.
It's a new way to use generative AI, so it's worth experimenting with the tool while it is open to the public.
Character Reference from Leonardo.AI
Leonardo.ai releases Character Reference
Leonardo.ai has made the "Character Reference" feature available to all users, both paid and free. This tool allows you to change the position, style, or context of a character while keeping the character's face consistent.
Collage with images generated with sref 76981987
To generate images with this style, just add --sref 76981987 at the end of your prompt. For example, for the top 4 images, the prompt is:
Photographic portrait --ar 16:9 --v 6 --sref 76981987
If you want to learn how to use Midjourney:
Anthropic is mapping the mind of an AI
The company Anthropic has identified how millions of concepts are represented within Claude Sonnet, one of their large-scale language models. And they have shared a first analysis of how it works.
Thank you for reading. See you next week!
Hi š Iām Erik Knobl, Product Designer by day and explorer of Generative Artificial Intelligence on weekends. I share my learnings in this newsletter. Consider subscribing to stay in touch.