Musk failed many times, OpenAI suddenly emerged, and the AI ​​war in Silicon Valley started ten years ago

Editor's note: "Those who claim to be most worried about the crisis that artificial intelligence will bring are actually the ones who are most determined to create it and enjoy the wealth it brings."

This feature report from the New York Times traces back the Silicon Valley AI battle that has been brewing since 2010:

1. Musk and Google co-founder Larry Page "fell out" at a party in 2015. The two had opposing attitudes towards the development of AI.

2. DeepMind founder Demis Hassabis first used chess to get Peter Thiel’s attention. With Thiel's support, DeepMind gained Musk's attention and investment and gradually grew.

3. "AI Godfather" Jeffrey Hinton held an "auction", and what he "sold" was himself and his team. In the end, Google successfully acquired Dr. Hinton's team.

4. This auction symbolizes the official entry of Silicon Valley giants into the AI ​​war. DeepMind "retired" and decided to sell.

5. After the sale, DeepMind originally had an "Ethics Committee" to ensure the safe development of AI, but this committee eventually became ineffective.

6. OpenAI breaks with Musk and moves towards cooperation with Microsoft.

7. Dario Amodei of OpenAI left to found Anthropic.

8. OpenAI passed Bill Gates’ “examination”, and GPT was regarded by the latter as a revolutionary innovation.

Original address: https://www.nytimes.com/2023/12/03/technology/ai-openai-musk-page-altman.html

Original author: Cade Metz, Karen Weise, Nico Grant & Mike Isaac

In July 2015, Elon Musk celebrated his 44th birthday with a three-day party thrown by his wife at a California wine country resort dotted with cabins. Only family and friends were invited, and the kids ran and played around the upscale Napa Valley property.

This was still a few years away from Twitter becoming X and Tesla becoming profitable. Mr Musk and his wife Talulah Riley – the actress who plays the beautiful but dangerous robot in the HBO sci-fi drama "Westworld" – are still close to the end of their second marriage. one year. The guest at the party was Larry Page, then CEO of Google. And artificial intelligence only started to become known to the public a few years ago, when it was used to identify cats on YouTube – with 16% accuracy.

When Musk and Page chatted by the poolside campfire after dinner that first night, artificial intelligence was the topic of their conversation. The two billionaires have been friends for more than a decade, with Mr. Musk sometimes joking that he would occasionally spend the night on Mr. Page's couch after an all-night gaming session.

But the atmosphere on that bright night quickly turned tense as the two began to debate whether artificial intelligence would ultimately enhance or destroy humanity.

As the discussion stretched into the cool night, the atmosphere became increasingly tense, and some of the 30-something partygoers had gathered to listen closer. Page describes his vision of a digital utopia in whispers, after suffering from an unusual condition that has affected his vocal cords for more than a decade. Humans will eventually merge with artificially intelligent machines, he said. One day, there will be many kinds of intelligence competing for resources, and the best will win.

If that happens, Musk said, we're screwed. Machines will destroy humanity.

With a croak of frustration, Page insists that we should pursue the utopia he describes. Finally, he called Musk a "speciesist," someone who favors humans over future digital life forms.

Musk later said that humiliation was "the last straw."

As the crowd dispersed, many looked stunned but also amused, assuming it was just the kind of esoteric debate that often occurs at Silicon Valley parties.

Eight years later, however, the debate between the two men looks prescient. Whether artificial intelligence will enhance the world or destroy it—or at least cause serious harm—has formed a debate among Silicon Valley entrepreneurs, chatbot users, academics, lawmakers, and regulators about whether we should control or let it go. A framework for ongoing debate on development.

That debate pitted a group of the world's richest people against each other: Musk, Page, Meta's Mark Zuckerberg, tech investor Peter Thiel, Microsoft's Satya Satya Nadella and OpenAI’s Sam Altman. They are all vying for a piece of a business that may one day be worth trillions of dollars, and for the power to shape it.

At the heart of this competition is a puzzling paradox. Those who claim to be most worried about artificial intelligence are the ones most determined to create it and enjoy the wealth it will bring. They justify their ambitions with a firm belief that only they can prevent artificial intelligence from posing a danger to the planet.

Shortly after that summer party, Musk and Page lost contact. A few weeks later, Musk had dinner with Altman, who was running a tech incubator at the time, and several researchers in a private room at the Rosewood Hotel in Menlo Park, California. This is a pretty popular place to chat about collaborations because it's close to the VC offices on Sand Hill Road.

That dinner led to the founding of a startup called OpenAI later that year. Backed by hundreds of millions of dollars from Musk and other investors, the lab promises to protect the world from Page's vision.

▲ Musk and Page split due to differences in AI concepts

Thanks to its ChatGPT chatbot, OpenAI has fundamentally changed the technology industry and demonstrated the risks and potential of artificial intelligence. OpenAI was valued at more than $80 billion, according to two people familiar with the company's latest funding round, although Musk and Altman's partnership did not last. Since then, the two have stopped talking to each other.

"There's division, distrust, egotism," Altman said. "The more people move in the same direction, the sharper the divisions. You see this in sects and religious groups as well. It's those who are closest to each other that have the most intense fights."

Last month, the infighting reached OpenAI's board of directors. Rebellious board members tried to force Ultraman out because they believed they could no longer trust him to build AI that could benefit humanity. For five days of chaos, OpenAI looked set to collapse, until the board backed down under pressure from key investors and employees who threatened to follow Altman out.

The dramatic battle within OpenAI has given the world its first glimpse into the raging battle between those who will determine the future of AI.

But a few years before OpenAI’s near-debacle, there was a little-known but fierce competition in Silicon Valley for technologies that are now rapidly reshaping the world. It is rewriting everything from how we educate our children to How to fight a war. The New York Times interviewed more than 80 executives, scientists and entrepreneurs, including two who attended Musk's 2015 birthday party, to tell this story of ambition, fear and money.

The birth of DeepMind

Five years before that Napa Valley party, and two years before the success of cat videos on YouTube, 34-year-old neuroscientist Demis Hassabis walked into Peter Thiel’s Over a cocktail party at a San Francisco townhouse, he realized he had struck gold. In Tierna's living room, which overlooks the local Palais des Arts and the Swan Pond, there is a chessboard. Dr. Hassabis was once the world's second-ranked chess player in the under-14 category.

"I prepared for that meeting for a year," Dr. Hassabis said. "I thought that would be my unique entry point: I knew he loved playing chess."

▲ Damis Hassabis

In 2010, Dr. Hassabis and two of his colleagues, who were living in the United Kingdom, were looking for funding to begin building an "artificial general intelligence," or AGI, a machine that could do anything a brain can do. At the time, few people were interested in artificial intelligence. After half a century of research, the field of artificial intelligence has failed to produce anything close to the human brain.

However, some scientists and thinkers have developed stubborn views on the negative impacts of AI. Many, like these three young men from the UK, are related to the internet philosopher and self-taught AI researcher Eliezer Yudkowsky. Mr. Yudokowsky was a community leader who called himself a "rationalist," who later became known as an "effective altruist."

They believe AI can find a cure for cancer or solve climate change, but they worry that AI robots might do things that go against the intentions of their creators. If machines become smarter than humans, rationalists argue, the machines may turn against their creators.

Thiel amassed a fortune through his early investments in Facebook and his early partnership with Musk at PayPal. He became fascinated by the singularity, a theme in science fiction that describes the moment when intelligent technology can no longer be controlled by humans.

With Thiel's funding, Yudokovsky expanded his AI lab and created an annual conference on the Singularity. Years ago, one of Dr. Hassabis's two colleagues met Yudokowsky. He got them a chance to speak at meetings and made sure they were invited to Till's parties.

Mr. Yudkowsky introduced Dr. Hassabis to Thiel. Hassabis thought there would be many people at the party trying to squeeze money out of Thiel. His strategy was to arrange another meeting. He told Teal that relations between the bishop and the knight were tense. The two pieces have the same value, but the best players understand that their advantages are very different.

▲ Peter Thiel

This worked. Charmed, Teal invited them back the next day to meet in the kitchen. The host had just finished his morning workout and his body was glistening with sweat. A steward handed him a Diet Coke. The three of them pitched their pitch, and soon Thiel and his venture capital firm agreed to invest £1.4 million (about $2.25 million) into their startup. He was their first significant investor.

They named the company DeepMind, a nod to "deep learning" (a way for AI systems to learn skills by analyzing large amounts of data), neuroscience, and the deep-thinking supercomputer featured in the science fiction novel "The Hitchhiker's Guide to the Galaxy." By the fall of 2010, they were building their dream machine. They firmly believe that they are uniquely positioned to protect the world because they understand the risks.

“I don’t think it’s a contradictory position,” said Mustafa Suleyman, one of DeepMind’s three founders. "These technologies will bring huge benefits. The goal is not to eliminate or suspend their development. The goal is to reduce their negative impacts."

After winning Thiel's support, Dr. Hassabis gradually entered Musk's field of vision. They met about two years later at a conference organized by Thiel's investment fund, which also invested in Musk's company, SpaceX. Dr. Hassabis arranged a tour of SpaceX headquarters. Afterwards, the two had lunch and talked in the cafeteria with rocket shells hanging from the ceiling.

Mr. Musk has explained that his plan is to colonize Mars as a way to escape overpopulation and other dangers on Earth. Dr. Hassabis responded that this plan will work – provided that super-intelligent machines do not follow humans to Mars and destroy humans on Mars.

Musk was speechless. He had never considered this particular danger. Musk quickly invested in DeepMind with Mr. Thiel to be closer to the creation of the technology.

With ample funding, DeepMind employs researchers who specialize in neural networks, complex algorithms modeled after the human brain. A neural network is essentially a giant mathematical system that spends days, weeks, or even months identifying patterns in vast amounts of data. Originally developed in the 1950s, these systems can learn to handle tasks on their own. For example, after analyzing hundreds of handwritten names and addresses on envelopes, they can read handwritten text.

DeepMind takes this concept further. It built a system that could learn to play classic Atari games like Space Invaders, Pong, and Breakout to demonstrate its potential.

This attracted the attention of Google, another giant in Silicon Valley, especially Larry Page. He saw a demonstration of DeepMind's machine playing an Atari game. He wanted to be a part of it.

Talent auction

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

Scientists were surprised by the accuracy of the technology built by Dr. Hinton and his students. One person who is particularly concerned about this is Yu Kai, an artificial intelligence researcher who met Dr. Hinton at a research conference and recently started working for the Chinese Internet giant Baidu. Baidu offered Dr. Hinton and his students $12 million to join the Beijing-based company, according to three people familiar with the matter.

Dr. Hinton turned down Baidu, but the money caught his attention.

▲ Jeffrey Hinton

The Cambridge-educated British expatriate, who has spent most of his time in academia, apart from occasional stints at Microsoft and Google, doesn't particularly value money. But his children are affected by neurodiversity, and the money would mean financial security.

"We don't know what our value is," Dr. Hinton said. He consulted with lawyers and acquisition experts, then came up with a plan: "We're going to organize an auction, and we're going to sell ourselves." The auction will take place at the annual Artificial Intelligence Conference at Harrah's Hotel and Lake Tahoe Casino. conduct.

Big tech companies are starting to take notice. Companies such as Google, Microsoft, and Baidu have begun to believe that the machines that neural networks will bring will not only have the ability to see, but also listen, write, speak, and eventually – even think.

Page had seen similar technology in his company's artificial intelligence lab, Google Brain, and thought Dr. Hinton's research could improve the work of his scientists. He essentially gave Alan Eustace, Google's senior vice president and engineer, an unnumbered check to hire any AI experts he needed.

Mr. Eustace and Jeff Dean, who runs Brain Labs, flew to Lake Tahoe and arranged to have dinner with Dr. Hinton and his students at a hotel steakhouse the night before the auction. Dr. Dean recalled that there was a strong smell of old cigarette smoke. They persuaded Dr. Hinton and his students to work at Google.

The next day, Dr. Hinton conducted the auction in his hotel room. He rarely sits due to an old injury. He turned a trash can upside down on a table, placed his laptop on top, and watched the bids roll in over the next two days.

Google made an offer. So does Microsoft. As prices rose, DeepMind quickly exited. The industry giants pushed the bid up to $20 million and then $25 million, according to auction details. Microsoft withdrew when the price exceeded $30 million, but rejoined the bidding at $37 million.

"We felt it was like a movie," Dr. Hinton said.

▲ Jeffrey Hinton and his two star students Ilya Sutskvi and Alex Krizhevsky

Then, Microsoft withdrew from the bidding for a second time. Only Baidu and Google were left, and they pushed their bidding prices to US$42 million and US$43 million. Finally, at $44 million, Dr. Hinton and his students terminated the auction. Although the bids are still rising, they want to work for Google. And the amount of money is staggering enough.

It’s a clear sign that deep-pocketed companies are determined to buy up the most talented AI researchers — something DeepMind’s Dr. Hassabis also realizes. He has been telling his employees that DeepMind will remain an independent company. He believes this is the best way to ensure his technology doesn't turn into something dangerous.

But when big tech companies entered the competition for talent, he decided he had no choice: It was time to sell.

By the end of 2012, Google and Facebook were seeking to acquire the London lab, according to three people familiar with the matter. Dr. Hassabis and his co-founders insisted on two conditions: DeepMind’s technology could not be used for military purposes and its AI technology must be overseen by an independent board of technologists and ethicists.

Google's bid was $650 million. Facebook's Mark Zuckerberg offered DeepMind's founder a higher price but disagreed with the terms. DeepMind was eventually sold to Google.

Zuckerberg decided to build his own artificial intelligence laboratory. He hired Yann LeCun, a French computer scientist who had also done groundbreaking artificial intelligence research, to run the lab. A year after Dr. Hinton’s auction, Zuckerberg and Dr. LeChun flew to Lake Tahoe to attend the same AI conference. In a suite at Harrah's Casino, Mr. Zuckerberg walked around in his socks, personally interviewing top researchers who were quickly offered millions of dollars in salary and stock.

Artificial intelligence was once ridiculed. Now Silicon Valley's richest people are spending billions to prevent themselves from being left behind in the race.

The missing ethics committee

When Musk invested in DeepMind, he broke an unofficial rule he wouldn't invest in any company he didn't run. The negative impact of this decision became apparent just a month or so after his birthday row – when he was faced with his former friend and fellow billionaire.

▲ Members of the Ethics Committee: Sergey Brin, Eric Schmidt, Larry Page

This occasion was the first meeting of the DeepMind Ethics Committee, held on August 14, 2015. At the insistence of the founders, the committee was formed to ensure that their technology would cause no harm if sold. Committee members gathered in a conference room outside Musk's SpaceX office, with windows overlooking his rocket factory, according to three people familiar with the meeting.

However, this is also where Musk's control ends. When Google acquired DeepMind, it bought the entire company. Musk was pushed out. Although he was making money financially, he was not happy.

Three Google executives now have firm control of DeepMind: Page; Google co-founder and Tesla investor Sergey Brin; and Google Chairman Eric Schmidt (Eric Schmidt). Other attendees included Reid Hoffman, another PayPal founder, and Toby Ord, an Australian philosopher who studies existential risk.

DeepMind's founders report that they are moving forward with their work, but they are aware that the technology carries serious risks.

Suleiman, co-founder of DeepMind, gave a speech titled "The Rising People Are Coming". He told the board that artificial intelligence could lead to an explosion of disinformation. He worries that in the coming years, as the technology displaces countless jobs, the public will blame Google for taking away their livelihoods. He believes that Google needs to share its wealth with the millions of unemployed people and provide a "universal basic income."

Musk agrees. But it's clear that his Google guests have no intention of redistributing wealth. Mr Schmidt said he believed the concerns were completely exaggerated. Mr. Page agreed in his usual low voice. He argued that AI would create more jobs than it takes away.

Eight months later, DeepMind achieved a breakthrough that shocked the AI ​​community and the world. A DeepMind machine called AlphaGo defeated one of the best Go players in the world. The game was broadcast live over the Internet and was watched by 200 million people around the world. Most researchers had thought it would take another decade for AI to have the innovative capabilities to do this.

Rationalists, effective altruists, and others worried about the risks of artificial intelligence claim that the computers’ triumph confirms their fears.

"This is another sign that artificial intelligence is advancing faster than many experts expected," Victoria Krakovna, who will join DeepMind as an "AI safety" researcher, wrote in a blog post.

DeepMind's founders grew increasingly concerned about how Google would use their invention. In 2017, they tried to separate themselves from the company. In response, Google increased the salaries and stock awards of DeepMind's founders and their employees. They chose to stay.

The ethics committee never met a second time.

separate

Convinced that Page's optimism about artificial intelligence was completely misplaced and angry about losing DeepMind, Musk built his own lab.

OpenAI was founded in late 2015, just months after he and Sam Altman met at the Rosewood Hotel in Silicon Valley. Musk poured money into the lab, and his former PayPal partners, Messrs. Hoffman and Thiel, joined the effort. The trio and others pledged $1 billion to the project, which Altman, then 30, would help run. To get the project started, they hired Ilya Sutskever from Google. (Dr. Sutskwei was one of the graduate students Google “purchased” at Dr. Hinton’s auction.)

Initially, Musk wanted to run OpenAI as a non-profit, away from the financial interests driven by companies like Google. But while Google is shocking the tech world with its Go performance, Musk is changing his mind about how OpenAI should operate. He desperately wanted OpenAI to invent something that would capture the imagination of the world and bring them closer to Google, but as a nonprofit, it wasn't able to get the job done.

In late 2017, he hatched a plan to try to wrest control of the lab from Altman and the other founders and turn it into a commercial operation, joining forces with Tesla and relying on the supercars the car company was developing. computer company, according to four people familiar with the matter.

When Altman and others objected, Musk opted out and said he would focus on his own artificial intelligence work at Tesla. He announced his departure to OpenAI employees in February 2018 on the top floor of the startup's offices in a converted truck factory, three people who attended the meeting said. When he said OpenAI needed to move faster, a researcher at the meeting countered that Mr. Musk's behavior was reckless.

Musk called the researcher a "stupid" and stormed off, taking his vast fortune with him.

OpenAI suddenly needed new funding. Altman flew to Sun Valley for a conference and bumped into Microsoft CEO Satya Nadella. The union of the two seemed natural. Altman knew Kevin Scott, Microsoft's chief technology officer. Microsoft acquired LinkedIn from OpenAI board member Hoffman. Nadella tasked Scott with getting it done. The deal was completed in 2019.

Altman and OpenAI, which formed a for-profit company under the original nonprofit umbrella, have $1 billion in new capital and Microsoft has found a new way to incorporate artificial intelligence into its massive cloud computing service.

But not everyone at OpenAI is happy.

Researcher Dario Amodei is associated with the Effective Altruism community and was already there when OpenAI was born at the Rosewood Hotel. Dr. Amoudi, who keeps twirling his curls as he talks, is leading a lab building a neural network called a large language model that can learn from large amounts of digital text. It generates text on its own by analyzing countless Wikipedia articles, digital books, and message boards. It also has an unfortunate habit of making things up. It's called GPT-3 and was released in the summer of 2020.

Researchers within OpenAI, Google, and other companies believe that this rapidly improving technology may be the path to AGI.

However, Dr. Amodei is dissatisfied with the Microsoft deal because he believes it is taking OpenAI in a very commercial direction. He and other researchers tried to pressure the board to oust Mr. Altman, according to five people with knowledge of the matter. But they failed and left. Like the DeepMind founders before them, they worried that their new company bosses would put commercial interests ahead of safety.

▲ Daniela Amodei and Dario Amodei co-founded Anthropic

In 2021, about 15 engineers and scientists created a new lab called Anthropic. Their plan is to build artificial intelligence the way effective altruists think it should be built—with very tight controls.

"Anthropic's co-founders are not trying to remove Sam Altman from OpenAI," Anthropic spokesperson Sally Aldous said. "The co-founders themselves came to the conclusion that they wished to leave OpenAI to start their own company, they made this clear to OpenAI's leadership and within weeks negotiated their exit on mutually acceptable terms."

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

reveal

When OpenAI secured an additional $2 billion in investment from Microsoft, Altman and another executive, Greg Brockman, went to Bill Gates' mansion on the shores of Lake Washington outside Seattle. Although the Microsoft founder is no longer involved in the company's day-to-day operations, he remains in regular contact with the company's executives.

Over dinner, Gates told them he wasn't sure large language models would work. He said he would remain skeptical unless the technology could handle tasks that require critical thinking — such as passing an Advanced Placement biology exam.

Five months later, on August 24, 2022, Altman and Brockman returned and brought with them an OpenAI researcher named Chelsea Voss. Ms. Voss was an International Biology Olympiad winner in high school. Mr. Nadella and other Microsoft executives were also present.

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

Mr. Brockman ran the system through multiple-choice questions on an advanced biology exam, and Ms. Voss scored the answers. The first question involves polar molecules, which are groups of atoms with a positive charge on one end and a negative charge on the other. The system answered the question correctly and explained its choices. "We only trained it to provide answers," Mr. Brockman said. "The conversational nature magically emerged." In other words, it was doing something they didn't really design it to do.

Out of 60 questions, GPT-4 only got one wrong answer.

Mr. Gates sat up straight, his eyes wide. He had a similar reaction in 1980 when researchers showed him the graphical user interface that became the basis of modern personal computers. He believes that GPT is so revolutionary.

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

OpenAI has defeated Anthropic’s Effective Altruist. Page's optimists at Google rushed to release their own chatbot, Bard, but it was widely seen as losing the race to OpenAI. Three months after ChatGPT was released, Google's stock fell 11%. Mr. Musk was nowhere to be found.

But all this is just the beginning.

# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo