Ambition, Fear, and Money in the AI Industry
"People who are well aware of the risks of artificial intelligence firmly believe that only by building it themselves can their safety be ensured. However, this belief has sparked deep distrust, leading to an increasingly fierce competition."
In July 2015, Elon Musk celebrated his 44th birthday at a wine resort in California.
This is a three-day party attended only by family and friends, with multiple cabins arranged for children to play on the luxurious grounds of Napa Valley.
At that time, there were still a few years left until Twitter reorganized into X and Tesla, making their first profit.
Musk and his wife Talula Riley (who played a beautiful and dangerous robot in HBO's science fiction drama Westworld) are about to end their second marriage.
One of the guests at the party was Google CEO Larry Page, who was still working at the time.
Artificial intelligence has just begun to attract public attention, and it has been used to identify cats with 16% accuracy on YouTube.
On the first night of the party, Musk and Page had a deep discussion about artificial intelligence by the pool fire pit after dinner.
These two billionaires have been friends for over a decade, and Musk sometimes jokes that he would occasionally stay overnight on Page's sofa after playing video games.
But that night, the debate between the two about whether artificial intelligence promotes human progress or leads to destruction became particularly intense.
As the discussion continued into the late night, the situation became increasingly tense, and more than 30 party guests gathered to listen.
Mr. Page, who has been plagued by vocal cord diseases for a long time, quietly described his digital utopian dream, which is that humans will eventually integrate with AI machines, compete for resources, and survive the strongest.
Musk responded that such development would lead to the extinction of humanity. Mr. Page emphasized the need to pursue this vision and even referred to Musk as a speciesist who favors humanity.
Musk described this insult as his "limit of patience". Many onlookers, although surprised by this, also find it interesting, believing that it is just one of the esoteric discussions that often occur at Silicon Valley parties.
However, eight years later, the debate between these two appears to foresaw the future. The debate over whether AI should enhance the world or destroy it has become a core issue among Silicon Valley entrepreneurs, chatbot users, scholars, and policymakers, involving whether AI technology should be regulated or allowed to develop unchecked.
This debate has put some of the wealthiest people in the world on the opposite side, including Musk, Page, Zuckerberg, Thiel, Nadella, and Ultraman, all vying for a place in this potential trillion dollar market and striving to shape its development direction.
This debate on AI has put some of the world's wealthiest individuals in opposition, including Musk, Page, Meta's Zuckerberg, tech investor Thiel, Microsoft's Nadella, and OpenAI's Ultraman. They are all vying for control over this business area that could be worth trillions of dollars in the future.
The core of competition is an incomprehensible paradox: those who are most concerned about AI are actually the most active in promoting its development and profiting from it. They believe that only they can effectively control AI and avoid its threat to the Earth.
Shortly after the summer party, Musk and Page stopped communicating. A few weeks later, Musk had dinner with Ultraman (who was managing a technology incubator at the time) and some researchers at the Rosewood Hotel in Menlo Park, California, a popular trading location in the venture capital industry.
This dinner led to the establishment of OpenAI, which received billions of dollars in support from Musk and other investors, promising to resist the threats posed by Page's vision.
OpenAI, with its ChatGPT chatbot, has fundamentally changed the technology industry and demonstrated the risks and potential of artificial intelligence to the world.
According to two sources familiar with the company's latest financing situation, OpenAI's valuation has exceeded $80 billion, but the collaboration between Musk and Ultraman has not continued, and the two have no longer communicated.
Ultraman said, "There are issues of disagreement, distrust, and personal interests. When people's goals tend to converge, differences become even sharper. This situation is common in religious groups and denominations, and close people often engage in intense arguments."
Last month, this internal conflict spread to the board of OpenAI. Some board members attempted to dismiss Ultraman because they no longer believed he could build AI beneficial to humanity.
In the midst of five days of chaos, OpenAI faced a crisis of disintegration until the board finally gave in under pressure from large investors and employees threatening to leave with Ultraman.
The internal dispute within OpenAI has shown the world for the first time the fierce struggle between the people who determine the future of AI.
Many years before the collapse of OpenAI, Silicon Valley had already engaged in a little-known but highly intense competition for control of technology that is rapidly changing the world today. These technologies have changed everything from child education to the way wars are fought.
The New York Times interviewed over 80 executives, scientists, and entrepreneurs, including two who attended Musk's birthday party in 2015. Together, they shared a story about ambition, fear, and money.
The Origin of DeepMind
Five years before the Napa Valley party and two years before the breakthrough in cat recognition on YouTube, 34 year old neuroscientist Demis Hassabis attended a cocktail party hosted by Peter Thiel at his townhouse in San Francisco and realized he had found an opportunity.
In Tyr's living room, you can see the city's art palace, swan pool, and a chessboard. Hassabis was once the second player in the global under-14 group.
"I prepared for that meeting for a year," Hassabis recalled. "I think this will be a unique way for me to attract his attention: I know he enjoys playing chess."
In 2010, Hassabis and his two colleagues in the UK were searching for funding to build "General Artificial Intelligence" (A.G.I.), a machine that could do anything that a human brain could.
At that time, after half a century of research, the field of artificial intelligence failed to produce any results close to the human brain, so few people were interested in it.
However, some scientists and thinkers are increasingly concerned about the potential risks of artificial intelligence. Like these three young British people, many have connections with internet philosopher and self-taught artificial intelligence researcher Elise Yudokorsky. Yudokorsky was the leader of a community that referred to itself as "rationalists" or later as "effective altruists".
They believe that artificial intelligence can find ways to cure cancer or address climate change issues, but at the same time, they are concerned that AI robots may exhibit behaviors that exceed the expectations of creators. Rationalists believe that if machines become smarter than humans, they may turn against creators.
Til became very wealthy by investing in Facebook early on and partnering with Musk on PayPal. He developed a strong interest in "singularity" - the moment in science fiction that describes intelligent technology beyond human control.
With funding from Thiel, Yudokorsky expanded his artificial intelligence laboratory and founded an annual conference on singularities. Prior to this, one of Dr. Hassabis's two colleagues met Yudokorsky and ensured that they had the opportunity to speak at the meeting, which led to an invitation to attend Thiel's party.
Yudokorsky introduced Dr. Hassabis to Tyr. Hasabis realized that many people at the party may be seeking investment from the host, so he planned to arrange another meeting.
He proposed an insight to Thiel: in chess, although elephants and horses have equal value, there are profound differences. He explained that although these two pieces have the same score, the best player understands that their respective strengths are completely different.
This method was successful. Tyr was attracted to them and invited them to his home again the next day, where everyone gathered in the kitchen. Tyr has just finished his morning exercise and is still wearing wet sportswear. He took a can of Diet Coke handed over by the butler.
These three people introduced their plan, and soon after, Thiel and his venture capital firm decided to invest £ 1.4 million (approximately $2.25 million) in their startup, becoming their first major investor.
They named the company DeepMind, not only as a tribute to "deep learning" (a method of AI systems learning skills by analyzing large amounts of data), but also as a reference to neuroscience and the "deep thinking" supercomputer in the science fiction novel "Hitchhiker's Guide to the Galaxy".
In the autumn of 2010, they began to realize their dream machine. They firmly believe that it is precisely because they have a deep understanding of the risks involved that they have unique insight and can protect the world.
"I don't think this is a contradictory position," said Mustafa Suleiman, one of DeepMind's three founders. "These technologies will bring enormous benefits. Our goal is not to stop or suspend their development, but to reduce the potential negative impacts they may bring."
After gaining Thiel's support, Dr. Hassabis began to approach Musk. Two years later, they met at a conference organized by Thiel Investment Fund, which also invested in Musk's SpaceX company.
Hasabis was given the opportunity to visit SpaceX headquarters. Afterwards, they dined and had a conversation in a restaurant decorated with rocket shells.
Musk shared his plan to colonize Mars in order to escape the overpopulation and other dangers on Earth. Hassabis replied that this plan is feasible - as long as super intelligent machines do not follow them to Mars and destroy humans.
This viewpoint left Musk speechless, as he had not previously considered this particular danger. Shortly thereafter, Musk and Thiel invested in DeepMind together, allowing him to get closer to the development of this technology.
After sufficient funding, DeepMind began recruiting researchers specializing in neural networks. Neural networks are complex algorithms that mimic the construction of the human brain. Essentially, they are large mathematical systems that can take days, weeks, or even months to identify patterns in large amounts of data.
These systems were originally developed in the 1950s and were able to learn to handle tasks autonomously, such as recognizing handwritten text after analyzing hundreds of handwritten names and addresses on envelopes.
DeepMind has taken this concept to a new level. They built a system that could learn to play classic Atari games such as Space Invader, Table Tennis, and Brick Fighter to showcase their potential.
This has attracted the attention of another Silicon Valley giant, Google, especially Larry Page. He saw a demonstration of DeepMind machines playing Atari games and wanted to join in.
Talent Competition
In the autumn of 2012, 64 year old Professor Jeffrey Hinton and two of his graduate students from the University of Toronto published a research paper showcasing the capabilities of artificial intelligence. They trained a neural network to recognize common objects such as flowers, dogs, and cars.
Scientists were surprised by the accuracy of the technology developed by Dr. Hinton and his students. Of particular interest is Yu Kai, an artificial intelligence researcher who met Dr. Xinton at a research conference. He recently joined China's internet giant Baidu.
It is understood that Baidu has offered a bid of $12 million to invite Dr. Sinton and his students to join its Beijing team.
Although Dr. Sinton rejected Baidu's proposal, this money still caught his attention. This British immigrant who received education in Cambridge has spent most of his career in academia, occasionally working at Microsoft and Google, and money is not the driving force for him. But he has a child with neurological diversity, so this money will mean economic security.
"We don't know how much we're worth," Dr. Sinton said. He consulted with lawyers and acquisition experts and decided to adopt a strategy: "We will organize an auction to sell ourselves." This auction will take place during the annual artificial intelligence conference held at the Harras Hotel in Lake Tahoe.
Large technology companies are beginning to notice this trend. Companies such as Google, Microsoft, and Baidu are beginning to realize that neural networks may be a key way for machines to not only see, listen, write, speak, but also ultimately think.
Page saw similar technologies in Google Brain's artificial intelligence laboratory and believes that Dr. Hinton's research can enhance the work of his scientific team. He authorized Google's Senior Vice President, Alan Eustace, to recruit artificial intelligence experts in any necessary way.
Eustace and Jeff Dean, who is in charge of the Brain laboratory, flew to Lake Tahoe and had dinner with Dr. Hinton and his students at the hotel steakhouse the night before the auction. Dean recalled that the smell of old smoke was very strong. They demonstrated the advantages of working at Google to Dr. Hinton.
The next day, Dr. Hinton organized an auction in his hotel room. Due to an old back injury, he rarely sits. He inverted a trash can on the table, placed his notebook on top, and began observing the bids for the next two days.
Google provided a quote, and Microsoft also participated in the bidding. As prices rose, DeepMind withdrew from competition. The giants have pushed up their bids to $20 million and $25 million. According to the auction documents, Microsoft withdrew when the price exceeded $30 million, but rejoined the competition at $37 million.
"We felt like we were in a movie," Dr. Hinton recalled.
After Microsoft withdrew again, only Baidu and Google were left to bid, with prices pushed up to $42 million and $43 million. In the end, when the bid reached $44 million, Dr. Hinton and his students decided to stop the auction. Although the bid is still rising, they are more willing to join Google, and the amount is huge.
This clearly indicates that financially strong companies are determined to acquire the best artificial intelligence research talents, as Dr. Hassabis also recognized in DeepMind. He has always promised his employees that DeepMind will remain independent, and he believes this is the best way to ensure that technology does not go dangerous.
But as large technology companies join the talent competition, he believes he has no choice but to sell the company.
By the end of 2012, both Google and Facebook were attempting to acquire this London laboratory. Dr. Hassabis and his co founders insist on two conditions: DeepMind's technology cannot be used for military purposes, and its general artificial intelligence technology must be supervised by an independent technology and ethics committee.
Google bid $650 million. Facebook's Mark Zuckerberg offered a higher quote, but did not agree to these conditions. In the end, DeepMind chose Google.
Zuckerberg decided to create his own AI laboratory and hired French computer scientist Jan Le Kun, who had done pioneering research in the field of artificial intelligence, to operate it.
A year after Dr. Sinton auctioned off, he and Dr. Le Kun went to Lake Tahoe to attend the same AI conference. Zuckerberg walked in his suite at the Haras Casino wearing only socks and personally interviewed some top researchers, who were soon offered millions of dollars in salary and stocks.
Artificial intelligence, once seen as a joke, is now causing Silicon Valley billionaires to invest billions of dollars to avoid falling behind.
The disappearing ethics committee
When Musk invested in DeepMind, he broke an informal rule of not investing in companies he did not personally manage. The negative impact of this decision became apparent shortly after his dispute with Page at the birthday party, when he faced his former friend and billionaire again.
Their reunion took place on August 14, 2015 at the first meeting of the DeepMind Ethics Committee. This committee was established at the request of the founder to ensure that their technology will not cause harm after sale. According to three informed sources, committee members gathered in the conference room outside Musk SpaceX's office, with his rocket manufacturing plant outside the window.
Musk's influence came to an end here. Google's acquisition of DeepMind means he will no longer be involved in company affairs. Although he has achieved success from a financial perspective, he is not satisfied with it.
Among Google executives who control DeepMind are Page, Google co-founder and Tesla investor Sergey Brin, and Google Chairman Eric Schmidt. There were other attendees at the meeting, such as Reed Hoffman, another founder of PayPal, and Australian philosopher Toby Ode, who studies "existential risk.".
The founders of DeepMind reported on their ongoing work and emphasized their awareness of the serious risks hidden in this technology.
DeepMind co-founder Suleiman gave a speech at the board of directors titled "The Uprising of the People is Coming". He pointed out that artificial intelligence may lead to a significant increase in false information. He also expressed concerns that this technology will replace a large number of jobs in the coming years, and that the public may accuse Google of depriving them of their livelihoods. Suleiman believes that Google should share its wealth with millions of people who cannot find jobs and support the implementation of a "basic income for all" policy.
Musk agrees with this point. But it is clear that his Google guests do not intend to distribute their wealth. Mr. Schmidt believes that these concerns have been exaggerated, and Mr. Page agrees with them in his usual low voice. Page believes that artificial intelligence will create more jobs than it eliminates.
Just eight months later, DeepMind achieved a shocking breakthrough that shook the entire artificial intelligence industry and the world. DeepMind's AlphaGo machine defeated one of the world's top Go players. This game attracted approximately 200 million viewers worldwide through live streaming on the internet.
Previously, most researchers believed that artificial intelligence would take at least another 10 years to demonstrate this creativity.
Rationalists, effective altruists, and other individuals concerned about the risks of artificial intelligence claim that the victory of this computer confirms their concerns.
"This indicates that the development speed of artificial intelligence has even exceeded the expectations of many experts," Victoria Krakowna, who is about to join DeepMind as a "AI Security" researcher, wrote in a blog post.
The founders of DeepMind are becoming increasingly concerned about how Google will leverage their inventions. In 2017, they attempted to become independent from Google, which responded by increasing the salaries and stock rewards of DeepMind founders and their team. In the end, they chose to stay at Google.
The ethics committee has never held a second meeting.
division
Doubts about Page's optimistic views on artificial intelligence and dissatisfaction with losing DeepMind prompted Musk to establish his own laboratory.
OpenAI was founded at the end of 2015, following his meeting with Sam Ultraman at the Rosewood Hotel in Silicon Valley. Musk provided financial support for this laboratory, and his former PayPal colleagues Hoffman and Thiel also joined in. They and others promised to invest $1 billion in the project. At the age of 30, Ultraman will be involved in managing this project. As a starting point, they poached Ilya Sutcliffe from Google (Sutcliffe was one of the graduate students acquired by Google in the Sinton auction).
At first, Musk hoped to run OpenAI as a non-profit organization to avoid being influenced by economic motives of companies like Google. But after Google shook the tech industry with its achievements in Go, Musk began to rethink the way OpenAI operates. He eagerly hoped that OpenAI could invent products that could attract global attention and narrow the gap with Google, but as a non-profit organization, OpenAI failed to achieve this goal.
By the end of 2017, Musk plans to take over the laboratory from Ultraman and other founders and transform it into a commercial company, collaborating with Tesla to utilize Tesla's developed supercomputers, according to four insiders.
When Ultraman and others opposed it, Musk chose to withdraw and focus on Tesla's AI work. In February 2018, he announced his departure decision to employees at the top level of the OpenAI office. When he stated that OpenAI needed to accelerate its pace, a researcher countered at the meeting that Musk was too reckless.
Musk referred to the researcher as a "foolish person" and angrily left, taking away his funds.
OpenAI urgently needs new sources of funding. Ultraman went to Sengu to attend a meeting and coincidentally met Microsoft CEO Nadella. Collaboration seems to be a natural choice.
Ultraman knows Microsoft's Chief Technology Officer Kevin Scott, who has previously acquired LinkedIn from OpenAI board member Hoffman. Nadella instructed Scott to complete this transaction. The transaction was completed in 2019.
Ultraman and OpenAI established a profitable company under their initial non-profit organization with $1 billion in new funding, while Microsoft integrated artificial intelligence into their extensive cloud computing services through this approach.
However, not everyone within OpenAI is satisfied with this.
Researcher Dario Amode, who is associated with the effective altruistic community, participated in the establishment of OpenAI at the Rosewood Hotel. He has been leading the development of a neural network called a large-scale language model in the laboratory, which can learn from a large amount of digital text. By analyzing countless Wikipedia articles, digital books, and forum posts, it can generate its own text. However, it also has a tendency towards misleading information. This model, known as GPT-3, was released in the summer of 2020.
Researchers from OpenAI, Google, and other companies believe that this rapidly advancing technology may be a way to achieve universal artificial intelligence.
But Amody is dissatisfied with the deal with Microsoft, as he believes it has pushed OpenAI towards extreme commercialization. He and other researchers attempted to pressure the board to step down Ultraman, according to five people familiar with the matter. They chose to leave after their attempt failed. Like the founders of DeepMind before, they were concerned that new company managers would prioritize business interests over security.
In 2021, a team of approximately 15 engineers and scientists established a new laboratory called Anthropic. Their goal is to develop artificial intelligence through strict control in a way that is perceived by effective altruists.
"The co founders of Antitopic did not attempt to remove Sam Ultraman from OpenAI," said Antitopic spokesperson Sally Aldos. "The co founders themselves decided to leave OpenAI and start their own company. They have made this decision clear to OpenAI's leadership and negotiated a mutually satisfactory resignation plan within a few weeks."
Two years later, Anthropic received $4 billion in investments from Amazon and $2 billion from Google.
Reveal
After OpenAI received an additional $2 billion investment from Microsoft, Ultraman and another executive, Greg Brockman, went to the shores of Lake Washington outside Seattle to visit Bill Gates' magnificent mansion there. Although Gates no longer participates in Microsoft's operations on a daily basis, he still maintains regular contact with the company's executives.
During dinner, Gates expressed doubts about the effectiveness of large-scale language models to them. He stated that unless this technology can accomplish a task that requires critical thinking, such as passing an advanced biology exam, he will remain skeptical.
Five months later, on August 24, 2022, Ultraman and Brockman visited Gates again and brought along an OpenAI researcher named Chelsea Voss. Voss won a medal in the International Biology Olympiad in high school. Nadella and other Microsoft executives also attended.
On a huge digital display screen outside Gates' living room, the OpenAI team showcased a technology called GPT-4 to the attendees.
Brockman had the GPT-4 system participate in a multiple choice test in advanced biology, rated by Voss. The first question involves polar molecules, which are atomic clusters with one end positively charged and the other end negatively charged. The system not only correctly answered this question, but also explained its choice. "It was originally only trained to provide answers," Brockman explained. "But the characteristics of dialogue are almost as natural as magic." That is to say, the system has shown some abilities that the designer had not expected.
There are a total of 60 issues, and GPT-4 only made one mistake.
Gates sat up straight in the chair, his eyes wide open. He had a similar reaction when he saw the graphical user interface that would become the foundation of modern personal computers in 1980. He believes that GPT technology has the same revolutionary value.
In October, Microsoft began integrating this technology into its online services, including the Bing search engine. Two months later, OpenAI released its ChatGPT chatbot, which is now used by 100 million people every week.
OpenAI has taken the lead in the competition with effective altruists in Antioptics. Google's optimists are eager to launch their own chatbot Bard, but it is widely believed to have fallen behind OpenAI. Three months after ChatGPT was released, Google's stock price fell by 11%. And Musk has disappeared without a trace.
But this is just the beginning.
