Claude 3 deciphers OpenAI s email ciphertext Musk angrily denounced that it should be renamed Closed

Mondo Technology Updated on 2024-03-08

Editor: Momoko I'm sleepy.

Following OpenAI's head-on declaration of war with Musk and the exposure of 8 years of emails, Musk himself finally responded.

He said, "If OpenAI changes its name to Closed AI, I will give up the prosecution."

Subsequently, he made all kinds of memes, and modified Altman's classic return to work card to Closed AI.

Uncover the essence of OpenAI - Closed AI.

What's more interesting is that netizens launched an attack on the content of the mask in the email, trying to decipher the hidden text.

Not only did I use the world's most powerful model, Claude 3, but I also used participle guessing.

Claude 3 Solving the Case: The Future of Humanity Is in the Hands of Google

In the email, Musk said that unfortunately, the future of humanity is now in the hands of .

So, whose hands are it exactly?

Netizen Pietro Schirano said that he used Claude 3 to decrypt part of the content of the obscured content in OpenAI's email.

Surprisingly, they use a per-word desorption, which means that the length of each redaction is proportional to the length of the word.

So, based on the context and word length, Claude made the following guesses about the original content:

First of all, the future of humanity that Musk said is now in Google's hands.

And then immediately after that, I believe that DeepMind alone spends more than $100 million a year.

This sentence can be seen at a glance and is wrong, after all, the sentence length of the original text ends before the word hands.

In the last big expunged section, Claude 3 speculates: they are building huge clusters and have a clear path to artificial general intelligence. It adapts to any problem and crushes humans on every intellectual task, while also running on inexpensive hardware.

In any case, with Claude 3's ciphertext interpretation, the content of the letter does not contradict it from the beginning.

Let's take a look at another email, a large part of the blacked out section was sent to Musk by someone from the OpenAI team.

It is too difficult to decipher this email, and the first step for netizen Varepsilon to start is word segmentation.

He guessed that it was supposed to be a letter written by Karpathy to Musk and copied to himself.

After some effort, the first sentence to be deciphered was: deepmind's alphazero implementation is over nine to twelve times the next

However, more progress is still ongoing.

Even OpenAI's generation researcher Will Depue took an interest in it and said that it looks interesting and I want to join it, but it probably won't work.

5 details revealed in the email

Chief scientist Ilya Sutskever's words in the email have become the focus of many netizens.

As we get closer to building AI, it makes sense to reduce open.

He argues that OPEN means that after the AI is built, its results should benefit everyone, but it doesn't mean that all the scientific research has to be shared.

Although in the short to medium term, sharing results for recruitment purposes is undoubtedly the right strategy.

In response to Ilya's reply, Musk used only one word to express his agreement that AI technology does not have to be open source.

In an email to Sutskever and Brockman, Musk said:

OpenAI is burning cash, and the current funding model is difficult to compete with big companies like Google. The most promising option is to merge OpenAI with Tesla and fund it through Tesla**.

Musk added that we may hope otherwise. But in my opinion, Tesla is the only company that can compete with Google. Even so, it is unlikely that it will become a counterweight to Google. But this probability is not zero.

While OpenAI is still ostensibly a non-profit organization.

But back in 2018, OpenAI and Musk discussed a for-profit shift — one to the current unusual and complex nonprofit-limited for-profit structure.

Musk noted that a profit-focused strategy is likely to generate more sustainable revenue over time** and, in the case of the current team, is likely to attract significant investment. But building a product from scratch can be overly complex, distracting from the focus of research.

He went on to say that the most promising approach is for OpenAI to use Tesla as its cash cow to help build self-driving solutions.

If it can be done well, and the transportation industry is large enough, then Tesla's market capitalization can be increased to O(100k) and this revenue can be used to fund AI jobs at the right scale. 」

By hard take-off, Ilya actually means – the sudden and rapid development of artificial intelligence to a stage beyond human intelligence, with no buffer period.

He emphasized that in this scenario, it is more difficult to develop secure AI than to develop insecure AI. If everything is open-sourced, it will be easy for bad guys with a lot of computing power to build insecure AI.

Musk and Altman's love-hate relationship

Between Musk, OpenAI, and Altman, there are still some secrets that have never been made public.

From falling in love to killing each other, what exactly happened between them?

Today, WSJ published a long article about how Musk's brotherhood with Altman came to an end.

More than a decade ago, at SpaceX's headquarters in Southern California, Sam Altman first felt Musk's ambitions, and that meeting changed the course of his life forever.

Today, as the CEO of OpenAI, Altman often recalls the huge contrast in his heart at that time.

The 20-year-old himself has just sold his first entrepreneurial project with an unsatisfactory **, and Musk, who is 14 years older than him, is planning a grand plan to launch a rocket to Mars.

Altman later recalled, I thought to myself, this is what conviction looks like.

Recently, Altman has experienced first-hand the former mentor's insistence on his beliefs: Musk directly challenged Altman himself.

In a lawsuit filed last week, Musk accused Altman of deviating from OpenAI's original mission in pursuit of profits, which OpenAI denied.

This intensifying love-hate relationship reveals the changing balance of power in the AI arms race.

In 2015, the two co-founded OpenAI. As a non-profit research organization, the initial tens of millions of dollars ($40 million, according to emails) were funded by Musk himself.

Now, Musk has left, and OpenAI has turned to commercialization, receiving billions of dollars in investment from Microsoft. OpenAI is leading the current AI boom, while Musk's own startup, XAI, can barely catch up with the organization that was once funded by him.

On a personal note, this is sad, Altman said in a memo sent to employees on the day of the lawsuit. He called Musk his personal hero.

I'd like to see Musk as a builder, someone who competes by building better technology, and I hope he'll be on our side, he wrote.

In a blog post posted yesterday by OpenAI, it is planned to dismiss all legal charges filed by Musk.

Over the years, Altman has relied on Musk's fame and wealth to help OpenAI get off the ground. They teamed up to try to prevent Google, which was far ahead of the AI field at the time, from becoming the industry hegemon.

However, around the reasons for the deterioration of their relationship, people around them have given different accounts:

Those who stand with Altman believe that Musk is jealous of his achievements in the field of AI. Compared with AI security, Musk is more concerned about how to surpass OpenAI.

Musk's supporters, on the other hand, argue that his concerns about AI security are real, and that he sees XAI as the key to developing a better alternative to OpenAI.

Altman and Musk were introduced to each other by Geoff Ralston, a partner at Y Combinator.

When the two first met, Altman had just sold Loopt for an undesirable **, which only paid off.

In his 20s, he was at a crossroads in his life, looking for peace of mind in a ashram in India and considering whether to start a new company or go all in investing.

In contrast, Musk had already successfully developed a commercial spacecraft capable of cargo from the space station and back to Earth, which he saw as a critical first step towards human interstellar life.

geoff ralston

As AI develops faster and faster, Musk is also becoming more and more uneasy.

According to Musk's lawsuit, his concerns originated in a 2012 meeting with Demis Hassabis, co-founder of AI company Deepmind, where Hassabis highlighted the potential risks that AI advances could pose to society.

What is not mentioned in the lawsuit is that Musk invested in DeepMind after the meeting in order to pay closer attention to the development of this technology.

Later, when Musk learned that Google was planning to buy Deepmind, he was shocked.

He made a bid with Luke Nosek, one of PayPal's co-founders, but it was unsuccessful — Google eventually bought Deepmind.

Altman has had a keen interest in AI since he was a child, and at the age of 18 he was already at the top of his list of questions he wanted to explore in depth.

As Altman's popularity in Silicon Valley continues to rise, he strives to put the spotlight of the tech world on the immense potential of AI.

In 2014, he said on his personal blog that AI could be the most significant technological development ever. A few days later, he was appointed head of Y Combinator, who later became a supporter of OpenAI.

Like Musk, Altman is deeply concerned about the potential dangers of this technology. In February 2015, he wrote that AI may be the greatest threat to the continued existence of humanity.

In March of that year, Altman asked Musk if he would be interested in co-authoring an open letter calling on the United States to pay attention to the issue of artificial intelligence.

By May, he had sent an email to Musk suggesting that Y Combinator start an artificial intelligence project similar to the Manhattan Project. Musk responded by saying that the idea might be worth further discussion .

Subsequently, the two began to work together to build a new AI lab - Musk named it OpenAI.

In a June email, Altman suggested that they should run the nonprofit as part of a five-member board of directors, and suggested that an open letter calling for AI regulation should be issued after the lab was officially established. Musk fully agrees with this.

Altman invited Greg Brockman, Stripe's CTO, to join.

Musk, on the other hand, helped recruit Ilya Sutskever, a top AI scientist who works at Google.

To make the nonprofit's compensation package more attractive, OpenAI plans to offer employees hired shares in Tesla and SpaceX, as well as the opportunity to participate in Y Combinator's investment projects.

Musk and Altman became the first co-chairs of OpenAI. According to former employees, it is actually Musk who has more influence and control in this organization.

Initially, Musk often showed up in the office to come up with bold ideas and ask employees what they thought about implementing artificial general intelligence (AGI).

Interestingly, at the time, OpenAI and another of Musk's start-up projects, brain-computer interface company Neuralink, shared office space.

At the same time, Musk is also the financial pillar of OpenAI.

The email leaked by OpenAI said that no, Musk had promised $1 billion in funding to avoid sounding hopeless compared to Google or Facebook spending, and promised to make up for any shortfalls.

At the time, Altman and Brockman's initial goal was $100 million.

According to the complaint, Musk donated a total of $44 million, including $15 million in 2016 and another $20 million in 2017, becoming the largest donor for two consecutive years. In addition, he also took on the rent of OpenAI for several years.

Musk urged OpenAI's research team to work on innovative projects to get a head start on the competition with DeepMind. One of the teams aims to defeat the world's top players in Dota 2.

But by 2017, OpenAI had not achieved any major research results.

Former employees revealed that Musk appeared increasingly impatient, and he increased the pressure on employees, sometimes even threatening to quit the project.

In the same year, Google published a study on a new type of AI model, Transformer**, which marked a new era in building large language models like tools like human chatbots.

*Noting that achieving this goal requires processing large amounts of data and enormous computing power.

In order to raise enough computing resources, Brockman and others proposed to transform OpenAI's architecture into a for-profit nature.

In this way, it will be able to obtain funding from investors, including Microsoft.

Musk objected to this, making it clear in his letters to Brockman, Sutskever, and Altman that they would either start their own business or continue to maintain OpenAI's non-profit nature. He stressed that he will not fund OpenAI until they have made a clear commitment, and the discussion ends here. 」

In response, Altman said that he is still very bullish on the non-profit model

In a blog post published on Tuesday, OpenAI said that Musk has recognized the need to establish a for-profit entity, but that he wants not only a majority stake and control of the initial board of directors, but also a CEO.

On the one hand, Musk has been trying to tighten his grip on OpenAI, and has even proposed to include OpenAI in Tesla.

In an email attached to the blog post, Musk said: Tesla is the only possible way to compete with Google. Still, the odds of becoming a Google rival are slim, but not impossible. 」

On the other hand, Musk also tried to attract OpenAI's researchers to join Tesla, an act that angered his colleagues.

The article notes that during the discussions, he interrupted financial support. At this point, OpenAI had to rely on the financial backing of LinkedIn co-founder Reid Hoffman to stay afloat.

In February 2018, OpenAI's top management rejected a proposal that Musk intended to control, so he chose to resign as co-chairman. Thereafter, Altman took over as CEO.

At an all-hands meeting, Altman thanked Musk for his contributions within the organization, and Musk also said at the meeting that he plans to conduct his own AI research at Tesla.

When a young researcher questioned Musk's decision, arguing that it could exacerbate the AI arms race, Musk called the researcher a fool and left in anger.

At the end of 2018, Musk again expressed his pessimism about the future of OpenAI to Altman, Brockman, and Sutskever via email.

He believes that without major adjustments in execution and resource allocation, the likelihood of OpenAI and Deepmind Google remaining competitive is zero.

Although Musk stopped making cash donations to OpenAI, he continued to cover its rental costs.

In March 2019, after OpenAI set up a profitable subsidiary, Altman still publicly supported Musk.

When investors started shorting Tesla**, Altman immediately stood up to defend Musk.

In November 2022, when OpenAI released ChatGPT, a chatbot that can write poems and generate computers, there was a rift in the relationship between the two.

Musk has publicly expressed concern about ChatGPT's race to speed up the development of powerful AI, while he has also begun to question how the nonprofit, which he co-founded, raised billions of dollars from Microsoft. These challenges eventually became the basis for his lawsuit.

Shortly after OpenAI released ChatGPT, Musk announced that he had cut off OpenAI's access to Twitter's data, as OpenAI had previously been considering using its data to train models.

After that, Altman invited Musk to visit OpenAI's headquarters and had a long private meeting about Twitter's decision-making and ChatGPT issues.

During that time, Musk revealed to Altman his intention to create a new general artificial intelligence company, XAI. Altman was puzzled, unsure if joining another AI race would ease Musk's concerns.

A few months after the release of ChatGPT, Musk began working to poach employees from OpenAI to his XAI, and even threatened to file a lawsuit against Altman and OpenAI.

By November, Musk had released his own chatbot, Grok, a less woke-up competitor than ChatGPT.

Recently, XAI began to prepare for a new round of financing, which is likely to make the competition between it and OpenAI more intense.

In the wake of Musk's lawsuit, Altman sent a memo to his team: The idea that benefiting humanity and building a business are contradictory and incomprehensible.

The Year of Science and Technology in China

Related Pages