Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

OpenAI engineers revealed that it took only 8 days to develop ChatGPT, and a long article revealed how top streams in Silicon Valley, such as Google DeepMind, were born.

2024-11-04 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

Shulou(Shulou.com)12/24 Report--

The emergence of OpenAI stems from a dispute between Musk and Larry Page. That was in 2015, when Tesla was just making a profit and Page was Google's CEO. It was a few weeks after the quarrel that OpenAI was born at a dinner.

Recently, OpenAI engineers astonishingly exposed that the development of ChatGPT took only 8 days!

The engineer said that a year ago today, he signed up for a preview of the study demonstrated around the world.

Within 8 days, the team completed the whole process of product production and launch. At that time, no one could predict how the world would change.

Here are some screenshots of the Crazy first Night.

As we all know, it takes many years to go from GPT-2 to GPT-3, but why is it only 8 days since ChatGPT was born?

It is reported that Anthropic, which was set up by some "defected employees" of OpenAI at that time, is about to release a large model product. In order to get ahead of them in releasing the AI chatbot, the OpenAI team wrote a web page and tuned an interface in Next.js.

Then, ChatGPT, which set off an AI storm all over the world, was born.

How AI companies and Unicorns were born today, the New York Times also published a long article revealing a dramatic story about the origins of Google, DeepMind, OpenAI and Anthropic.

Self, fear and money: how AI ignites

Musk celebrated his 44th birthday in July 2015, and his wife held a three-day party for him.

That year, Tesla had just made a profit, and Larry Page was still Google's CEO. Page, as more than 10 friends of Musk, was also invited to the party.

That night, when the two men talked about whether artificial intelligence would benefit or destroy mankind, the discussion turned into a quarrel.

Among the more than 30 friends present, they all began to calm down and listen to their argument.

Because of the abnormal diseases of vocal cords, Page has a vision for digital utopia.

He believes that humans will eventually merge with artificial intelligence machines. And one day, all kinds of intelligence will compete for resources, and the best intelligence will win.

'If that happens, machines will destroy humans and we are doomed to fail, 'Mr. Musk said.

Page believes that his ideal utopia is worth pursuing. Finally, he called Musk a "breed", believing that he preferred human beings to future digital life forms.

Musk later said the insult was the last straw.

Many of the people present were dumbfounded, but eight years later, they found that the argument between the two men was going on everywhere.

Will AI benefit or destroy mankind? This issue has sparked a protracted war among Silicon Valley founders, users, academics, legislators and regulators.

This debate has pitted some of the richest people in the world against each other-Musk, Page, Xiaoza, Peter Thiel, Nadella, Sam Altman.

Everyone is fighting for a cause that could one day be worth trillions of dollars.

At the heart of this game is an inexplicable paradox.

Those who say they are most worried about AI are also those who have the strongest determination to create AI and enjoy its wealth. They use firm belief to prove that their ambitions are reasonable, and only they can stop AI from destroying mankind.

That summer, Musk and Page stopped talking to each other after the party.

A few weeks after the birth of OpenAI, Musk had dinner with several researchers at Altman, which then ran a technology incubator.

It was this dinner that led to the birth of OpenAI.

Backed by hundreds of millions of dollars from Musk and other investors, OpenAI promised to protect the world from Page's vision.

Now, OpenAI has fundamentally changed the technology industry with ChatGPT, valuing OpenAI at more than $80 billion, according to insiders.

Now, however, Musk and Altman have parted ways.

Last month, OpenAI's board of directors also had infighting. The "traitor" on the board tried to get rid of Altman because they no longer believed that he would make AI for the benefit of mankind.

The drama within OpenAI also allows the world to see for the first time that there is a fierce struggle among those who decide the future of AI.

In fact, a few years before the OpenAI upheaval, there was a little-known fierce struggle in Silicon Valley for the initiative of this technology.

The New York Times interviewed more than 80 executives and scientists, including two who attended Musk's birthday party in 2015, to recreate the story of ambition, fear and money.

DeepMind was born five years before Musk's birthday party when Demis Hassabis, a 34-year-old neuroscientist, walked into a cocktail party at Peter Thiel's San Francisco townhouse and realized he had made money.

In Thiel's living room, there is a chess board, and Hassabis was once the second-largest chess player in the world under the age of 14.

"I prepared for that meeting for a year. I know this will be my unique charm: he likes chess. "

In 2010, Hassabis and colleagues were looking for funding to build AGI.

At that time, few people were interested in AI.

At the time, all three young people from the UK were associated with Eliezer Yudkowsky, a self-taught AI researcher, a community rationalist led by Yudkowsky who later developed into an effective altruist.

They believe that AI can cure cancer and solve climate change, but they worry that AI robots may surprise their creators.

Thiel became very rich through his early investment in Facebook and his early partnership with Musk in PayPal. He developed a strong interest in singularity.

With funding from Thiel, Yudkowsky expanded his artificial intelligence lab and held an annual conference on singularities.

A few years ago, one of Hassabis's colleagues met Yudkowsky and he arranged for them to speak at the meeting to make sure they were invited to Thiel's party.

Subsequently, Yudkowsky successfully introduced Hassabis to Thiel.

Hassabis told Thiel that there was a tension between bishops and knights and that the best chess players would understand that their values seemed to communicate first, but their advantages were very different.

Hassaibs succeeded. Thiel was so fascinated that after finishing his morning exercise, he invited them to the kitchen to party with himself.

Soon, Thiel and his venture capital firm agreed to invest $2.25 million in their startup, the first big investment they received.

They named the company DeepMind, a tribute to deep learning, neuroscience and the Deep Thought supercomputer in the Galaxy Odyssey Guide.

By the fall of 2010, they began to build the machines of their dreams. They believe they understand the risks, so they can protect the world.

"I don't think our positions are contradictory," said Mustafa Suleyman, one of the three founders of DeepMind. "these technologies have brought huge benefits. Our goal is not to eliminate them or suspend their development, but to mitigate disadvantages. "

DeepMind founders: Mustafa Suleyman (left), Demis Hassabis (center) and Shane Legg (right)

After receiving Thiel's investment, Hassabis entered Musk's social circle.

About two years later, they met at a meeting of Thiel's investment fund organization. Hassabis visited SpaceX headquarters and chatted with Musk in the cafeteria.

Musk explained that his plan was to colonize Mars to avoid dangers such as overpopulation on Earth. Hassabis replied that the plan would work-as long as super intelligence would not follow humans to Mars or destroy them.

Musk was speechless. He didn't think of the danger.

Soon, Musk invested in DeepMind with Thiel so he could get closer to the technology.

DeepMind, with a lot of money in hand, hired researchers who specialize in neural networks.

Neural network is a complex algorithm based on human brain images. It is essentially a huge mathematical system that takes days, weeks or even months to identify patterns in a large number of numbers.

These systems were first developed in the 1950s and can learn to handle tasks on their own. For example, after analyzing the scribbled names and addresses on hundreds of envelopes, the system can read handwritten text.

DeepMind takes this concept a step further. It has set up a system where you can learn to play classic Atari games such as Space invaders, Ping-Pong and Breakthrough.

This attracted the attention of Google, another Silicon Valley giant, especially Larry Page.

After watching the demo of Deep Mind's machine playing Atari, he also wants to invest.

In the fall of 2012, Jay Geoffrey Hinton, a 64-year-old professor at the University of Toronto, and two graduate students published a paper in which they trained a neural network to identify common objects, such as flowers, dogs and cars.

People are very surprised by this technology, especially attracted the attention of AI researcher Yu Kai.

Baidu provided $12 million to Hinton and his academics and invited them to join Baidu.

Hinton didn't choose Baidu, but the money caught his attention.

As a Cambridge-educated British expatriate, Hinton spends most of his time in academia, occasionally at Microsoft and Google.

He doesn't care so much about money, but because of a child with a neurological mutation, the money means economic stability.

After consulting lawyers and experts, he proposed the plan to organize an auction.

Google, Microsoft, Baidu and other big technology companies believe that neural networks are the way to AI.

Page saw similar technology in Google's brain lab, and he believes Hinton's research will be of great benefit to Google's brain.

He gave Alan Eustace, Google's senior vice president of engineering, a blank check to hire any AI talent Google needed.

The night before the auction, Eustace and Jeff Dean, who led Google's brain, flew to Lake Tahoe, where Hinton and academia were located, had dinner with them and persuaded them to come to work at Google.

The next day, Hinton hosted the auction in his hotel room.

He seldom sits down because of an old back injury. He put a trash can upside down on the table, put his laptop on it, and bids rolled in over the next two days.

Google and Microsoft made a bid, and as the price continued to rise, DeepMind was launched quickly.

The bid was $20 million, then $25 million. Microsoft pulled out when the price exceeded $30 million, and then rejoined the bid at a price of $37 million.

'it 's like watching a movie, 'says Hinton.

Then Microsoft quit for the second time.

Only Baidu and Google were left with a bid of $42 million and $43 million. Finally, at the price of $44 million, Hinton and his students stopped the auction.

Bids are still climbing, but they want to work for Google. And the amount of money is already amazing.

The auction shows that large companies with deep pockets are determined to buy the most talented AI researchers.

Previously, Hassabis has been telling employees that DeepMind will always be an independent company, and that this is the only way to ensure that AI does not endanger humans.

But as big technology companies began to join the talent competition, he knew he had no choice: it was time to sell DeepMind.

Google and Facebook are scrambling to acquire DeepMind London Labs by the end of 2012, according to three people familiar with the matter.

Hassabis and its co-founders insist on two conditions: DeepMind's technology cannot be used for military purposes, and its AGI technology must be overseen by an independent committee of technical experts and ethicists.

Google offered $650 million. Facebook Zuckerberg offered more, but did not agree to these terms. So DeepMind was sold to Google.

Zuckerberg is determined to build his own artificial intelligence laboratory. He hired French computer scientist Yann LeCun.

A year after the Hinton auction, Zuckerberg and LeCun flew to Lake Tahoe to attend the same artificial intelligence conference.

Zuckerberg wandered around a casino suite, talking to top AI researchers, who soon received multimillion-dollar salaries and shares.

AI, once ridiculed, is asking Silicon Valley's richest people to spend billions of dollars to prevent themselves from falling behind.

The lost Ethics Committee when Musk invested in DeepMind, he broke his old rule of not investing in any company he didn't run.

After a month of quarrelling with Page, he found that the drawbacks of the decision were obvious.

On August 14, 2015, the DeepMind Ethics Committee held its first meeting. The members met in a conference room outside Musk's SpaceX office, but Musk's control ended there.

When Google bought DeepMind, it bought the whole company. Musk is out.

Three Google executives who firmly control DeepMind attended the meeting: page; Sergey Brin, Google co-founder and Tesla investor; and Eric Schmidt, Google chairman. Other participants included Reed Hoff, another founder of PayPal, and Toby Ord, an Australian philosopher who studies "risks".

The founders of DeepMind reported that they were moving forward with their work, but also realized that there were serious risks to the technology.

Suleyman, co-founder of DeepMind, gave a speech entitled "here comes the pitchfork (The Pitchforkers Are Coming)".

He told the board that AI could lead to an explosion of false information. He worries that the technology could replace countless jobs in the next few years, and the public will accuse Google of losing their jobs.

He believes Google needs to share wealth with millions of people who cannot be out of work and guarantee a "universal basic income".

Musk agreed. But it is clear that Google executives do not intend to redistribute their wealth.

Schmidt says he thinks these concerns are completely exaggerated.

Page also agreed. He believes that artificial intelligence will create more jobs than it will take away.

Eight months later, DeepMind made a breakthrough that shocked the AI community and the world.

AlphaGo beat the best human chess player in the world in the go game. More than 200 million people around the world watched the game.

Most researchers believe that it will take another 10 years for AI to do this.

Rationalists, effective altruists and others worried about the risks of AI claim that AI's victory confirms their fears.

Victoria Krakovna, who is about to join DeepMind as the AI Security Institute, wrote, "this is a sign that AI is growing so fast that it is even faster than many experts expected. "

The founders of DeepMind are increasingly worried about what Google will do with its own invention.

In 2017, they tried to break away from the company. Google responded by raising salaries and stock awards for DeepMind founders and their employees. So they stayed.

The ethics committee has never held a second meeting.

The breakup Musk was convinced that Page was too optimistic about AI and angry about losing DeepMind, so he set up his own laboratory.

Musk injected money into OpenAI, which was founded at the end of 2015, as did his former Paypal friends Hoffman and Thiel.

Several people invested 1 billion dollars in the project and poached Ilya Sutskever from Google in order to get OpenAI started. (Ilya is one of the graduate students that Google bought at Hinton auction.)

At first, Musk wanted to run OpenAI as a non-profit organization, unaffected by economic incentives.

But when Google amazed the tech world with the gimmick of go, Musk changed his mind. He is eager for OpenAI to invent the world's imagination and narrow the gap with Google.

However, OpenAI does not work as a non-profit organization.

At the end of the 17th, he laid out a plan to wrest control of OpenAI from Altman and other founders and turn it into a commercial company in partnership with Tesla, relying on supercomputing developed by the latter, according to people familiar with the matter.

When Altman and others raised objections, Musk resigned.

When he explained that OpenAI needed to speed up action, one researcher retorted that he was reckless. Musk called him a jerk and rushed out with his rich purse.

As a result, OpenAI needs new financing. When Altman was flying to Sun Valley for a conference, he met Microsoft CEO Nadella. The new cooperation makes sense.

Altman knows Microsoft's CTO Kevin Scott. Microsoft bought LinkedIn from Hoffman, a member of the OpenAI board. Nadella told Scott to finish it. The transaction was completed in 2019.

Altman and OpenAI formed a for-profit company under the original non-profit organization. They have $1 billion in new capital, and Microsoft can build AI into its vast cloud computing service.

Not everyone in OpenAI is happy.

Dario Amodei is a researcher associated with an effective altruistic community. His lab released GPT-3 in the summer of 2020.

Researchers at OpenAI, Google and other companies believe the technology could be a path to AI. But Dr. Amodei is not satisfied with the Microsoft deal, which he believes is taking OpenAI into a real business direction.

He and other researchers went to the board to try to oust Altman, according to five people familiar with the matter. After the failure, they left, just like the founders of DeepMind.

In 2021, about 15 engineers and scientists created a new laboratory called Anthropic. Their plan is to build AI-- in a way that effective altruists believe should be done through very strict control.

"the co-founder of Anthropic did not try to remove Sam Altman from OpenAI," said Sally Aldous, a spokesman for Anthropic. They wanted to leave OpenAI to start their own company, so they revealed this to the leadership of OpenAI and withdrew from negotiations on mutually agreed terms within a few weeks. "

Two years later, Anthropic accepted an investment of $4 billion from Amazon and another $2 billion from Google.

After revealing that GPT-4 received a $2 billion investment from Microsoft in OpenAI, Altman and another executive, Greg Brockman, visited Bill Gates' mansion in Seattle. At that time, the Microsoft founder was no longer involved in the daily work of the company, but kept in touch with senior executives on a regular basis.

At dinner, Gates told them that he doubted whether the large language model would work.

Unless LLM can perform tasks that require critical thinking, such as passing the AP Bio's college biology exam, he says.

Five months later, on August 24, 2022, Altman and Brockman came to Gates again and brought another OpenAI researcher, Chelsea Voss. She won a medal in the International Biology Olympiad in high school.

Also present at the meeting were Nadella and other Microsoft executives.

At the scene, OpenAI staff demonstrated GPT-4 technology.

Brockman asked GPT-4 to do multiple-choice questions in the Advanced Biology exam, and Voss graded the answers.

The first question concerns polar molecules, and GPT-4 gives the right answer and explains why.

Brockman explained that they were only training GPT-4 to provide answers and were not designed for it to explain, but it did so automatically.

Of the 60 questions, GPT-4 made only one mistake.

Gates sat up in his chair and opened his eyes wide.

The last time he had such a reaction was in 1980, when researchers were showing him the graphical user interface that would become the basis of modern personal computers.

There is no doubt that he believes GPT is revolutionary.

By October, Microsoft had integrated GPT-4 into services such as Bing search engine. Today, 100 million people use ChatGPT every week.

OpenAI defeated Anthropic's effective altruist. Google's optimists rushed to release Bard, but it is accepted that Google has lost to OpenAI.

Three months after the release of ChatGPT, Google shares fell 11%. And this is just the beginning.

Reference:

Https://www.nytimes.com/2023/12/03/technology/ai-openai-musk-page-altman.html?smid=nytcore-android-share

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report