New Gallup poll on attitudes toward AI: China 93% believe AI will mostly help (top 1) US 36% believe AI will mostly help (bottom 5)

if one billion people used personal ai agents, how many agents will companies need as a single personal agent may try to interact with dozens of different companies for a single task? now multiply that by the number of unique tasks an agent is commissioned for each day. now give those agents more autonomy so they’re proactive and not just initiating task when prompted but may be doing 10x-1000x more tasks in a day on their own. then there is the response:query token ratio which could be 100:1 and let’s not forget that those enterprise agents may themselves interact with other enterprises agents both within and between the enterprise and now that we have all of these agents and low cost agent infrastructure everywhere, jevons paradox really starts to kick in and they become ubiquitous and then some of those enterprise agents start initiating their own transactions … you can see where this is going

Goldman Sachs earlier said: "Token use by AI agents is expected to multiply 24X by 2030" AI agents are now creating the first serious cost test for the AI boom. As was reported this week, Uber and Microsoft are already rethinking expensive agent usage. A chatbot may answer once, but an agent plans, calls tools, checks results, edits mistakes, and repeats the loop. That loop can make one user request consume 10x, 50x, or even far more tokens than a normal answer. Goldman’s bullish case is that monthly token use could reach 120 quadrillion by 2030, while inference cost per token keeps falling 60%-70% per year. The fight is now between agent productivity and token waste.

training and inference are pulling from the same fleet but the fungibility only runs one way, since a frontier post-training run needs coherent clusters with real interconnect and inference doesn’t, which means training can always pull capacity down from serving but serving can never backfill training. so if you’re sprinting a post-training run, and gpt-5.6 landing 8 days after fable came back off the commerce hold looks like a sprint to me, you’re pulling capacity that was already allocated to serving, which is the whole thing, because a run that was scheduled is already carved out of the revenue plan and doesn’t cost you anything you were counting on. that capacity doesn’t book anything for the quarter even though it’s still depreciating and still drawing power, and you can’t really rent your way around it either because nobody is selling you 30MW of coherent training capacity on six weeks notice. so the only fleet you can actually move is your own. then july comes and the run finishes, capacity goes back to serving, the model ships, and Friar tells investors the run rate is at $40b and jumped 20% in July with business customers up 32%. that’s arr not recognized revenue so it isn’t apples to apples with the quarter, but the direction is clear. which to me says the soft quarter and hot july are the same story. the compute went into building the next model instead of serving customers, so q2 ate the cost and july got the payoff a ~400M gap (OAI’s missing revenue assuming 25% QoQ rev growth) annualizes to somewhere around 20-40MW, which against a fleet measured in gigawatts is a couple percent of the pool that can actually run the job. Arkady basically priced this on the $nbis call, $40-$50M/MW for time-bounded needs up to six months, specifically post-training sprints and model releases. that tier exists because the burst is real and recurring i’m not saying this is what happened, i don’t have the segment data, but a soft quarter followed immediately by a sharp reacceleration is exactly the shape you’d get if the compute went somewhere else for a few weeks and the alternative reading requires demand to have softened and then un-softened inside of six weeks, which is a stranger thing to believe than a quarter where compute was reallocated from inference to an unscheduled training run

It's basically ~standard terminology though? We've been talking about how AGI ("Artificial General Intelligence") would lead to ASI ("Artificial Super Intelligence") for a decade. Just drop the Artificial ("redundant, cleaner") and you're there :)

The United States is committed to diplomacy and dialogue on super intelligence. But international dialogue cannot be allowed to drift toward global governance. As @POTUS said, the U.S. totally rejects any attempt to construct a globalist scheme of control for super intelligence.

.@perrymetzger argues more capable AI could make cyberattacks and pandemics easier to stop, not harder: "We already are at the point where AI systems are providing a massive uplift in computer security across the board." "This last Patch Tuesday, I think there were around 975 major vulnerabilities patched. I, as a computer security person, think this is amazing. We're finally clearing out all of the dry fuel that could create a fire." "The way that Hugging Face defended itself was through the use of an AI system." "Shotgun sequencing of wastewater, of air, of soil samples, all over the world constantly. You can detect emerging diseases very, very fast. This was not possible without AI." "If your immune system shuts down, you start being eaten by microorganisms immediately. But you don't think about your immune system. Our society is kept together by this huge array of defensive systems." "AI is just going to end up becoming another thing that is incorporated into the defensive systems. Hopefully we save so many orders of magnitude more lives that it's worth it."

IDK if you polled some unbiased sample of experts in the field (granted, "experts" and "unbiased" are doing a lot of work there), I imagine most of them would roughly agree with the idea that AI models "think," at least for more advanced tasks.

the level of taste that opus 5.5 has is mind blowing and taste is quite obviously difficult to get right. it won’t be long before models eclipse human creativity and intelligence and tbh they’re kinda already doing it… just look at the creations you enjoy on this platform, are they human made?

He means, what happens when some EAs fake such a thing, and have pre-prepared media appearances set up to whip up maximum fear, and already have legislation drafted for various scared politicians to pass.

Guys, you shouldn't use frontier models for everything GPT 5.6 luna can summarize your excel spreadsheet

For Meta Muse to displace $BKNG and $ABNB, it would need to: (1) Access special unadvertised prices only available via Booking app and logged-in users. (2) Yes, an agentic AI agent can easily login and do that, but that doesn't mean they can find that $6 rental car deal elsewhere. (3) So they still have to book through Booking, or in the case of Airbnb, where most properties are exclusively listed. This friend and wife travel all the time, super creative with finding best deals, and they often find the best only in Booking app. Not website. Most likely scenario? $META does affiliate marketing arrangements with travel aggregators.

Very important that anti-AI protests avoid the omnicause shenanigans characteristic of wokeness. There are good reasons for conservatives and liberals, critics and champions of current US foreign policy, to oppose accelerationism. This is a transcendent issue.

Holy crap... Remember the Australian guy whose OpenClaw hacked his gym?! he is a THE EA CONFERENCE DIRECTOR SINCE 2015!! these people lie about everything!!

UPDATE: AIs have achieved the highest *possible* score on the Mensa Norway IQ test - 151 3 years ago: cognitively impaired human (64 IQ) 2 years ago: average human 1 year ago: genius human Today: literally off the charts Next year?

Geoffrey Hinton (Godfather of AI) says increasing AI productivity could actually make society worse More goods. More services. More of everything, for everyone. That is what a huge jump in productivity is supposed to deliver. The reason it won't has nothing to do with the technology failing. It has to do with who owns it. The productivity shows up. The profits go to the big companies and the rich. The people whose jobs disappear get nothing back. So the gap widens. And the gap is the actual danger here, not the machines. Hinton: "as soon as you increase that gap, you get fertile ground for fascism" He is not arguing against the technology. He says it will obviously help healthcare and education. His fear is narrower. If the profits only travel upward, something built to help everybody ends up making society worse.

I resigned from Google today. I enjoyed my work and loved the people, but my GDM team was working on a new generation of chips to make AI much faster and cheaper, and I think AI is already progressing too fast, so I had to quit.