OpenAI is in trouble... again - Weekly News Roundup - Issue #468
Plus: is Rabbit R1 a scam?; Humane is looking for a buyer; FDA clears Neuralink for a second human trial; Nvidia reports another big quarter; China builds the world's first drone carrier; and more!
Hello and welcome to Weekly News Roundup Issue #468. It was another week full of interesting announcements and developments.
We’ve had the Microsoft Build conference, where the tech giant showcased new AI products and services. We’ve also had the AI Seoul Summit, where AI leaders discussed ways to enhance AI safety while fostering innovation, inclusivity, and mutual prosperity in AI development. I will have separate articles about these events published in the next few days, along with the delayed article about Google I/O (I apologise for the delay, life got in the way).
Meanwhile, OpenAI found itself in controversy over using a voice similar to Scarlett Johansson’s in its demos without Johansson’s permission, which is the main topic of this week’s news roundup.
In other news, the FDA cleared Neuralink for a second human trial, and Wired interviewed the person who received the first Neuralink implant. We also saw a startup emerging from stealth mode, working on the radical idea of whole-body transplantation. Meanwhile, Coffeezilla exposes Rabbit R1, Nvidia reports another good quarter, Elon Musk’s xAI is close to a $24 billion valuation, humanoid robot Sophia delivers a commencement speech in New York, and more!
I hope you enjoy this week’s issue!
OpenAI’s introduction of their latest and best model, GPT-4o, was almost perfect. The new model amazed the public with its almost human-like responsiveness. Conversations with it felt like talking to a human, not an AI. GPT-4o climbed to the top of many leaderboards. It also brought many people to the ChatGPT app, generating $4.2 million between May 13th and 17th, according to mobile market analytics company Appfigures. OpenAI’s Spring Update successfully overshadowed Google’s annual event, Google I/O, which took place a day after OpenAI released GPT-4o. Instead of focusing solely on Google’s latest AI models and AI-powered products and services, people were comparing them to what OpenAI had shown the day before.
The demos showcasing what GPT-4o and ChatGPT can do looked like science fiction has become a reality. After seeing them, many people drew comparisons to the movie Her, saying that the voice used in the demo, named Sky, sounded very close to the voice of Scarlett Johansson. For Johansson, the similarity was too close.
According to a statement issued by Scarlett Johansson, who played the role of an AI assistant named Samantha in the movie, Sam Altman personally approached her with a proposal to voice the new AI assistant, which she declined. OpenAI confirmed that this is indeed what has happened. Two days before the new version of ChatGPT was shown to the world, Altman contacted Johansson’s agent asking to reconsider the offer. According to OpenAI, the request was to join as a future additional voice. Before Johansson could reply, OpenAI released the new version of the app, with Sky being the voice chosen for the demos.
“When I heard the released demo, I was shocked, angered, and in disbelief that Mr. Altman would pursue a voice that sounded so eerily similar to mine that my closest friends and news outlets could not tell the difference,” wrote Scarlett Johansson in her statement. After two letters requesting OpenAI to explain in detail the process by which they created the voice of Sky. On May 19th, six days after the Spring Update event, OpenAI published a post explaining how the voices for ChatGPT were chosen. The post states that the “voice of Sky is not Scarlett Johansson's, and it was never intended to resemble hers” and apologises to Johansson for not communicating better. The same day the post was published, OpenAI took down the voice of Sky from the pool of available voices in ChatGPT.
The Washington Post adds more context to the story by citing the agent representing the actress behind Sky’s voice (whose identity hasn’t been revealed to protect her privacy). According to the agent, neither Johansson nor the movie Her were ever mentioned by OpenAI, and what we heard in the demos is the actress’s natural voice. The Washington Post also revealed what kind of actors OpenAI was looking for: they should sound between 25 and 45 years old with “warm, engaging [and] charismatic” voices. Additionally, OpenAI requested that the actors be nonunion. OpenAI issued a casting call in May 2023, in the middle of the screenwriters’ and actors’ strikes, which ruled out many actors. According to OpenAI, each actor received compensation above top-of-market rates, and this will continue for as long as their voices are used in OpenAI’s products.
It seems the situation has calmed down as I am writing this article. OpenAI honoured Johansson's request to explain themselves and removed Sky’s voice from the ChatGPT app. According to Johansson, the voice was removed "reluctantly." Johansson has not taken any legal action.
OpenAI technically honoured Scarlett Johansson’s request and did not use her voice. They just used the voice of another actress who sounds similar to Johansson and whose voice references the movie Her. If Johansson and OpenAI were to go to court, it would be up to a judge to decide on the matter.
However, the controversy around Sky’s voice exposed how OpenAI operates. In the same post explaining how the voices for ChatGPT were chosen, OpenAI revealed that they selected five voices from over 400 submissions. I’ve listened to the remaining four voices, and they all sound generic, like someone from the US. Yet, out of those over 400 submissions, OpenAI chose to include one that resembled Scarlett Johansson’s portrayal of Samantha. Not only that, they used that voice specifically in demos seen by millions of people. Such things do not happen by accident. Her is one of Altman’s favourite science fiction movies. Altman even tweeted the word “her” during the Spring Update.
With strong links between what OpenAI presented on May 13th and the movie Her, OpenAI’s claims that Sky’s voice “was never intended to resemble” Scarlett Johansson’s voice sounds hollow. Additionally, Johansson revealed that Altman's plan was for her participation in the project to “bridge the gap between tech companies and creatives and help consumers feel comfortable with the seismic shift concerning humans and AI," adding that Altman felt her voice “would be comforting to people.” That plan backfired spectacularly, to say the least.
Scarlett Johansson’s voice was not the only controversy OpenAI got involved in this week. Shortly after GPT-4o was released, Ilya Sutskever, one of the co-founders of OpenAI and its long-serving Chief Scientist, left the company. Together with Jan Leike, an AI safety researcher, Sutskever joined a number of safety-oriented OpenAI employees who left the company in recent months (I wrote more about that in the last week’s news roundup).
These departures raised concerns about their impact on the safety culture at OpenAI and the direction the company is heading. It was also revealed that OpenAI employees were asked to sign a non-disparagement agreement before leaving the company, which would prevent them from saying anything negative about the company, its products, services, or its leaders. Violating this agreement or declining to sign it means they can lose all vested equity they earned during their time at OpenAI. With OpenAI worth over $80 billion, we are talking about possibly millions of dollars in vested equity. Altman acknowledged that non-disparagement agreements are a thing at OpenAI but were never enforced. He also wrote he wasn’t aware of them. If any former OpenAI employee is worried about it, they can contact him to fix that, Altman tweeted.
The recent controversies surrounding OpenAI erode trust in the company. For a company that promises to bring about artificial general intelligence (AGI) to benefit all of humanity, the recent actions raise questions about whether OpenAI, under Sam Altman’s leadership, is the right entity to create AGI and to safely manage such potentially history-defining technology.
If you enjoy this post, please click the ❤️ button or share it.
Do you like my work? Consider becoming a paying subscriber to support it
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🦾 More than a human
US FDA clears Neuralink's brain chip implant in second patient
Neuralink has received a green light from the FDA to implant its brain chip in a second person after it proposed to fix a problem that occurred in its first patient. Earlier this month, Neuralink shared results from its first human trial, which allowed a paralysed patient to control a computer and play games. However, 85% of the implant threads have detached and are not working. Neuralink opted to avoid another brain surgery, choosing instead to issue a software update and keep the implant to gather more information.
Neuralink’s First User Is ‘Constantly Multitasking’ With His Brain Implant
In this interview, Wired speaks with Noland Arbaugh, the first human to receive Neuralink’s brain implant. Arbaugh shares what the selection process was like, and what it felt like waking up from the surgery and using the device for the first time. He also discusses his experience working with the Neuralink team and how the device has changed his life. Additionally, Arbaugh addresses the problem of the implant's threads pulling out and reducing the device's effectiveness and explains what it took to overcome that obstacle.
Having a head for longevity
BrainBridge is a startup working on the radical idea of whole-body transplantation, which involves taking a patient's head and attaching it to a healthy, brain-dead donor body with the help of an advanced robotic system. The team at BrainBridge hopes their procedure could offer new hope to patients suffering from untreatable conditions. Currently, the company is focused on developing computer models for comprehensive feasibility studies and refining its processes. BrainBridge hopes that, after these studies are complete, the first surgery could happen within the next eight years.
Frozen human brain tissue was successfully revived for the first time
Scientists from China have discovered a combination of chemicals that allows human brain tissue to be frozen and thawed while maintaining its normal function. Their experiments showed that thawed organoids made from human brain tissue closely resembled their unfrozen counterparts in appearance, growth, and function, even after being frozen for 18 months. This breakthrough in freezing human brain tissue could improve investigations of brain development in the lab for health research. It might even pave the way for the freezing and revival of entire brains in the future.
🧠 Artificial Intelligence
▶️ $30,000,000 AI Is Hiding a Scam (17:10)
The first AI-first devices - Humane AI Pin and Rabbit R1 - did not work as people hoped they would. Humane AI Pin got into hot waters for being a trash device and now it’s Rabbit’s turn. If Coffeezilla, who exposed multiple scams and has a perfect track record, comes after someone, you know something shady is happening. In the first video in a series about Rabbit R1, Coffeezilla exposes the people behind the device, their actions during the crazy days of NFTs, and how that connects with their AI assistant. In the second video, he compares what was promised to what was delivered and uncovers Rabbit’s misleading marketing and claims.
Humane, the creator of the $700 Ai Pin, is reportedly seeking a buyer
Humane, the creators of the failed AI Pin, are looking for a buyer. The company has reportedly priced itself between $750 million and $1 billion, and the sale process is in the early stages, according to a report published by Bloomberg.
NVIDIA Announces Financial Results for First Quarter Fiscal 2025
Nvidia published its financial results for Q1 2025 and continues to cash in on the generative AI boom. The chip maker reported a profit of $26 billion in the quarter, up 18% from the previous quarter and up 262% from a year ago. As expected, the data center segment was the main driver of revenue, bringing in a record $22.6 billion, up 23% from the previous quarter and up 427% from a year ago.
EU’s AI Act set to enter force next month
The EU’s AI Act is set to enter into force next month after receiving a final rubber stamp from the European Council. However, the rules won’t fully apply for another 24 months, and high-risk systems will have an additional year to meet their obligations.
Musk’s xAI Nears Close of Deal Valuing Startup at $24 Billion
According to Bloomberg, Elon Musk’s AI company, xAI, is set to close a $6.5 billion fundraising round in June with a potential post-money valuation of over $24 billion.
Hugging Face is sharing $10 million worth of compute to help beat the big AI companies
Hugging Face is offering $10 million in GPU compute to help developers create new AI technologies, products, and services. The goal of this program, called ZeroGPU, is to make state-of-the-art AI technologies accessible not just to big tech companies but also to small developers, academics, and startups, and to counter the centralization of AI advancements.
The Recipe for an AI Revolution: How ImageNet, AlexNet and GPUs Changed AI Forever
This post takes us back to the late 2000s and tells the story of how Fei-Fei Li created the ImageNet dataset. This, together with Nvidia introducing CUDA, laid the groundwork for Geoffrey Hinton, Alex Krizhevsky, and Ilya Sutskever to create AlexNet, the neural network that started the deep learning boom in 2012 and the decade of breakthroughs in AI.
An Ordinary Beijinger's Thoughts on AI
I recommend checking out this week’s ChinAI Newsletter, one of the best newsletters on Substack relaying what is going on in AI in China. The main story is about what an ordinary Beijinger thinks about interacting with AI. It also features a report about the state of AI safety in China and how China is using AI news anchors to deliver its propaganda.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
🤖 Robotics
China Builds World’s First Dedicated Drone Carrier
China has secretly launched its fourth aircraft carrier, believed to be the world’s first dedicated fixed-wing drone carrier, in a shipyard on the Yangtze River. The ship might serve a training role or act as an experimental platform for drone operations. The exact purpose and status of this new carrier remain mysterious.
‘Robot-phobia’ takes a toll on food and hotel workers
Businesses in the hospitality industry are increasingly looking into robots to close the labour gap. However, according to a new study that surveyed 321 lodging and 308 food service employees from across the US, bringing in more robots and automation increases workers’ job insecurity and stress. This, in turn, leads to greater intentions to leave their jobs, potentially worsening the hospitality labour shortage.
In a first, humanoid robot Sophia delivers commencement speech in New York
Humanoid robot Sophia has become the first robot to deliver a commencement speech to over 2,000 attendees at a college in Buffalo, New York. University officials claim the decision to bring in Sophia aligns with their theme for the year, AI. Some students, however, did not approve of this decision and protested through an online petition that received close to 2,000 signatures.
▶️ SliceIt!: Simulation-Based Reinforcement Learning for Compliant Robotic Food Slicing (2:09)
In this video, researchers demonstrate how they trained a robotic arm to slice food with a knife, bringing the dream of kitchen robots one step closer to becoming a reality.
🧬 Biotechnology
IndieBio’s SF incubator lineup is making some wild biotech promises
In this article, TechCrunch takes a closer look at IndieBio’s 15th cohort of biotech startups. This diverse group includes a startup promising a faster and cheaper method of sequencing DNA, a battery tech startup, a company producing synthetic fertilizer, another working on biomining, and more young companies hoping to solve big challenges with biology.
These Artificial Blood Platelets Could One Day Save Lives
Platelets, crucial for blood clotting, have a short shelf life, making them scarce even when blood donations are plentiful. To address this, researchers have developed artificial platelets. These synthetic platelets mimic natural ones and can be freeze-dried and stored long-term. Initial animal studies have yielded positive results. Human trials are anticipated in a few years to confirm their efficacy and safety.
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!
My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"
Thank you for sharing, Exciting week in tech!
I wonder if the university commencement theme that reportedly aligned with hyped-up Sophia was "Built-in Robotic Obsolescence..."?