At the center of each empire there is an ideology, a system of beliefs that pushes the system forward and justifies the expansion, even if the cost of this expansion directly challenges the declared mission of ideology.

For European colonial powers, it was Christianity and the promise to save souls as they extracted resources. For the Empire Ai today, it is general artificial intelligence “to benefit all humanity”. And Openi is his evangelist head, which spreads zeal throughout the sector in a way that has reformulated the way the IA is building.

“I was interviewing people whose voices we shake up from the fervor of their beliefs taking place”, Karen Hao, journalist and author of best seller of “Empire of Ai”, He told Techcrunch we have Recent episode From Equity.

In his book, Hao compares the artificial intelligence industry in general and in particular Openi to an empire.

“The only way to truly understand the scope and scope of Openi’s behavior … It is actually recognizing that they have already become more powerful than practically any national state in the world and have consolidated an extraordinary love not only the political power of economic power,” said Hao. “They have terrarformed the earth. They are reducing our geopolitics, all our life. And therefore you can only describe it as an empire.”

Openii has described Agus as “a highly autooman system that exceeds humans in the greatest number of economically precious works”, one who in some way “will raise humanity by increasing abunance, disturbing the economy and helping in the discovery of new scientific knowledge that limit the possibility”.

These promised nebulosis have fueled the exponential growth of industry: its enormous resources needs, oceans of scraped data, aimed and arranged energy networks available to release systems not tested in the world. All in the service of a future that many experts say may never arrive.

Hao says that this path was inevitable and that downsizing is not the only way to obtain greater progress in the AI.

“You can also develop new techniques in algorithms,” he said. “You can improve existing algorithms to reduce data amnt and calculate they have to use.”

But that tactic would have meant to sacrifice the speed.

“When you define the search to build a beneficial act like the one in which Victor takes everything – which is what Openai did – then the most important thing is the speed compared to anything else,” said Hao. “Speed ​​on efficiency, speed compared to safety, speed compared to exploratory research.”

For Openi, he said, the best way to guarantee speed was to take existing techniques and “just do the intellectually economic thing, which is to pump more data, more supercomputer, in the existing techniques”.

Openi fixed the internship and, rather than he remained behind, other technological companies have decided to lift themselves.

“And because the artificial intelligence industry has successfully captured most of the best artificial intelligence researchers in the world and those researchers have not existed for some time in the academic world, so you have a whole discipline that is now modeled by the agenda of these companies, rather than a real scientific exploration,” said Hao.

The expense was and will be astronomical. Last weekend, Openai said it expresses burning $ 115 billion in cash By 2029. Meta said to Juny that he would have spent until $ 72 billion We build artificial intelligence infrastructures this year. Google expects to hit $ 85 billion In capital expenses for 2025, most of which will be spent to expand the integrity and cloud infrastructure.

In the meantime, the posts of the objectives continue to move and the highest “benefits for humanity” have still materialized, even if the damage mount. Damage the loss of work, the concentration of wealth and To the chatbots that feed Délusioni and Psychosis. In his book, he also workers in development in developing countries such as Kenya and Venezuela who were exposed to disturbing content, including the material for sexual abuse on minors, and Wrere paid very low wages – about $ 1 to $ 2 years – in roles such as moderation and data labeling.

Hao said it is a false compromise to put the progress of the AI ​​against current damage, especially when other forms of AI have offered real benefits.

He gave Google Deepmind a tip to the Nobel Prize AlphafoldWhich is trained on the sequence data of amino acids and complex proteins folding structures and now the 3D structure of proteins can be accurate from their amino acids – deeply useful for the discovery and understanding of drugs of drugs.

“These are the types of artificial intelligence system we need,” Hao said. “Alphafold does not create mental health cise in people. Alphafold does not lead to colossal environmental damage … because it is trained on substantially less infrastructures. [the datasets don’t have] All the toxic shit you dried when you were scraping on the internet. ”

OR Beat China in the race aiso that Silicon Valley can have a liberalizing effect on the world.

“Literally, the opposite happened,” said Hao. “The gap continued to fill the United States and China, and Silicon Valley was a illiberalizing effect on the world … and the only actor who came out unharmed, it could be supported, it is Silicon Valley herself.”

Of breed, many will support that Openai and other artificial intelligence companies have benefited humanity by releasing chatgpts and other large language models, which promise enormous gains in productivity by automating tasks such as coding, writing, research, customer support and other knowledge activities.

But the way in which Openai is structured not partly with part, partly complicates the way in which it defines and measures its impact on humanity. And this is further complicated by the news this week Openii has reached an agreement with Microsoft This brings him closer to the tests.

Two Treen Openii Safety researchers told Techcrunch to fear that the Ai Laboratory was to confuse it on a mission for professional and non -anti -professional, that because people have fun using chatgpt and other products built on LLMS, this appears the benefit box of humanity.

Hao echoed to these concerns, describing the dangers of being so consumed by the mission that reality is ignored.

“Even if the evidence accumulates that what they are building is actually damaging the meaning of love for people, the mission continues to paper all this,” said Hao. “There is something really dangerous and dark in this, of [being] So wrapped in a system of beliefs that you have built that you lose contact with reality. “