Leopold Aschenbrenner looks a bit like the villain of a 90s Bravo photo love story, blonde hair, big smile. But the scientist, who originally comes from Germany, is above all a high-flyer type: According to his LinkedIn profile, Ashenbrenner has At the age of 19, Ashenbrenner graduated at the top of his class from Columbia University in the US. That was 2021. After a few jobs in research and investment, Ashenbrenner started at OpenAI in the superalignment field in 2023. This is the team that deals with the safety of superintelligence. After a year and a half and many quarrels in the AI company, Ashenbrenner was fired.
Advertisement
Shortly after, the Economist has a 165 page essay publishedwhich is getting a lot of attention at the moment. Ashenbrenner is clearly one of those who could be described as AI ultras. For example, he believes that in just a year or two, AI will become smarter than most college graduates. For example, by the end of the decade, AI will become smarter than all of us, including Ashenbrenner himself, “we will have superintelligence in the true sense of that word.” It also seems a bit suspicious when he writes about “The Project,” that is, the project he writes in capital letters. This project leads to either competition or “total war” with the Chinese Communist Party.
Reasons for Aschenbrenner’s ouster from OpenAI
It is said that this concern was the reason Ashenbrenner was ousted from OpenAI. In a podcast, he said that there was a security incident at OpenAI, after which he put together a memo that he plans to share with some board members. The memo discussed the possible theft of “critical algorithmic secrets from foreign actors.” Business Insider writes. The human resources department described his concerns about espionage by the Chinese Communist Party as “racist and unconstructive” and warned him. In addition, according to Aschenbrenner, he was accused of passing on confidential information. It is said that this was about a brainstorming document that he shared with scientists to obtain their evaluation.
This is perhaps the most exciting thing to share about the essay. Aschenbrenner has been in contact with various tech entrepreneurs, employees and investors in San Francisco and Silicon Valley. In his essay you can read how some of these people see the future, what they talk about – and what mindset prevails in them.
While some scientists believe that it will take a long time to make AI truly intelligent and reliable, Aschenbrenner believes that the next breakthrough is just around the corner. And this is despite the fact that he himself writes that “the magic of deep learning is that it just works.” This magic is called a “black box” by other experts because no one is quite sure how and why it works so well. And that’s why it’s really clear to many that progress will not necessarily be linear. For example, Meta’s AI chief Naila Murray said that she sees clear limits on large language models and that scaling does not reflect the intelligence that we humans have. And even if AI is constantly making good progress, there are still some limiting factors: energy and data.
Where does the energy for AI come from?
Aschenbrenner is also clear when it comes to the resources needed for AI. He believes that by the end of the decade, individual training clusters will be used, which will then consume more than 20 percent of all electricity produced in the United States. Nvidia’s rise and sale in the stock market is just the beginning. “Trillions of dollars in investment will produce hundreds of millions of GPUs per year,” the essay says. “Where do I get 10GW? (Power for $100B+, Trending 2028 Clusters) is a popular topic of conversation in SF,” Aschenbrenner says. He carefully calculates the needs and investments. “Surprisingly, even a 100GW cluster is easy to implement,” says Aschenbrenner. Aschenbrenner relies on natural gas; you’ll only need about 1,200 new boreholes. For comparison: According to the EnBW, the total net output of all generation plants in Germany is around 252.8 gigawatts (GW).,
Read this also
Aschenbrenner’s calculations also include the fact that OpenAI’s value continues to grow. While the company was worth one billion US dollars in August 2023, it was already worth two billion US dollars in February 2024. So Aschenbrenner expects a valuation of ten billion US dollars by the end of 2024 or the beginning of 2025. And he writes: without any new “next-generation models”.
Bill Gates also recently commented on the topic of energy. His calculations are also similar to those of Silicon Valley. Gates says that AI will compensate for what it consumes. That is, AI algorithms will lead to savings through better calculations and progress. OpenAI CEO Sam Altman is already investing in a start-up that relies on fusion power plants. Energy purchase contracts should be negotiated with OpenAI, which does not yet exist. Helion has already signed contracts with Microsoft. The power plant in Washington state is scheduled to be operational by 2028.
According to his own website, Aschenbrenner himself has now founded an investment company that specializes in AI, of course. If the hype continues and expectations stay high, he too could benefit.
In Aschenbrenner’s view, the data that AI needs comes from robots, which are manufactured by robots, in factories that the robots built.
(EMW)
