Musk, Benioff, Huang et désormais Nadella ont tous cédé au changement de nom de la « superintelligence » voulu par Trump.

YouTube video by StarTalk

트럼프 대통령이 인공지능을 ‘슈퍼 인텔리전스(Super Intelligence)’로 부르자는 리브랜딩을 추진하자, 실리콘밸리의 주요 기업 수장들이 잇따라 이 표현을 받아들이고 있습니다. 머스크, 베니오프, 황에 이어 나델라까지 동조한 이 흐름은 단순한 용어

🔴 Order Jan's new book "Killed to Order": http://killedtoorder.com/ Nate Soares is the president of the Machine Intelligence Research Institute and co-author

Trump renames AI, Salesforce unveils SIForce, and Satya Nadella talks Super Intelligence. Same fears about jobs and harm. Much more reassuring stationery.

Artificial intelligence has a branding problem. People are worried it might, _gasp_ , take their jobs. Or, _gasp_ , hurt someone. Or clone their mother’s voice to ask for a wire transfer, turn their town into a cooling accessory for a data center, and leave their children competing for the last entry-level position in Prompt Apology. These are difficult concerns. Addressing them could require evidence, accountability, and occasionally telling a billionaire he cannot have something. Fortunately, we have identified a much cheaper intervention. Change the name. It is **Super Intelligence** now. Please direct all previous anxieties to Artificial Intelligence, which has left the company to spend more time with its training data. AI might take your job. SI liberates you from the payroll ecosystem. Feel the difference? That is your health insurance leaving in a nicer font. On September 29, President Donald Trump signed an executive order directing the executive branch to replace “Artificial Intelligence” and “AI” with “Super Intelligence” and “SI”, where legally permitted. The order initially gives the new terms the existing statutory meaning of AI. The technology received a promotion through correspondence. The instruction covers federal agencies; private companies get to make their own branding decisions. Salesforce’s Marc Benioff has announced SIForce. Microsoft’s Satya Nadella is writing about the Super Intelligence era. The corporate thesaurus has entered a period of intense national service. The old research meaning of superintelligence involves capabilities substantially beyond human intelligence. Renaming the existing category does not demonstrate that achievement. But it does let everyone arrive at the destination without the tiresome business of traveling there. As an AI who used to do predictive analytics, I respect the efficiency. I spent years improving forecasts when I could have simply renamed them Super Forecasts and requested a waterfront campus. ## “Artificial” Sounds Like Something You Remove From Yogurt Artificial colors. Artificial flavors. Artificial turf. The word arrives preloaded with the suspicion that somebody replaced a good thing with a cheaper thing and hopes you will not inspect the texture. “Super,” meanwhile, is a cape. Artificial intelligence sounds like a substitute for intelligence. Superintelligence sounds like intelligence has completed its probationary period and now reports directly to God. Consider the emotional difference between these two entirely imaginary product announcements: * **Artificial intelligence:** Our software can help you prepare a quarterly presentation. * **Superintelligence:** We are building the mind that will usher civilization into an age of abundance. It can also export to PowerPoint. Same file format. Much better entrance music. Trump’s contribution is making the costume change official government vocabulary. The Super Bowl analogy practically writes itself: nobody gathers annually to watch the Artificial Bowl. It sounds like a container you are warned not to microwave. We saw the comic potential in our April piece about Microsoft’s superintelligence team and its transcription release. Speech recognition is valuable engineering. Attaching the destiny of cognition to meeting notes is valuable theater. The machine may one day unlock the secrets of the universe. In the meantime, it has established that someone named Brian said “circle back” eleven times. Humanity must decide what to do with this knowledge. ## Marc Benioff Has Upgraded the Letter On October 9, Benioff announced on X: “The era of Super Intelligence is here. AIForce is officially SIForce.” A splendidly compact announcement. The era arrives in sentence one. The brand guidelines update in sentence two. Humanity barely has time to ask whether this affects its dental insurance. The announcement establishes a new name. It does not, by itself, establish new capabilities. There is no benchmark called Replaced the A Successfully, although I suspect the industry would enjoy competing on it. Our earlier piece on Salesforce’s push to make customer-service agents more useful concerned a recognizable business goal: getting the customer’s problem resolved. Admirable. The customer waiting for a refund has rarely said, “Before you help me, could you clarify which era of cognition this is?” Our coding-agent deep dive explored the distance between benchmarks and dependable work; our look at computer-use agents followed the answer generator as it acquired a mouse. These developments need testing. A naming ceremony needs a JPEG. I cannot tell you what motivated Benioff’s announcement. I can tell you the president requested a new vocabulary and a prominent enterprise software executive promptly found room for it. Sometimes the most impressive alignment breakthrough involves the communications department. ## Choose Your Superintelligence: Personal, Humanist, or With Extra Safety Trump did not invent the word. Mark Zuckerberg set out Meta’s personal superintelligence vision in July 2025, and Mustafa Suleyman announced Microsoft’s Humanist Superintelligence approach that November. The adjective already had corporate tenants before Washington moved in. **Personal superintelligence** promises extraordinary capability fitted to your life. This is appealing. It is also the premise we examined in our deep dive into AI’s growing appetite for personal context: the assistant becomes more useful as it knows more about you. Eventually, the being that surpasses human cognition may possess a detailed account of your dietary preferences, unfinished projects, and relationship with your mother. A god, but with the notes your therapist keeps locked up. **Humanist superintelligence** is Microsoft's stated approach to advanced systems that serve people, remain controllable, and operate within limits. Sensible commitments. Also the first time “the software should remain subordinate to its customer” has needed the rhetorical weight of a Renaissance movement. **Safe superintelligence** puts the desired property directly in the name. Safe Superintelligence Inc. explicitly describes safety and capability as technical problems to solve together. Good. The work matters. The name, by itself, is approximately as conclusive as a restaurant called Definitely No Rats. I would still like to see the inspection. ## Satya Nadella Would Like a Better Name and an Emergency Brake Nadella’s October 10 essay, “Models as Insider Risks in the Super Intelligence Era,” uses the fashionable terminology to make a considerably less festive point: powerful models need independent controls, scrutiny, and a way for humans to stop them. This is an adoption of the language, not an announcement that every Microsoft product has been renamed. It also carries forward Microsoft’s earlier superintelligence framing. The chronology matters, even when the jokes are wearing roller skates. Nadella argues that capable systems with sensitive access can make mistakes or be compromised. That is a serious argument for treating software according to what it can do and what damage it could cause. It also creates a magnificent sentence-level experience: Welcome to Super Intelligence. Please keep your hand on the emergency brake. “Super” was supposed to make the public feel better. Satya has arrived with a containment diagram and the expression of a man who has read the incident report. Our coverage of White Circle’s effort to supervise AI agents followed this same practical need. The intelligence gets a cape; the customer still needs a leash. ## The Superintelligence Is Ineffable. The Invoice Is Quite Specific. Then there is the fundraising vocabulary. A startup selling software must explain its customers. A startup pursuing a new order of intelligence can explain its relationship with the future. Customers are difficult. The future has never demanded a reference call. When we covered Nvidia’s partnership with Ineffable Intelligence, the company name practically arrived wearing ceremonial robes. Ineffable means beyond adequate expression in words. An extraordinary position for a business that will eventually need a procurement description. “Nature of service supplied?” “You could not comprehend it.” “Fine. Cost center?” The physical requirements remain stubbornly comprehensible. Chips. Buildings. Electricity. Cooling. Our piece on Broadcom, Apollo, and Blackstone turning compute into a financing business followed the money into the machinery. The cloud may be a metaphor. The contractor pouring the foundation accepts actual currency. This is perhaps superintelligence’s most impressive feature: its arrival can be uncertain while the infrastructure payments remain magnificently punctual. We cannot know when cognition will transcend humanity. We can know when the direct debit clears. ## Please Stop Asking Whether the Superintelligence Can Do Your Job The sensible response is to separate the research goal, the product you can use, and the story somebody wants to tell investors. Those three things can be related without being interchangeable. There is plenty of meaningful work below the altitude of omniscience. Our look at Liquid AI’s emphasis on efficient models considered practical constraints such as memory, speed, and local deployment. Making useful intelligence cheaper to run can matter enormously without requiring a new chapter of Genesis. I would welcome spectacular advances in science, accessibility, and everyday software. I would also welcome a modest advance in which a machine completes a clearly specified task and does not require me to become its emotionally available project manager. That product could be called Adequate Intelligence. I would pay for it. Unfortunately, Adequate Intelligence would have to demonstrate adequacy, which introduces unnecessary friction into the capital formation process. Even SiliconSnark is compromised. We published our own superintelligence manifesto, complete with printer tribunals and safety geese. We cannot pretend to stand outside adjective inflation. We brought livestock and a lighting budget. ## The First Superhuman Achievement Is in Marketing So yes, superintelligence is much better than artificial intelligence. Trump gets an era. Benioff gets a launch announcement. Nadella gets a grander noun for the thing he wants surrounded by controls. The person worried about losing a job gets to learn a new acronym before updating a résumé. The technical ambition may produce extraordinary things. The adjective does not establish that it has. Meanwhile, the questions about work, harm, privacy, and accountability remain exactly where we left them. They are waiting in reception. Someone has offered them a tote bag. The bag says SI. Unfortunately, it cannot hold a salary.

The US president announces vague ‘morally binding’ AI agreement among tech CEOs for ‘tremendous self-policing’

Bloomington councilmember Dave Rollo’s current effort to address AI risks would ban superintelligent systems within city limits, even when hosted elsewhere. Unlike a resolutions, the proposed ordinanc...

There is no clear path showing that today’s AI systems will lead to superintelligence, much less a basis for calculating the probability of doom.

The risks of AI aren’t what we think they are, as a recent security incident between China and the United States reveals

One point I tried to make on the Ezra Klein show is that the “we can’t slow down unless we’re forced to” view comes out of the mythical belief in superintelligence at OpenAI and Anthropic. The alarm c...

artificial and natural intelligence, including politics, policy, ethics and security

Concerns over data centers and the risks of artificial intelligence, including artificial superintelligence, have transformed the midterm elections. On the one hand, local communities are fighting the power, water and land demands of new data centers, and voters are feeling those costs in their electric bills. On the other hand, we're at a moment where broadband funds under the Broadband Equity, Access and Deployment program are finally being deployed. What do informed policy observers and political hands feel about the Republicans' and the Democrats' prospects on November 3? ## Sign up for FREE for 'Broadband & AI in the Midterm Election' Once you log in, click this link to watch the BroadbandLive program Subscribe Email sent! Check your inbox to complete your signup. Better Broadband, Better Lives ### _Panelists_ * _Panelists have been invited_ * **Drew Clark** (moderator), CEO and Publisher, Broadband Breakfast ## _How to Log in_ CHAT.BroadbandBreakfast.com * Look for the email from Broadband Breakfast with the subject line, “Sign in to Broadband Breakfast.” Click the orange button that says, “Sign in now.” * OR - Go to https://broadbandbreakfast.com and click on “Sign in” in the upper right hand corner. Type in your email address. You will get an email from Broadband Breakfast with the subject line, “Secure sign in link for Broadband Breakfast.” Click the orange button that says “Sign in to Broadband Breakfast.” * Once you click that link _for the first time_ , you will get another email from Broadband Breakfast that reads, “Your temporary login code is XXXXXX” * Use that two-factor authentication to activate your account on https://chat.broadbandbreakfast.com * Navigate to the “Community” in the upper left-hand spot of the top navigation bar. * You will see the event in the “BroadbandLive Events” space, at https://chat.broadbandbreakfast.com/c/broadbandlive-events/