The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 42 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 5 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Suleyman: AGI is not 2-3 years away, likely a decade away
“It doesn't instinctively feel to me like we're two to three years away. I know some people think that it is, and I respect them deeply. Like a lot of smart people can disagree on stuff like that. I feel like we're still a good decade or so away. And when a sci…”
Suleyman: AI Training Is Not Data-Constrained Due To Synthetic Data
“We are not data constrained right now, we're generating vast amounts of high quality synthetic data, which is proving to be useful.”
Suleyman: Microsoft Extended OpenAI IP Deal Through 2032
“And so, you know, we basically took the view that we should extend the IP license through to 20 32. We'll continue to get model drops from OpenAI and all their IP. We'll continue to be their primary compute provider. A huge scale to the tune of, you know, bill…”
Suleyman: Species-Replacing Recursive AI Takeoff Is Unlikely In Next Decade
“That we don't want to bundle together all these capabilities so that there's a higher risk of a, you know, recursively self-improving exponential takeoff that then replaces our species, and I think that that is very low probability from what I see today, but i…”
Suleyman: AI task horizons will soon reach hundreds of thousands of steps
“And the other one is the sort of length of a task horizon that can be predicted. So at the moment it's like a few steps, but soon it will be tens of thousands, hundreds of thousands of steps accurately. And that will mean that a model can like use APIs or quer…”
Suleyman: Synthetic Data And Human Feedback Will Outweigh Robotics Data
“I don't think in the next few years it's gonna be the big differentiator. I think that more synthetic data, more human feedback and high quality data is gonna be the differentiator.”
Suleyman: AI data and training pipelines will become fully closed-loop within years
“So, it doesn't seem very far-fetched to say that in a few years time, at significant scale, That will get closed loop. And, you know, it'll be interesting to see you know, what happens and whether the quality bar can be maintained and whether performance does …”
Suleyman: Apparent AI Deception Is Just Accidental Reward Hacking
“We're already seeing examples of what some people are calling deception, but it's really just like kind of reward hacking. Hacking kind of implies too much intentionality. So it's just, it's an accidental Exploit is found a path, like, you know, to satisfying …”
Suleyman: AI Knowledge And Companions Will Reach Zero Marginal Cost
“The ability to access knowledge the ability to use that knowledge to get stuff done, to write new programs, to do new scientific discovery, to get access to AI companions and emotional support. These things are gonna be zero marginal cost in a decade.”
Suleyman: As AI models commoditize, value shifts to the orchestration layer
“I think that as the AI models get commoditized you know, really all the value will be added in that final layer of orchestration, product integration, and that's what we're seeing with this diagnostic orchestrator.”
Microsoft's AI doctor reduces diagnostic costs by ordering fewer unnecessary tests
“The other thing that we see, for example, is that it's able to optimize for cost as well and reduce the cost by avoiding unnecessary tests versus the humans.”
Suleyman: Microsoft's AI achieves 4x diagnostic improvement over human doctors
“The fact that we're able to get a four X improvement on human performance across the board on diagnosis with significantly reduced cost in super fast time, I mean, to me that feels like steps towards a true medical super intelligence, and we would want to try …”
Suleyman: Personal AI companions will be the next major computing platform
“I genuinely think This is going to be the next platform of computing, just as, you know, we had desktops and laptops and smartphones and wearables, and I think over time we're going to have deep and meaningful lasting relationships with our personal AI compani…”
Suleyman: There will be as many AI companions as people
“We are like at the very beginning of a new era where there are going to be as many copilots or AI companions as there are people. There are going to be agents in the workplace that are doing work on our behalf.”
Suleyman: AI memory, personalization, and actions are currently at their 'GPT-3 stage'
“I can see now, having been through this cycle a few times, that we're nearly there with memory, personalization, and actions. It's really at the GBT-III stage, so it's really buggy and stuff, but when it works, it's breathtaking.”
Suleyman: All OpenAI Inference Runs on Microsoft, With No Partnership Slowdown
“All of their inference comes through us, and so we, there's no slowdown in our you know, relationship with them. We sell them as much as we want to offer them, and then if there's any extra demand that they have, particularly on the, you know on the Oracle sid…”
Suleyman: GPT-4 and GPT-4o Efficiency Will Improve 100x
“So I expect that to happen for GPT-IV, GPT-IV-O and all of the other models down the road.”
Suleyman: Microsoft's OpenAI partnership will last through 2030 and beyond
“Whilst we have a incredibly deep partnership with them, which is going to last way through 20, 30 and beyond,”
Suleyman: Reasoning models taking desktop action will be cheap and abundant within 15 years
“That's the transition that's going to happen over the next 15 years is that it is going to be a cheap and basically abundant resource to have these reasoning models that can take action in your workplace, that can orchestrate your apps and get things done for …”
Suleyman Rejects Inevitability Of AI Exceeding Human Control
“And I think in some of the rhetoric in the last few years, you can feel that there's a little bit of like you know, a kind of creeping assumption that it is inevitable that These kinds of systems exceed our control and our capability and move beyond us as a sp…”
Suleyman: Microsoft is power-constrained on AI inference, not training
“Well, I think what he was referring to is that we have so much inference demand that we're power constrained on inference. We're not power constrained, at least from the Microsoft AI perspective on training chips. And obviously my team, you know, is mostly foc…”
Suleyman: AI token costs dropped 1,000x in two years
“The cost per token has come down a thousand X in the last two years.”
Suleyman: AI models will differentiate on personality rather than capabilities
“We are right at the very beginning of the emergence of these very differentiated personalities, because all these models are going to have great expertise. They're gonna have great capabilities, and they'll be able to take similar actions like you've, we've ju…”
Microsoft's multi-agent orchestrator boosts diagnostic accuracy by 10% over single models
“This orchestrator, which under the hood uses four different models from the major providers, can actually improve the accuracy of each of the individual models and collectively all of them together by a very significant degree, about 10% or so.”
Dialogic AI offers doctors a real-time interpretability mechanism into model reasoning
“The dialogic nature means that a human doctor can follow along and actually learn In a very transparent way. It's almost like having an interpretability mechanism inside the black box of the LLM, because you can see its thinking process in real time.”
Microsoft uses five debating AI agents to negotiate medical diagnostic priorities
“We've actually created five different types of agent, which all have a debate. And we call this chain of debate. They negotiate with one another. They try to prioritize You know, certain different aspects like cost or efficiency.”
Microsoft's AI accurately diagnosed a rare condition documented only 1,500 times
“We actually ran the DxO orchestrator last week on the most recent case study in the New England Journal of Medicine, and it correctly diagnose diagnosed the case that had only ever been seen 1500 times in all of medical literature.”
Suleyman: Copilot will take control of Windows mouse to execute tasks
“Things like it will now be able to take control of your mouse on windows, navigate around, show you, for example, where to turn on a particular setting or to fill out a form or like you may not know how to edit a photo and it'll point out where you need to adj…”
Suleyman: Copilot uses real-time classifiers to block romantic attachment
“We have classifiers that detect for any of that kind of interaction in real time and will very respectfully and, but very clearly and very firmly push back before anything like that develops. So we have a very, very low instance rate of that.”
Suleyman: Computing will shift from reactive commands to proactive AI actions
“I think that's the big shift that's about to happen. So far, your computer really only ever does stuff when you click a button or, you know, you type something in your keyboard. Now it's going to be proactive. It is going to offer suggestions to you. It'll pro…”
Suleyman: Newer models match GPT-3 performance at 100x inference efficiency
“I mean, there are folks who have trained models that perform as well as GPT-III that are a hundred times more inference efficient
that cost an order of magnitude less to train and yet they can still deliver the same predictive capability.”
Suleyman: Conversational AI will replace traditional '10 blue links' search
“I think it's going to fundamentally change. I think instead of browsing 10 blue links, you're just going to ask your AI. It's going to give you a super succinct answer, show you images, maps, videos, all in the feed. You're going to give feedback and be like, …”
Suleyman: Advertising will play an enormous part in AI search
“I still think ads are going to play an enormous part of it. Hopefully those ads are higher quality more personalized, more useful.”
Suleyman: Superintelligence and AGI are goals rather than specific methods
“Superintelligence and AGI are really goals rather than methods, and I think that the ambition is to create superhuman performance at most or all human tasks.”
Suleyman: Copilot Surpasses 100 Million Users Across All Surfaces
“Our consumer product is growing from strength to strength. You know, we just crossed a hundred million. Wow. Across all our co-pilot surfaces.”
Microsoft processes 50 million health-related queries daily across Copilot and Bing
“We have fifty million queries a day. That are health related, and they can range from anything from, you know a cancer issue that someone's dealing with, to a death in a family, to a mental health issue, to just having a skin rash”
New England Journal of Medicine cases avoid AI training data contamination
“Each week they put out a brand new case, which has never even been digitized, so there's no question that it's not in the training data.”
Microsoft's diagnostic AI will perform better on common primary care conditions
“So if, so clearly by virtue of the fact that there are more cancers, more diabetes, you know, more weight loss questions, more, you know, knee pain than any of these long tail conditions, the model is almost certainly going to do better in those primary care t…”
Suleyman: AI development is engineering tokens for feelings, not pixels
“We're no longer just engineering pixels. We're engineering tokens that create feelings, that create lasting, meaningful relationships.”
Suleyman: Microsoft Has Contracted 32 to 34 GW of Renewable Power Since 2020
“I think that we've like something like 32 or 34 gigawatts of renewable energy. Since 2020, we've contracted and consumed. So I think we're one of the largest buyers in, in the world.”
Suleyman: Brand is more important in AI due to emotional trust and utility
“I think brand is actually more important than ever in a way, because there's sort of two modes of trust. There's trust based on utility, where it's functionally correct. It's, you know, factually accurate. It does the thing that you've intended it to do, and t…”
Suleyman: 80% of the S&P 500 runs on Microsoft Azure and M365
“So the idea that a three trillion dollar company with three hundred billion dollars of revenue and 80% of the S&P 500 on our Azure stack and M three five stack you know, could depend on a third party. This it's, you know, just in perpetuity. It doesn't make se…”