Source: Geek Park "The Second 'AI World War'"

Original Authors: Cade Metz, Karen Weise, Nico Grant, and Mike Isaac

Original Title: Self, Fear, and Money: How the Fuse of Artificial Intelligence Was Lit

Image Source: Generated by Unbounded AI

The AI war triggered by large models began 8 years ago, not 1 year ago.

Editor's Note:

In June 1919, the signing of the Treaty of Versailles seemed to mark the official end of the horrific First World War, and a long-lasting peace was on the horizon. Only Marshal Ferdinand Foch of France predicted: This is not peace, but a 20-year armistice. The marshal's words proved prophetic, and in 1939, Germany launched a blitzkrieg against Poland, officially starting the Second World War.
Moving this history of war into the current heated AI war, one can find astonishing similarities between the two. It seems that the AI+large model wave led by OpenAI started last year, but in fact, the seeds of this "war" were sown 8 years ago in the "AI debate" between Musk and Google co-founder Larry Page.
Over the course of 8 years, figures such as Musk, Peter Thiel, Sam Altman, Larry Page, Demis Hassabis, Gates, and Satya Nadella, who are closely tied to the now popular AI companies OpenAI, Anthropic, Microsoft, and DeepMind, formed alliances and ultimately gave birth to the first artificial intelligence war around 2016 and the large model war in 2022, reshaping the landscape of the artificial intelligence war. And driving it all behind the scenes were fear, self, and money.

In July 2015, Elon Musk celebrated his 44th birthday, and his wife hosted a three-day party at a vineyard resort in California, surrounded by family and friends, with children running around upscale residences in Napa Valley.

This was a few years before Twitter became X, and Tesla also had a profitable year. Musk and his wife, Talulah Riley, an actress who played a beautiful but dangerous robot in HBO's sci-fi series "Westworld," were still a year away from their second marriage surrender. Party guest Larry Page was still the CEO of Google. Artificial intelligence had only recently caught the public's attention, being used to identify cats on YouTube with an accuracy of 16%.

After the first night's dinner, as Musk and Page sat by the poolside fire pit, artificial intelligence became a major topic of conversation. These two billionaires had been friends for over a decade, and Musk would occasionally joke that he would crash on Page's couch after playing video games all night.

But that clear night quickly became controversial as the two argued whether artificial intelligence would ultimately elevate or destroy humanity.

As the discussion continued into the cold hours, it became increasingly heated, and some of the 30 or so attendees gathered closer to listen. Mr. Page, who had been plagued by vocal cord issues for over a decade, quietly described his vision of a digital utopia. He said that humans would eventually merge with AI machines. One day, there would be multiple intelligences vying for resources, and the best one would win.

Musk said that if that happened, we were doomed. Machines would destroy humanity.

With a frustrated tone, Mr. Page insisted on pursuing his utopia. Finally, he called Musk a "speciesist," someone who favored humans over future digital life forms.

Musk later said that this insult was "the last straw."

As the crowd dispersed that night, many seemed stunned, even amused, and thought it was just another esoteric debate that often erupted at Silicon Valley parties.

But eight years later, the debate between the two seemed prescient. The question of whether artificial intelligence would elevate or destroy the world—or at least cause serious harm—has sparked ongoing debates among Silicon Valley founders, chatbot users, scholars, legislators, and regulators about whether the technology should be controlled or unleashed.

This debate has pitted some of the world's wealthiest people against each other: Musk, Page, Meta's Mark Zuckerberg, tech investor Peter Thiel, Microsoft's Satya Nadella, and OpenAI's Sam Altman. All are vying for a share of this business (which may one day be worth trillions of dollars) and shaping its power.

At the core of this competition is a perplexing paradox. Those who claim to be most concerned about artificial intelligence are the most determined to create and enjoy the wealth of artificial intelligence. They firmly believe that only they can prevent AI from endangering the earth, thus proving their ambition.

Shortly after that summer gathering, Musk and Page stopped speaking. A few weeks later, Musk, then running a tech incubator, and several researchers dined at a private room in the opulent Rosewood hotel in Menlo Park, California, near the capital office on Sand Hill Road, a popular trading spot for the joint venture.

That dinner led to the founding of a startup called OpenAI later that year. With the support of hundreds of millions of dollars from Musk and other backers, the lab pledged to protect the world from Page's vision.

With its ChatGPT chatbot, OpenAI fundamentally changed the tech industry and introduced the world to the risks and potential of artificial intelligence. According to sources familiar with the company's latest round of financing, OpenAI is valued at over $80 billion, but the collaboration between Musk and Altman did not succeed. The two never spoke again.

"There's disagreement, distrust, and hubris," Altman said. "The closer people are to the same direction, the more controversial the disagreement. You can see this in sects and religious groups. Even the closest people can have fierce arguments."

Last month, this internal strife reached OpenAI's board. Rebel board members attempted to oust Altman, believing they could no longer trust him to build AI for the benefit of humanity. In a tumultuous five days, OpenAI seemed on the verge of collapse until the board relented under pressure from major investors and employees threatening to follow Altman out.

The drama within OpenAI gave the world its first glimpse of the fierce battles between those who will determine the future of artificial intelligence.

But a few years before OpenAI's near-collapse, Silicon Valley witnessed a little-known but fierce competition for control of technology, which is now rapidly reshaping the world, from how children are educated to how wars are waged. The New York Times interviewed over 80 executives, scientists, and entrepreneurs, including the two who attended Musk's birthday party in 2015, to tell this story of ambition, fear, and money.

The Birth of DeepMind

Five years before the Napa Valley party, and two years before the breakthrough with cats on YouTube, 34-year-old neuroscientist Demis Hassabis walked into a cocktail party at Peter Thiel's row house in San Francisco and realized he had made a lot of money. Thiel's living room had a chessboard overlooking the city's Palace of Fine Arts and the swan pond. Dr. Hassabis had been the world's second-best player in the under-14 category.

"I prepared for this meeting for a year," Dr. Hassabis said. "I thought it would be my unique charm: I knew he liked chess."

In 2010, Dr. Hassabis and two colleagues in the UK were looking for funding to start building "general artificial intelligence" (AGI), a machine that could do anything the brain could do. There was little interest in artificial intelligence at the time. After half a century of research, the field of artificial intelligence had failed to produce anything remotely close to the human brain.

Nevertheless, some scientists and thinkers remained concerned about the shortcomings of artificial intelligence. Many, like these three young men from the UK, had connections to internet philosopher and self-taught artificial intelligence researcher Eliezer Yudkowsky. Mr. Yudkowsky was a leader in a group of self-proclaimed rationalists, later known as effective altruists.

They believed that artificial intelligence could find cures for cancer or solve climate change, but they were concerned that AI robots might do things their creators hadn't thought of. Rationalists believed that if machines became smarter than humans, they would betray their creators.

Thiel, through early investments in Facebook and early collaboration with Musk at PayPal, had become very wealthy. He had a strong interest in the singularity, a metaphor from science fiction describing a moment when intelligent technology is no longer under human control.

With Mr. Thiel's support, Mr. Yudkowsky expanded his AI lab and hosted an annual conference on the singularity. A few years earlier, one of Dr. Hassabis's two colleagues had met Mr. Yudkowsky, who secured them a speaking slot at the conference, ensuring they would be invited to Mr. Thiel's gathering.

Mr. Yudkowsky introduced Dr. Hassabis to Mr. Thiel. Many at Dr. Hassabis's party would try to squeeze money out of the host. His strategy was to arrange another meeting. He told Mr. Thiel that there was deep tension between bishops and knights. Both pieces were of equal value, but the best players knew that their strengths were vastly different.

Effective. The next day, Mr. Thiel was captivated and invited everyone back, gathering in the kitchen. Their host had just finished a morning workout, dressed in shiny workout clothes, still sweating. The butler handed him a glass of Diet Coke. The three presented their proposal, and soon Mr. Thiel and his venture capital firm agreed to invest £1.4 million (about $2.25 million) in their startup. He was their first major investor.

They named the company DeepMind, a nod to "deep learning," a way for artificial intelligence systems to learn skills by analyzing large amounts of data; neuroscience; and the supercomputer Deep Thought from the science fiction novel "The Hitchhiker's Guide to the Galaxy." By the fall of 2010, they began building their dream machine. They wholeheartedly believed that because they understood the risks, they were uniquely positioned to protect the world.

"I don't think it's a contradictory position," said Mustafa Suleyman, one of the three founders of DeepMind. "These technologies can bring enormous benefits. Our goal is not to eliminate them or halt their development. Our goal is to mitigate the downsides."

After winning Mr. Thiel's favor, Dr. Hassabis worked to get into Mr. Musk's orbit. About two years later, they met at a conference organized by Thiel's investment fund, which also invested in Musk's company SpaceX. Dr. Hassabis led everyone on a tour of SpaceX headquarters. Later, rocket casings hung from the ceiling, and the two had lunch and talked in the cafeteria.

Musk explained that his plan was to colonize Mars to escape overpopulation and other dangers on Earth. Dr. Hassabis replied that as long as superintelligent machines didn't track and destroy humans on Mars, the plan would work.

Mr. Musk was speechless. He hadn't thought of that particular danger. Mr. Musk soon invested in DeepMind with Mr. Thiel, bringing him closer to the creation of the technology.

With ample funding, DeepMind hired researchers specializing in neural networks and complex algorithms based on brain imaging. Neural networks are essentially massive mathematical systems that take days, weeks, or even months to identify patterns in large sets of numerical data. These systems were first developed in the 1950s and could learn to perform tasks on their own. For example, after analyzing hundreds of envelopes with scribbled names and addresses, they could read handwritten text.

DeepMind further developed this concept. They built a system that could learn to play classic Atari games like "Space Invaders," "Pong," and "Breakout" to demonstrate what was possible.

This caught the attention of another Silicon Valley giant, Google, especially Larry Page. He saw a demonstration of DeepMind's machine playing Atari games. He wanted in.

Talent Auction

In the fall of 2012, 64-year-old University of Toronto professor Geoffrey Hinton and two graduate students published a research paper, showcasing the capabilities of artificial intelligence. They trained a neural network to recognize common objects like flowers, dogs, and cars.

Scientists were amazed at the accuracy of the technology developed by Dr. Hinton and his students. According to three sources, Chinese internet giants offered Dr. Hinton and his students a $12 million signing bonus to join their company.

Dr. Hinton declined, but the money caught his attention.

The British expatriate, educated in Cambridge, had spent most of his career in academia, apart from occasional stints at Microsoft and Google, and was not particularly driven by money. But he had a neurodivergent child, and the money meant economic security.

"We didn't know how much we were worth," Dr. Hinton said. He consulted lawyers and experts on acquisitions and came up with a plan: "We'll organize an auction, and then we'll sell ourselves." The auction would take place during the annual AI conference at Harrah's hotel and casino at Lake Tahoe.

Large tech companies took notice. Google, Microsoft, Baidu, and others began to believe that neural networks were a path to machines that could not only see but also hear, write, speak, and ultimately think.

Mr. Page saw similar technology in his company's AI lab, Google Brain, and believed that Dr. Hinton's research could elevate the work of his scientists. He gave Google Engineering Senior Vice President Alan Eustace a blank check to hire any artificial intelligence expert he needed.

The night before the auction, Brain Lab head Mr. Eustace and Jeff Dean flew to Lake Tahoe, bringing Dr. Hinton and his students to dinner at a steakhouse inside the hotel. Dr. Dean recalled that the smell of old cigarettes was overpowering. They presented their reasons for joining Google.

The next day, Dr. Hinton conducted the auction in his hotel room. Due to an old back injury, he rarely sat down. He turned a trash can upside down on the table, placed his laptop on top, and watched as the bids rolled in for the next two days.

Google made an offer. Microsoft did too. As the price rose, DeepMind quickly dropped out. According to detailed auction documents, industry giants raised the bid to $20 million, then to $25 million. When the price exceeded $30 million, Microsoft dropped out, but re-entered the bidding at $37 million.

"We felt like we were in a movie," Dr. Hinton said.

Then Microsoft dropped out for the second time. Only Baidu and Google remained, pushing the bidding to $42 million, $43 million. Finally, Dr. Hinton and his students stopped the auction at $44 million. The bids were still climbing, but they wanted to work for Google. And the money was staggering.

It was a clear sign that well-funded companies were determined to acquire the most talented artificial intelligence researchers—something Dr. Hassabis, the founder of DeepMind, did not overlook. He had always told his employees that DeepMind would remain an independent company. He believed it was the best way to ensure their technology wouldn't become a dangerous thing.

But as large tech companies entered the talent race, he felt he had no choice: it was time to sell.

According to three sources, by the end of 2012, Google and Facebook were considering acquiring the London lab. Dr. Hassabis and his co-founders insisted on two conditions: DeepMind's technology could not be used for military purposes, and its AGI technology must be overseen by an independent committee of technical experts and ethicists.

Google bid $650 million. Facebook's Mark Zuckerberg offered a higher reward to the founders of DeepMind, but he did not agree to these conditions. DeepMind was sold to Google.

Mr. Zuckerberg was determined to build his own AI lab. He hired French computer scientist Yann LeCun to run the project, who had also done pioneering AI research. A year after Dr. Hinton's auction, Mr. Zuckerberg and Dr. LeCun flew to Lake Tahoe to attend the same AI conference. As Mr. Zuckerberg paced around the suite at Harrah's casino in his socks, he personally interviewed top researchers, who soon received millions of dollars in salaries and stock.

AI was once ridiculed. Now, the richest people in Silicon Valley are spending billions to avoid being left behind.

The Lost Ethics Committee

When Musk invested in DeepMind, he broke his own informal rule—he wouldn't invest in any company he didn't run. The drawbacks of his decision were already apparent, and just about a month after his birthday, he found himself face to face with his former friend and fellow billionaire.

This meeting was on August 14, 2015, the first meeting of the DeepMind Ethics Committee. The committee was established at the insistence of the founders of this startup company to ensure that their technology would not cause any harm after being sold. According to three sources, the members met in a meeting room outside Musk's SpaceX office, with a view of his rocket factory through a window.

But this was the end of Mr. Musk's control. When Google acquired DeepMind, it acquired the entire company. Mr. Musk was out. He was ahead economically, but he was not happy.

Present at the meeting were the three Google executives who now firmly controlled DeepMind: Mr. Page, Mr. Brin, Google's co-founders, and Sergey Brin, and former Google chairman Eric Schmidt. Other attendees included another PayPal co-founder, Reid Hoffman, and Australian philosopher Toby Ord, who researches "existential risks."

The founders of DeepMind reported that they were advancing their work, but they were aware of the serious risks associated with this technology.

DeepMind co-founder Mr. Suleyman delivered a speech titled "The Pitchforkers Are Coming." He told the board that artificial intelligence could lead to a surge in misinformation. He feared that as this technology replaced countless jobs in the coming years, the public would accuse Google of stealing their livelihoods. He believed that Google needed to share its wealth with millions of people who would no longer be able to find work and provide a "universal basic income."

Mr. Musk agreed. But it was clear that his Google guests were not ready to start redistributing their wealth. Mr. Schmidt said he believed these concerns were completely unnecessary. Mr. Page, in his usual low voice, expressed agreement. He believed that the job opportunities created by artificial intelligence would outweigh the job opportunities it took away.

Eight months later, DeepMind made a breakthrough that shocked the AI world and the world. A DeepMind machine called AlphaGo defeated one of the world's best Go players in the ancient game of Go. The match was broadcast live on the internet, with 200 million viewers worldwide. Most researchers believed it would take another 10 years for artificial intelligence to achieve this level of intelligence.

Rationalists, effective altruists, and others concerned about the risks of artificial intelligence claimed that the computer's victory confirmed their fears.

"This once again shows that the pace of development of artificial intelligence is faster than many experts expected," wrote Victoria Krakovna, who was about to join DeepMind as a "AI safety" researcher, in a blog post.

The founders of DeepMind became increasingly concerned about how Google would handle their invention. In 2017, they attempted to break away from the company. Google's response was to increase the salaries and stock rewards for the founders and their employees. They stayed put.

The ethics committee never met for a second time.

Parting Ways

Musk was convinced that Page's optimistic view of artificial intelligence was completely wrong and was angry about losing DeepMind, so he established his own lab.

OpenAI was founded in late 2015, just a few months after he met with Sam Altman at the Fairmont Hotel in Silicon Valley. Musk injected a large amount of funding into the lab, with his former friend Hoffman and Thiel from PayPal also joining. The three and others pledged to invest $1 billion in the project, with 30-year-old Altman helping to run the project. To get them started, they poached Ilya Sutskever from Google. (Dr. Sutskever was one of the "purchased" graduate students in Dr. Hinton's auction.)

Initially, Musk wanted to operate OpenAI as a non-profit organization, free from the economic incentives of Google and other companies. But as Google's AlphaGo stunt left the tech world in awe, Musk had changed his view of how the company should operate. He was eager for OpenAI to invent something that would inspire the world's imagination and close the gap with Google, but as a non-profit organization, it hadn't accomplished that.

According to four sources, at the end of 2017, he devised a plan to seize control of the lab from Altman and the other founders and turn it into a commercial operation, partnering with Tesla and relying on the supercomputer the car company was developing. This was related to the matter.

When Altman and others fought back, Musk resigned and said he would focus on his AI work at Tesla. Three attendees said that in February 2018, he announced his resignation to OpenAI's staff at the top floor of a startup office in a converted truck factory. When he said OpenAI needed to pick up the pace, a researcher at the meeting argued that Mr. Musk was being too rash.

Musk called the researcher a "jerk" and left in a huff with a hefty financial backing.

OpenAI suddenly needed new funding. Altman flew to Silicon Valley and met with Microsoft CEO Satya Nadella. The collaboration seemed natural. Mr. Altman knew Microsoft's Chief Technology Officer Kevin Scott. Microsoft had acquired LinkedIn from OpenAI board member Hoffman. Mr. Nadella told Mr. Scott to make it happen. The deal was completed in 2019.

Mr. Altman and OpenAI formed a for-profit company under the original non-profit organization, with $1 billion in new capital, and Microsoft had a new way to integrate artificial intelligence into its vast cloud computing services.

Not everyone within OpenAI was pleased.

When OpenAI was born, Dario Amodei, a researcher connected to the effective altruism community, was at the Fairmont Hotel. Dr. Amodei, who was constantly twirling his curly hair while speaking, was leading the lab's efforts to build a neural network called the Large Language Model, which could learn from a large amount of digital text. By analyzing countless Wikipedia articles, digital books, and message boards, it could generate text on its own. It also had a habit of fabricating things. It was called GPT-3, and it was released in the summer of 2020.

Researchers from OpenAI, Google, and other companies believed that this rapidly improving technology could be the path to AGI.

But Dr. Amodei was not satisfied with the deal with Microsoft, as he believed it would lead OpenAI in a truly commercial direction. According to five sources, he and other researchers went to the board, attempting to oust Altman. After failing, they left. Like the founders of DeepMind, they were concerned that the new corporate overlords would prioritize commercial interests over safety.

In 2021, a team of about 15 engineers and scientists created a new lab called Anthropic. The project was to build artificial intelligence in a way that effective altruists believed should be done—and with very strict controls.

"The co-founders of Anthropic did not attempt to oust Sam Altman from OpenAI," said Anthropic spokesperson Sally Aldous. "The co-founders themselves concluded that they wanted to leave OpenAI to start their own company, and they communicated this to OpenAI's leadership and negotiated the terms of their exit, which were agreed upon by both sides within a few weeks."

Anthropic accepted a $4 billion investment from Amazon, and two years later, received a $2 billion investment from Google.

Revelation

After OpenAI received another $2 billion from Microsoft, Altman and another executive, Greg Brockman, visited Bill Gates at his lakeside mansion in the outskirts of Seattle, Washington. The Microsoft co-founder was no longer involved in the company's day-to-day affairs but stayed in regular contact with company executives.

During dinner, Mr. Gates told them that he was skeptical about the effectiveness of large language models. He said he would remain skeptical until the technology completed a task that required critical thinking—such as passing an AP biology test.

Five months later, on August 24, 2022, Mr. Altman and Mr. Brockman returned, accompanied by an OpenAI researcher named Chelsea Voss. Ms. Voss had won a medal in the International Biology Olympiad in high school. Nadella and other Microsoft executives were also present at the meeting.

On a large digital display outside Mr. Gates' living room, the OpenAI team demonstrated a technology called GPT-4.

Mr. Brockman administered advanced biology tests with multiple-choice questions, and Ms. Voss graded the answers. The first question involved polar molecules, atomic groups with a positive charge at one end and a negative charge at the other. The system answered correctly and explained its choice. "It's just been trained to provide answers," Mr. Brockman said. "The conversational nature has almost magically disappeared." In other words, it was doing something they hadn't really designed it to do.

There were 60 questions. GPT-4 had only one wrong answer.

Mr. Gates sat up in his chair, wide-eyed. In 1980, when researchers showed him the graphical user interface that would become the foundation of modern personal computing, he had a similar reaction. He believed GPT was revolutionary.

By October, Microsoft had added the technology to its online services, including the Bing search engine. Two months later, OpenAI released the ChatGPT chatbot, which is currently used by 100 million people per week.

OpenAI defeated the effective altruists at Anthropic. Page's optimistic faction at Google hastily released their own chatbot, Bard, but it was widely considered a failure in the competition with OpenAI. Three months after the release of ChatGPT, Google's stock price fell by 11%. Mr. Musk was nowhere to be found.

But this was just the beginning.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。