Article reprint source: AI DreamWorks

Source: SenseAI

Image source: Generated by Unbounded AI

"Now I am Death, the destroyer of worlds" - this sentence appears twice in Nolan's film "Oppenheimer". It is not only the most famous sentence in "Bhagavad Gita", but also Oppenheimer's thoughts on his nuclear bomb test. As the founder of OpenAI, Sam Altman, his birthday is the same day as Oppenheimer, and he has always compared himself to the bomb maker. At the same time, he is also a mindfulness practitioner and has known Jack Kornfield, Jobs's meditation teacher, for many years. This article details Altman's growth and development process, his thoughts on mindfulness and technology, social change, and the future of OpenAI. Sam Altman said he can see our future. Is he the Oppenheimer of our time? "

Sense

We try to propose more divergent deductions and reflections based on the content of the article, and welcome exchanges.

• A conscious digital person is a narrative subject that often appears in science fiction. Recently, scholars from MIT published an article proving that LLM is not just a superficial probability model, but can understand time and space and form an understanding of the world. Regardless of how we define "understanding of the world", this technological evolution led by OpenAI is a social revolution that affects society's politics, civil rights, economic inequality, and education.

• Altman's Moore's Law of Everything states that in this world, for decades, the price of everything from housing, education, food, clothing, etc. has fallen by half every two years. The socioeconomic changes driven by technology are accelerating, and public policies need to be adjusted accordingly. For the regulation of artificial intelligence technology, due to the ambiguity of generative AI technology, it may be more appropriate to regulate it within its specific scope of application rather than establishing a regulatory agency modeled on the International Atomic Energy Agency. Of course, the creators of the technology determine the direction of its development.

This article has 9886 words in total. It takes about 15 minutes to read carefully.

01. At the Mindfulness Summit

From left to right: Soren Gord Hamer (founder of Wisdom 2.0); Jack Kornfield; Open AI CEO Sam Altman

This spring, Sam Altman, the 38-year-old chief executive of OpenAI, sat down with Jack Kornfield, Silicon Valley’s most beloved Buddhist monk. It was Wisdom 2.0, an event dedicated to combining wisdom with “the great technologies of our time,” at the Yerba Buena Center for the Arts in San Francisco. The two men sat in giant white upholstered chairs on a stage with dark mandala backs. Even the host seemed puzzled by Altman’s presence.

He asked, "Why are you here?"

"Yeah, well..." Altman said. "I'm definitely interested in the topic" - formally, mindfulness and artificial intelligence. "But meeting Jack was one of the great joys of my life. I'm happy to explore any topic with Jack."

The agenda became clear when Kornfeld, 78, whose books include “The Wise Heart,” which has sold more than 1 million copies, made his opening remarks.

“My experience is that Sam… I would use the language of, he’s a very servant leader.” Kornfield is here to bear witness to Altman’s exceptional character. He’ll answer a question that haunts many of us: How safe should we feel about Altman, since this relatively young man in charcoal Chelsea boots and a gray waffle sweater seems to be in control of how artificial intelligence will enter our world?

Kornfield said he had known Altman for several years. They meditated together. They explored the question: How could Altman "build values ​​-- the bodhisattva vow, to care for all beings"? How could compassion and care be "programmed in some way, in the most profound way"?

Throughout Kornfield's speech, Altman sat cross-legged, his hands clasped in his lap, his posture impressive, his facial expression firmly conveying patience (although his expression also made it clear that patience was not his natural state). "I'm going to embarrass you," Kornfield warned him. Then the monk once again addressed the crowd: "He has a pure heart."

Altman spent much of the rest of the panel talking about his views. He knows people are afraid of AI, and he thinks we should be, too. So he feels he has a moral responsibility to speak up and answer questions. “It would be very irrational not to do so,” he said. He believes we need to work together as a species to decide what AI should and shouldn’t do.

By Altman’s own assessment—as evidenced by his many blog posts, podcasts, and video events—we should be pleased with his status as an AI leader, but not great. As he understands it, he is a brilliant but not genius “tech bro” with the qualities of Icarus and a few distinctive traits. First, he has “absolutely delusional self-confidence.” Second, he has “a long view of the arc of technological and social change.” Third, as a Jew, he is both optimistic and expects the worst. Fourth, he is good at assessing risk because his brain is not fooled by what other people think.

The downside: He's not an emotional or demographic fit for the role. "There might be people who are more into the role," he admitted on the Lex Friedman podcast in March. "There might be people who are more charismatic." He knows he's "out of touch with most people's real lives." He occasionally makes surprising statements. For example, like many in the tech bubble, Altman uses the phrase "median human," as in "AGI (artificial intelligence), to me," "is the equivalent of a median human that you could hire as a co-worker."

In Yerba Buena, the moderator asked Altman: How does he plan to assign value to artificial intelligence?

Altman said one of his ideas is to "gather as many humans as possible" to reach a global consensus. You know: decide together: "These are the value systems that should be added, and these are the limits that the system must not do."

The whole place gradually quieted down.

"The other thing I'd like to do is have Jack write down ten pages that say, 'Here are the collective values ​​and here are how we make the system do this.' That would be great."

The whole place became even quieter.

Altman isn’t sure whether the revolution he’s leading will be seen in the long run of history as technological or social. It will be “bigger than the standard technological revolution,” he believes. Yet he also knows that having spent his adult life around tech founders, “it’s always annoying to say, ‘This time it’s different,’ or, ‘You know, my stuff is super cool.’ ” Revolution is inevitable, he’s convinced of it. At a minimum, AI will upend politics (deep fakes have become a major concern in the 2024 presidential election), labor (AI has been at the heart of the Hollywood writers’ strike), civil rights, surveillance, economic inequality, the military, and education. Altman’s power, and how he’ll use it, are all the problems we face right now.

Yet it was hard to discern who Altman really is; how much we should trust him; and how much he integrates the concerns of others, even as he stands on stage intent on calming them. Altman said he would slow the revolution as much as possible. Still, he told attendees, he believed everything would be OK. Or probably would be. We — that little royal word played a big role in his rhetoric — should “decide what we want, decide we’re going to execute it, and accept the fact that the future is going to be very different, and probably for the better.”

This statement was not well received either.

"There was a lot of nervous laughter," Altman said.

Then he waved his hands and shrugged. "I could fool you and say, 'Oh, we can totally stop it.' But I think it's..."

02. Global Values

Altman didn’t finish the thought, so I resumed the conversation at OpenAI’s offices on Bryant Street in San Francisco in late August. The street outside is the flea market of the new capitalism: self-driving cars, dogs sunning themselves next to sidewalk tents, bus stops where public transportation systems fail, shops serving $6 lattes. Inside OpenAI is the low-key, bland tech enterprise: grab a Pellegrino from the mini-freezer or put our logo on it.

In person, Altman is more charming, more serious, more calm, more silly than expected—more of him. He is likeable. His hair is grizzled. He wears the same waffle-knit sweater that has quickly become his trademark. I am the 10 billionth reporter he has hosted this summer. As we sit down in a soundproof booth, I apologize for asking him to do yet another interview.

He smiled and said, "Nice to meet you."

After my talk with Kornfeld, one person said to me, “You know, I came in very nervous that OpenAI was going to make decisions about the values ​​of AI, and you convinced me that you weren’t going to make those decisions, and I was like, ‘Great.’ And then they said, ‘No, now I’m even more nervous. You’re letting the world make these decisions, and I don’t want that.’”

Even Altman found himself on stage answering questions about global values ​​a bit unnatural. "If I hadn't been involved, I would have said, why should these people decide my fate?" he told Tad Friend of The New Yorker in 2016. Seven years later, and with plenty of media training, he has softened his tone. "I have a lot of sympathy for government projects like OpenAI."

The new nice guy image is hard to square with Altman’s will to power, one of the traits for which he is best known. “He’s the most ambitious and sane person I know in Silicon Valley, and I know 20,000 people in Silicon Valley,” one friend in his inner circle told me.

Still, Altman explains his rise with a nonchalant attitude. "I mean, I'm a Jewish kid from the Midwest with what was, at best, an awkward childhood. I was one of the few ... " he muttered to himself. "You know, the first few dozen or so most important technology projects. I couldn't imagine that happening to me."

03. From Childhood to Founding OpenAI

Altman grew up in the suburbs of St. Louis, the eldest of four siblings: three boys, Sam, Max, and Jack, each two years apart, and a girl, Annie, nine years younger than Sam. If you weren't raised in a middle-class Jewish Midwestern family -- and I speak from experience -- it's hard to imagine the underlying self-confidence such a family instills in a son. "One of the best things my parents did for me was to constantly (multiple times a day, I think) affirm their love and the belief that I could do anything," Jack Altman once said. "The confidence that resulted was fantasy, narcotic, weapons-grade. It was like an extra valve in your heart."

The story about Sam usually goes something like this: He was a whiz kid -- a "rising star in the world of technological wizardry," according to the St. Louis Post-Dispatch. He began repairing his family's VCR at age 3. In 1993, for Altman's eighth birthday, his parents -- dermatologist Connie Gibbsteen and real estate agent Jerry Altman -- bought him a Mac LC II. Altman described the gift as "the dividing line in my life: before computers and after computers."

The Altmans ate dinner together every night. At the dinner table, they would play games like "Square Roots": Someone would call out a big number and the boys would guess. Anne would hold up a calculator and check who was closest. They also played "20 Questions" to guess the nightly surprise dessert. The family also played ping-pong, pool, board games, video games, and charades, where everyone knew who won. Sam preferred those games. Jack recalled his brother's attitude: "I have to win, and I have the final say." The boys also played water polo. "He would disagree, but I would say I was better," Jack told me. "I mean, there was no doubt I was better."

Sam is gay and came out in high school. This surprised even his mother, who had always thought of Sam as "just one of those tech dudes who was half-and-half." As Altman said in a 2020 podcast, the private high school he attended "wasn't the kind of place where you could really come out and talk about being gay." When he was 17, the school invited a speaker for National Coming Out Day. A group of students objected, "mostly based on religious beliefs, but also some who thought gay people were bad people." Altman decided to speak to the student government. He barely slept the night before. His last words, he said in the podcast, were, "Either you tolerate open communities or you don't, and you don't get to choose."

Altman enrolled at Stanford in 2003, just as Silicon Valley was beginning to recover from the dot-com bubble. That same year, Reid Hoffman cofounded LinkedIn. In 2004, Mark Zuckerberg cofounded Facebook. At that moment, the eldest son of a suburban Jewish family didn’t become an investment banker or a doctor. He became an entrepreneur. During his sophomore year, Altman and his boyfriend, Nick Sivo, began working on an early geo-tracking program, Loopt, for locating your friends. At the time, Paul Graham and his wife, Jessica Livingston, and others had just created the Summer Founders Program as part of their venture capital firm, Y Combinator. Altman applied. He won a $6,000 investment and the chance to spend a few months with like-minded nerds in Cambridge, Massachusetts. Altman worked so hard that summer that he got scurvy.

At Loopt, though, he didn’t stand out. “Oh, another bright young guy!” says Hoffman, who served on OpenAI’s board until January and remembers his impression of the young Altman. It was enough to raise $5 million from Sequoia Capital. But Loopt never caught on with users. In 2012, Altman sold the company to Green Dot for $43.4 million. Even Altman doesn’t consider it a success.

Sam Altman's story of growing up: From left: With his siblings at age 14 or 15 in the 2000s. At age 23, promoting Loopt at Apple's 2008 WWDC. Photos by Anne Altman; CNET/YouTube.

“Failure always sucks, but it sucks really, really bad when you’re trying to prove something,” Altman told me. He left the company “very unhappy” but with $5 million, which he used, along with funding from Peter Thiel, to start his own venture capital fund, Hydrazine Capital. He also took a year off and read a ton of books, traveled, played video games, and “became a total tech guy,” he said. “I was like, I’m going to go live in the Ashura Temple for a while, and it changed my life. I’m sure I’m still anxious and stressed in a lot of ways, but the way I felt was that I felt very relaxed and happy and at peace.”

In 2014, Graham asked Altman to take over as president of Y Combinator, which by then had helped start Airbnb and Stripe. Graham had described Altman in 2009 as "one of the five most interesting startup founders of the last 30 years" and later described him as "Bill Gates when he started Microsoft ... a naturally imposing, confident guy."

During Altman's tenure as YC president, the incubator received about 40,000 applications from new startups each year. The incubator heard in-person pitches from 1,000 of those companies. A few hundred of them received YC funding: usually around $125,000, along with mentorship and networking (including weekly dinners and group office hours), in exchange for a 7% stake in the company. Running YC can be both the greatest job in Silicon Valley and the worst job. From the perspective of venture capitalists (some of whom, some say, spend a lot of their time and energy), running YC can be both the greatest job in Silicon Valley and one of the worst jobs.

For most of his tenure, Altman and his brothers lived in one of his two San Francisco houses, one in SoMa and the other in the Mission. He preached ambition, solitude, and scale. He believed in the value of hiring from a network of acquaintances. He believed in not caring too much about what other people thought. "The big secret is that you can bend the world to your will to a great extent—most people don't even try," he wrote on his blog. The most successful founders didn't set out to create companies. Their mission was to create something closer to a religion, and at some point, starting a company turned out to be the easiest way to do that. The bigger sin, he believed, was to back yourself into a corner with a small idea and not think big enough.

Altman had a pretty fascinating life. He became incredibly wealthy. He invested in boyish dream products, like developing a supersonic plane. He bought a house in Big Sur and filled it with guns and gold. He raced his McLaren.

He also embraces a utilitarian philosophy of technocracy known as effective altruism (EA). EA does almost anything to make a lot of money, on the theory that its followers know best how to spend it. This ideology values ​​the future over the present and imagines a rapturous singularity when humans and machines merge.

Deep in this framework, in 2015, Altman co-founded the nonprofit OpenAI with Elon Musk and four others—Ilya Sutskever, Greg Brockman, John Schulman, and Wojciech Zaremba. The 501(c)(3)’s mission: to create “a computer that thinks like a human in every way, and uses it for the best benefit of humanity.” The idea is to build good AI and dominate the field before the bad guys build bad AI. OpenAI pledges to open source its research in keeping with EA’s values. If someone—or someone they deem “values-aligned” and “safety-conscious”—is ready to achieve AGI before OpenAI, they will assist the project, not compete with it.

Altman has served as YC president for several years. He sends countless texts and emails to founders every day and tracks how quickly people respond because, as he writes on his blog, he believes response time is "the most significant difference between great founders and mediocre founders." One of the differences". In 2017, he considered running for governor of California. He once was at a dinner "complaining about politics and state government, and someone said, 'You should stop complaining and do something,'" he told me. "I was like, 'Okay.' He published a platform, the United Slate, outlining three core principles: technological prosperity, economic fairness, and personal freedom. A few weeks later, Altman dropped his candidacy.

In early 2018, Musk attempted to take control of OpenAI, claiming that the organization was falling behind Google. By February, Musk left, leaving Altman in charge.

A few months later, in late May, Altman's father suffered a heart attack while boating on Creve Coeur Lake outside St. Louis. He was 67. He died in a hospital shortly afterwards. Anne told reporters that at the funeral, Sam gave each of Altman's four children five minutes to speak. She used her time to rank the emotional expressiveness of her family members. She put Sam and her mother last.

04. Moore’s Law of Everything Behind GPT

Altman published an article in March 2021 titled "Moore's Law for Everything." The article begins: "My work at OpenAI reminds me every day that the scale of socioeconomic change is coming faster than most people think...If public policy does not adjust accordingly, most people will end up worse off than they are now."

Moore's Law for microchips states that the number of transistors on a chip doubles approximately every two years, while the price drops by half. Altman's "Moore's Law of Everything" assumes "a world in which the price of everything from housing to education to food to clothing drops by half every two years for decades."

As Altman wrote this, he had left YC to focus on OpenAI full-time. In the spring of 2019, the first thing the company did under his leadership was to launch a for-profit subsidiary. Building AI had proven to be very expensive; Altman needed money. By the summer, he had raised $1 billion from Microsoft. Some employees quit, uneasy about the mission’s move away from “the greatest good of humanity.” Surprisingly few were unhappy with the shift, however.

“What, Elon, who’s worth a hundred billion, is going to be like ‘Naughty Sam,’ ” said one of Altman’s friends inside. Altman refused to take a stake in the company, and OpenAI initially capped investors’ profits at 100 times. But many thought it was just a visual move. A billion times a hundred is a lot of money. “If Elizabeth Warren came and said, ‘Oh, you turned this into a profitable company, you evil tech guy,’ ” said Altman’s friend, “everyone in tech would say, ‘Go home.’ ”

Altman continued to race his cars (one of his favorites was a Lexus LFA, which was discontinued in 2013 and, HotCars reports, “costs at least $950,000”). In the early days of the pandemic, he donned an IDF gas mask. He bought a ranch in Napa. (Altman is a vegetarian, but his partner, Oliver Mulherin, a computer programmer from Melbourne, “loves cows,” Altman says). He bought a $27 million house on Russian Hill in San Francisco. His friends grew. In 2021, Diane von Furstenberg called him “one of my closest, very, very close friends. Meeting Sam was like meeting Einstein.”

Meanwhile, as OpenAI began selling rights to its GPT software to businesses, Altman hatched a series of side projects to prepare for AI to transform the world. He invested $375 million in Helion Energy, a speculative nuclear fusion company. If Helion can succeed, Altman hopes to control one of the cheapest energy sources in the world. He invested $180 million in Retro Biosciences. Its goal is to extend the human lifespan by ten years. Altman also conceived and raised $115 million for the Worldcoin project, which scans the irises of people around the world by having them gaze at an orb called an Orb. Each iris fingerprint is then linked to a crypto wallet, into which Worldcoin will deposit currency. This will solve two problems created by AI: distinguishing between humans and non-humans - which will be necessary once AI further blurs the line between humans and non-humans; and issuing currency - also two problems created by AI.

This isn’t the portfolio of someone as ambitious as Zuckerberg, who seems somewhat stuffy compared to Altman, content to “build a ruling city-state,” as the tech writer and podcaster Jathan Sadowski puts it. This is the portfolio of someone as ambitious as Musk, someone who takes an “imperialist approach.” “He really sees himself as the greatest man in the world, a Nietzschean superman,” Sadowski says. “He’s going to simultaneously create the things that destroy us and save us from them.”

Then, on November 30, 2022, OpenAI released ChatGPT. The software attracted 100 million users in two months, becoming the greatest product launch in the history of technology. Two weeks ago, Meta released Galactica, but the company took it down three days later because the robot could not distinguish between truth and falsehood. ChatGPT also lied and hallucinated. But Altman released it anyway and considered it a virtue. The world needs to get used to this. We need to make decisions together.

“Sometimes being unethical is what sets a successful CEO or product apart,” a former colleague who worked with Altman in OpenAI’s early years told me. “Facebook isn’t that interesting technically, so why is Zuck winning?” He could “scale faster and build products without getting into this mess.”

05. Oppenheimer Moment

In May 2023, Altman began a global tour of 22 countries and 25 cities. It was said that it started out as a meeting with ChatGPT users, but later turned into a kind of "socialite party". Altman, sometimes in a suit and sometimes in a gray sweater, showed himself to the diplomats as the inevitable new technological superpower. He met with British Prime Minister Rishi Sunak, French President Emmanuel Macron, Spanish Prime Minister Pedro Sánchez, German Chancellor Olaf Scholz, Indian Prime Minister Narendra Modi, South Korean President Yoon Suk-yeol and Israeli President Isaac Herzog. He stood for a photo with European Commission President Ursula von der Leyen. In the photo, she looks elegant and indifferent, while he looks like "Where's Waldo" - the mobile phone in the front trouser pocket is clearly visible, and a pair of green eyes are piercing with fatigue and cortisol.

Then, Ultraman returned home and seemed to unload not only his gear but also his psychological baggage. From late June to mid-August, he tweeted a lot. If you want to get a more realistic understanding of him, Twitter is the best place to go.

The base model is trained on large-scale datasets and can perform a wide range of tasks. Developers use the base model as the basis for powerful generative AI applications, such as ChatGPT.

Is the action tonight Barbie or Oppenheimer?

Altman released a poll and Barbie lost, 17% to 83%.

Okay, Oppenheimer.

The next morning, Altman returned and expressed his disappointment.

I hoped Oppenheimer's film would inspire a generation of kids to become physicists, but it really missed the mark.

Let’s make that movie!

(I think social networks have successfully done this for startup founders).

Careful readers will be puzzled by Altman's work. For years, Altman has compared himself to the bomb maker. He once told reporters that he and Oppenheimer share a birthday. He once relayed Oppenheimer's words to Cade Metz of the New York Times: "Technology comes because it is possible." So Altman is not surprised that Christopher Nolan did not create an inspiring work in his biopic. Oppenheimer spent the rest of his life feeling ashamed and regretful about his actions in making the atomic bomb. "Now I have become Death, the destroyer of worlds" - this is both the most famous sentence in the Bhagavad Gita and what Oppenheimer told NBC News about his thoughts during the Trinity test. (This sentence also appears twice in the movie).

Altman has been tying himself to Oppenheimer on his world tour, discussing (in nonspecific terms) the existential risks posed by AI and arguing (very specifically) for a regulator modeled on the International Atomic Energy Agency, which was approved by the United Nations in 1957, four years after the agency was founded. The agency’s mission—to work for international peace and prosperity—sounds like a great analogy to casual listeners. But it vexed experts.

One line of criticism is about its political cynicism. “You say, ‘I’ll regulate,’ and then you say, ‘This is a very complex and specialized topic, so we need a complex and specialized agency to do this,’ but you know full well that agency will never be established,” Sadowski said. “Or if it is established, hey, that’s OK because you built its DNA,” Sadowski said.

Another problem is ambiguity. Heidy Khlaaf, an engineer who specializes in evaluating and verifying safety protocols for drones and large nuclear power plants, explained to me that to reduce the risk of a technology, you need to define precisely what the technology can do, how it can help and hurt society. (Maybe someone will use AI to invent superbugs; maybe someone will use AI to launch nuclear bombs; maybe AI itself will turn against humanity—the resolution of each case is unclear). Moreover, Khlaaf argues that we don’t need a new agency. AI should be regulated like other technologies, within the scope of its use. AI built using copyrighted material should be regulated by copyright law. AI used in aviation should be regulated accordingly. In the end, if Altman is serious about strict safety protocols, then he will take what he considers to be lesser harms more seriously.

“If you can’t even make your systems not discriminate against black people”—a phenomenon known as algorithmic bias that affects everything from how job applicants are ranked to which faces are labeled most attractive—“how can you stop it from destroying humanity?” Clough asks. The harms in engineered systems are complex. “One tiny software bug can take down the New York power grid. Trained engineers know that.” Every one of these companies, every top player in AI, has the resources and the basic engineering knowledge to figure out how to reduce the harms in these systems. Choosing not to do so is a choice.”

On the same day that Altman published the Oppenheimer vs. Barbie poll, he also published this article:

Everything "creative" is a recombination of what happened in the past, plus the product of epsilon and the quality of the feedback loop, plus the number of iterations.

People think they should maximize ε, but the trick is to maximize the other two.

Throughout the summer and fall, OpenAI faced increasing pressure for allegedly training its models and making money on datasets rife with copyrighted works. Michael Chabon organized a class-action lawsuit after learning that his book had been used to teach ChatGPT without permission. The Federal Trade Commission launched an investigation into the company for alleged widespread violations of consumer protection laws. Now, Altman believes that creativity doesn’t actually exist. Whatever those striking writers or angry illustrators think of their personality or value, they are just remixing old ideas. Just like OpenAI’s products.

In terms of wording, mathematical terms lend a veneer of reliability. Mathiness is a term coined in 2015 by Paul Romer, a Nobel laureate in economics, to describe mathematical language that is designed not to clarify but to mislead. “The beauty of mathematical language is its ability to convey the truth about the world in simple, clear language—E = MC2,” Noah Giansiracusa, a professor of mathematics and data science at Bentley University, told me. I read his tweets over and over again and still have no idea how to parse them or what he was trying to say.

Giansiracusa translated Altman's words into notation. "With C for creative things, R for remixing things from the past, Q for the quality of the feedback loop, and N for the number of iterations, did he mean C = R + epsilon*Q*N or C = (R + epsilon)*Q*N?" Giansiracusa said Altman's wording didn't make the order of operations clear. "Does the 'and N' -- and the number of iterations -- mean something other than multiplication? Or does it mean ..."

May 22, 2023: Photo with Spanish President Sánchez. Photo: Pool Moncloa/Jorge Villar

06. The future of OpenAI

On June 22, 2023, Altman donned a tuxedo and attended a White House dinner with his partner Oliver Mulherin.

He looks like something out of a 1990s time-travel movie, physically too small and too young to carry his powers. But he’s mostly acting. He looks sharp in his tuxedo, a stark contrast to another famous Sam from “Silicon Valley,” Sam Bankman-Fried, whose scruffy aesthetic now looks like evidence of moral decay. Since Altman took over as CEO of OpenAI, the company has not only downplayed its nonprofit identity, but it’s also no longer very open, and no longer a company. It’s no longer very open, no longer making its training data and source code public, no longer making it possible for others to analyze and improve its technology. It’s abandoned its work “in the greatest good of humanity.” But what can be done? At a global tour in Munich, Altman asked a packed house whether they wanted OpenAI to open source its next-generation LLM, GPT-5, once it’s released.

The audience answered in the affirmative.

"Wow, we definitely wouldn't do that, but it's interesting to know," Altman said.

In his office in August, Altman was still spouting his ideas. I asked him what he had been working on for the past 24 hours. “One thing I was working on yesterday was: We’re trying to figure out if we can align AI with human values,” he said. “We’ve made good progress with the technology. Now there’s a harder question: Whose values?”

He also has lunch with the mayor of San Francisco, works on reducing his 98-page to-do list, and lifts weights (although, he says, somewhat resignedly, "I've given up trying to get really big"). He welcomes new employees. He has dinner with his brother and Ollie. He goes to bed at 8:45 p.m.

Imperialism is disorienting in its beautiful guise. He told reporters that one of Altman's most cherished things was the mezuzah (holy scroll on the doorpost, a Jewish talisman posted on the doorstep) that his grandfather had kept in his pocket all his life. He and Ollie want to have children as soon as possible, and he likes big families. He laughs so hard that sometimes he has to lie on the ground to catch his breath. He "has to find ways to incorporate people's wishes into our construction." He knows that "artificial intelligence is not a pure story with only benefits", "something will be lost here", and "loss aversion is super friendly and natural. People don't want to hear stories about themselves being victims."

His public persona, he said, "is only relevant to me."

References

https://nymag.com/intelligencer/article/sam-altman-artificial-intelligence-openai-profile.html