Today I\u2019m talking with Mustafa Suleyman, the CEO of Microsoft AI. And I\u2019m actually going to keep today\u2019s intro short \u2014 I\u2019m working from my wife\u2019s family farm this week, as you\u2019ll see in the video, but also this is a real burner of an episode.<\/p>\n<\/div>\n
We covered everything from Mustafa\u2019s approach to training new models to his criticisms of Anthropic talking about Claude as though it is conscious. Of course, we also talked about Microsoft\u2019s relationship with OpenAI, how Mustafa is thinking about all the negative polling and political pushback around AI right now, and whether any of the consumer products are good enough to overcome it.<\/p>\n<\/div>\n
Like I said, it\u2019s a burner.<\/p>\n<\/div>\n
Okay: Mustafa Suleyman, CEO of Microsoft AI. Here we go.<\/p>\n<\/div>\n
This interview has been lightly edited for length and clarity.<\/em><\/p>\n<\/div>\n Mustafa Suleyman, you are the CEO of Microsoft AI. Welcome back to <\/strong>Decoder<\/em><\/strong>.<\/strong><\/p>\n<\/div>\n Great to be with you again.<\/p>\n<\/div>\n I\u2019m very excited to talk to you. Our previous conversation was one of my favorite conversations \u2014 about AI, how it should make us feel, and what it\u2019s for \u2014 that I\u2019ve had in all the conversations we\u2019ve had. <\/strong><\/p>\n<\/div>\n There are some big changes at Microsoft, maybe some very important recontextualization about how people feel about AI that I want to talk to you about in particular. And then there\u2019s Microsoft Build, the big Microsoft developer conference, which <\/strong>featured lots of new announcements<\/strong><\/a> and lots of big ideas about what computers are for and maybe where they should be that I want to get into.<\/strong><\/p>\n<\/div>\n Let\u2019s start at the very start. This is some deep <\/strong>Decoder <\/em><\/strong>stuff that is important to understand before all the rest of it. Since you joined Microsoft, you have restructured how AI works there. Your role has changed. The last time I talked to you, you were in charge of a bunch of consumer products. That has since been set aside. You\u2019re now training new models; you\u2019re on the frontier. <\/strong><\/p>\n<\/div>\n Explain how Microsoft AI is structured now and how it\u2019s structured inside Microsoft.<\/strong><\/p>\n<\/div>\n I guess the last 15 to 18 months or so we\u2019ve been on this journey to reestablish our relationship with OpenAI, and it\u2019s taken a minute. I think it culminated in a new contract that we got done in October of last year. And there were lots and lots of different provisions in that, including cementing and extending the partnership, but crucially freeing us up to be able to pursue superintelligence independently as well as keep buying and licensing their models.<\/p>\n<\/div>\n So since October, I\u2019ve been assembling the Superintelligence team, building clusters of sufficient scale to train frontier models, and hiring a team focused on superintelligence. And so that was quite a big shift for us because it sort of enabled me to focus just on the superintelligence mission, and that has then culminated in a few things that we announced this week at Build. We have seven new models across all the modalities and so on. So it\u2019s been a pretty big shift, and I think a long time in the planning, and a great relief for us to now be in the game and pursuing the absolute frontier over the next few years.<\/p>\n<\/div>\n Was this the plan when you were hired at Microsoft?<\/strong><\/p>\n<\/div>\n It\u2019s certainly been the plan for the last 18 months. I mean, I think the relationship with OpenAI has gone through lots of ups and downs. And in many ways, I think it is going to go down as one of the most successful partnerships in history. It\u2019s been great for OpenAI, and it\u2019s been great for Microsoft, and all good relationships evolve, and I think this is just the next stage in our evolution.<\/p>\n<\/div>\n Let me ask you about that evolution specifically. We all just saw the trial between Elon Musk and OpenAI and Sam Altman. Microsoft was <\/strong>involved in that trial<\/strong><\/a> in the sense that every so often a lawyer from Microsoft would stand up and say, \u201cAnd we weren\u2019t around.\u201d And someone would say yes, and that was that.<\/strong><\/p>\n<\/div>\n But obviously, what came out during that trial, what has been clear during this entire time, is that the original notion was that OpenAI would be a research lab and provide models, while Microsoft would build the products. Microsoft had expertise in going to market; it had expertise in enterprise, it was trying to regain a foothold in consumer in a variety of ways. This would be a platform shift, and the research work would be over at OpenAI, and the product work would be inside of Microsoft. <\/strong><\/p>\n<\/div>\n That\u2019s the thing that changed: OpenAI wanted to make more and more consumer products. Obviously, given your new role and your new focus, Microsoft more and more wants to make its own models. Why the split? What didn\u2019t work in that relationship?<\/strong><\/p>\n<\/div>\n I mean, I think OpenAI is led by an incredibly ambitious founding team, and Sam himself. And so naturally, as they started to get more traction and generate a ton of revenue, they saw opportunities to go full stack. So it wasn\u2019t just that they started working on consumer products. Obviously, ChatGPT was incredibly successful. They also started working on their own data centers. They started creating their own chip. There are lots of rumors flying around about their own consumer hardware devices. They started taking models direct to market through ChatGPT Enterprise. So across the stack, they were kind of broadening way beyond research over the last two, three, four years. And naturally, the same is also true for Microsoft. I mean, I think the partnership\u2019s now five or six years old, and still has another four, five, six years to run.<\/p>\n<\/div>\n Likewise, we\u2019re one of the largest technology companies in the world. We have 493 of the 500 largest companies that store and process most of their data on our systems, use Azure, use M365 and Teams. I think people often underappreciate how enormous we are and how big our distribution is in enterprise. And so, long term, and I do mean over five, six, seven, 10 years, we have to make sure that we\u2019re completely sustainable, and we\u2019re not just a recipient of somebody else\u2019s IP that we then slightly modify and adapt and put into production for our products, but we actually can stand on our own two feet and create world-class models.<\/p>\n<\/div>\n I mean, superintelligence is coming. I think it\u2019s just around the corner. And so I think it\u2019s going to be basically the most valuable technology of all time. There\u2019s sort of no way that, long-term, we could be structurally dependent on a third party for providing that IP for all eternity.<\/p>\n<\/div>\n So that\u2019s been the transition that obviously was triggered when OpenAI and so on had their board issue. But then as I came in and my team came in, we started building that out, we\u2019re on that transition. And I think we\u2019re in a great spot because we can take a fairly steady, careful, long-term optimal position, both for OpenAI, which I think has done incredibly well out of this, and for us.<\/p>\n<\/div>\n I want to spend some time on superintelligence. I just want to put a pin in it now because I just want to kind of understand the transition for one more turn here.<\/strong><\/p>\n<\/div>\n There\u2019s a moment in the trial, sort of very funny message from Microsoft CEO, Satya Nadella, he says, \u201cI don\u2019t want to be Intel and have OpenAI be Microsoft,\u201d which is very funny in the context of Microsoft CEO himself saying, \u201cI don\u2019t want to be the provider, and have them be the platform that provides all the value and collects all the value and maybe we\u2019ll be swapped out. I don\u2019t want ChatGPT to run on Azure, and then OpenAI will get all the value, and then maybe they can swap us out,\u201d just as what happened with Windows and Intel over time.<\/strong><\/p>\n<\/div>\n Is that a realization? Did Nadella come to you? What was that meeting like where you said, \u201cOkay, OpenAI had its board issues. We need to get back on the frontier and stand on our own two feet.\u201d What did that conversation look like, and how was that decision made?<\/strong><\/p>\n<\/div>\n I mean, obviously that\u2019s Satya\u2019s decision as well as Amy, Brad, and many other people in the company. But I think it\u2019s as with anything: these are slow-moving changes in the company, as it comes to realize that the direction that we\u2019re taking needs a little bit of tweaking and adjustment. And so that was happening way before the November board incident, and I think it just builds up over time as you look at the kind of constellation of different fronts around which we\u2019re competing directly, increasingly, and all the tension that comes from that. But also just knowing that partnerships like that don\u2019t last forever.<\/p>\n<\/div>\n I mean, OpenAI wants to be a trillion-dollar public company, has incredible revenues, and is growing like crazy. They want to have the freedom to operate and be able to buy compute from all sorts of other places, build their own compute, and partner with whoever they want. So the contract was formed at a time when the companies were very different in terms of size and scale and balance of needs and stuff. I think it made sense for that moment, but then it became pretty clear that this is something that we have to be able to own and control ourselves and do right by our own customers.<\/p>\n<\/div>\n As I said, we have an incredible distribution on enterprise, which I think is just completely unrivaled in the world. And so we have to make sure we\u2019re building the best things for our customers. That looks slightly different to a company that has been jointly optimizing both for the consumer, with ChatGPT, and for the enterprise, and also for the fundamental science mission of superintelligence, which includes a whole bunch of different directions which are overlapping but could arguably be said to be orthogonal to the consumer and the enterprise directions too. Naturally, I think that\u2019s how partnerships evolve, and they get reset periodically.<\/p>\n<\/div>\n Yeah, but building a frontier model is very expensive, I\u2019m told. Reliably told, this is a very expensive project. At some point, Amy Hood, the CFO of Microsoft, has to say, \u201cYep, you\u2019ve got the budget.\u201d When did that happen? Was that just a text message? Was there a meeting? Tell me about the specifics there.<\/strong><\/p>\n<\/div>\n I think, look, we sort of made the decision in the early part of last year, which obviously informed all the contract negotiations, which then all got resolved and signed in October. And it is a significant investment, but we have a long time to make it. I mean, we\u2019ve already made significant investments in our own self-sufficiency mission.<\/p>\n<\/div>\n Our Maia 200 chip is actually an outstanding chip, as one example, right? We are now able to manufacture and ship a chip that is 30 percent cheaper than a GB200 inside of our own clusters. And now that we can co-design our own models with it, the MAI-Thinking-1 model<\/a> that we\u2019ve just released actually delivers 1.4x performance per watt improvement on top of the 30 percent improvement that you get from running on a Maia 200 once we co-optimize the models for our tasks.<\/p>\n<\/div>\n So the value of making sure that you own and control your own stack and direct the entire co-design effort end-to-end for the use cases that are most important to us \u2014 which is obviously agentic coding, our developers, our enterprises \u2014 that clearly pays the dividends that justify the investment that we have to make over the next few years.<\/p>\n<\/div>\n You said self-sufficiency mission, which is a very polite way of saying you want to stand on your own two feet; you want to do your own thing. I\u2019m told there\u2019s some controversy inside of Microsoft about a line <\/strong>my colleague Hayden Field wrote in a piece describing Build<\/strong><\/a>. I\u2019m just going to read this. This is from Hayden. It\u2019s a great line. She said, \u201cThis year\u2019s Microsoft Build had the vibe of a freshly single divorc\u00e9e posting a thirst trap on Instagram.\u201d<\/strong><\/p>\n<\/div>\n The breakup is completed, and it\u2019s time to flex. Here\u2019s our new model. We\u2019re going to stand on our two feet. You\u2019re out there saying you\u2019re going to build models at the frontier and compete with the leading labs. Is that the feeling inside of Microsoft that you\u2019re free to be on your own?<\/strong><\/p>\n<\/div>\n Definitely not. No, not at all. Look, I mean, obviously that\u2019s a cool headline and a fun phrase. But the reality is that we are in partnership with OpenAI for years and years to come. I mean, we\u2019re running way north of 2030. They still produce the best models in the world. GPT-5.5 is an outstanding model. The Codex, the cybersecurity models that are coming through, are amazing, and they\u2019re powering the majority of what we do.<\/p>\n<\/div>\n So naturally, that\u2019s going to continue. And so I think that\u2019s just a natural course of these sorts of partnerships. I don\u2019t think it\u2019s anything untoward or surprising. I think OpenAI is very understanding and supportive of that. I mean, they\u2019ve obviously been an incredibly fast-growing company, and they understand that we have to pursue our own agenda as well. So it\u2019s very normal.<\/p>\n<\/div>\n Let me ask you the other <\/strong>Decoder <\/em><\/strong>question, and then I want to get into the announcements at Build, and certainly superintelligence.<\/strong><\/p>\n<\/div>\n The last time we spoke, you said your framework for making decisions operated on a six-week cycle, given how fast AI was moving. That made sense then. Things have settled, maybe. Maybe some things are more in focus. What is your decision-making framework now?<\/strong><\/p>\n<\/div>\n We still operate by the same cycle rhythm. At the end of each cycle, we have a one-week meetup in person. I\u2019m a real believer in this, even though we\u2019re still an in-office culture, four days a week. In fact, the week after next, my entire Superintelligence team comes together in person in Boston for four days. That is for all of our retrospectives on how Build went, what we learned, what we didn\u2019t get right, what we need to improve, our planning for the next cycle, which is going to run for eight weeks this time with a one-week meetup afterwards, and that\u2019s all laid out for the entire year. So the whole organization knows that that\u2019s the rhythm by which we operate.<\/p>\n<\/div>\n And I think it\u2019s actually really important to emphasize that timeframe, because quarterly planning gets a little bit blurry and a bit abstract. I think six to eight weeks, depending on where it falls in the calendar, is actually the optimal time for making very clear, fortifiable missions.<\/p>\n<\/div>\n So we also, in addition to the rhythm of these six-to-eight-week cycles, operate by squads. The squads are mixed interdisciplinary subgroups that are focused on a specific mission, and they don\u2019t necessarily ladder up to the manager. They actually are run by a DRI, and the DRI is often an IC, and their job is\u2013<\/p>\n<\/div>\n That\u2019s \u201cdirectly responsible individual\u201d and \u201cindividual contributor.\u201d<\/strong><\/p>\n<\/div>\n Yeah, exactly. Thank you. And I think we\u2019ve taken the approach of separating the role of the manager from the role of the DRI that executes on a specific mission. I think that\u2019s because being a great DRI is exhausting. You\u2019re literally all-in 24 hours a day, and you\u2019re pushing as hard as you possibly can. Being a manager is often about being a coach, offering support, giving guidance, feedback, unblocking all sorts of things, helping with people\u2019s career growth. And so I think keeping those separate allows us to rotate DRIs every two or three cycles so that some people can try sort of different positions and have rotation. It\u2019s a great, very flexible structure that allows us to be pretty nimble, I think.<\/p>\n<\/div>\n Let\u2019s talk about Build. I wanted to start with superintelligence. You\u2019ve mentioned it several times now. I was just at Google IO. Demis Hassabis, who used to be your colleague when you were at Google, <\/strong>ended that keynote by saying<\/strong><\/a> that we were in \u201cthe foothills of the singularity, and that AGI was coming with all the power of Google.\u201c<\/strong><\/p>\n<\/div>\n You\u2019re saying superintelligence is here. Are these all the same things? Are we using different language to describe AGI? Are there differences? How would you define superintelligence in your context versus the singularity in Demis\u2019s?<\/strong><\/p>\n<\/div>\n I mean, obviously I didn\u2019t say it was here. I said it\u2019s coming. And I think there\u2019s a lot of fluidity around these phrases. But I think what we can clearly see that what\u2019s happening right now is that there is log-linear hill climbing across all modalities, and that means that there is a very direct relationship between each order of magnitude of compute that we apply, each incremental increase in data, and climbing on benchmarks, whether they\u2019re public benchmarks, internal benchmarks, they\u2019re targets that we focus on with reinforcement learning environments. And that is a very important observation.<\/p>\n<\/div>\n Those predictions that I think we\u2019re all making \u2014 I understand why some people are sort of skeptical of them or raise questions, but they\u2019re very grounded in the sort of empirical observations of over a decade of increase in performance of these models. I mean, essentially the same general-purpose architecture has seen 12 orders of magnitude more computation applied, a trillion-fold increase in FLOPS over 15 years, and basically has worked in audio, in image, in text, in code, and in many other time series prediction tasks. And so we\u2019re basically extrapolating out that more orders of magnitude of compute will enable us to continue to climb in this log-linear way inside of other environments.<\/p>\n<\/div>\n And then it raises the question of, are we going to be able to train models that can invent new knowledge, not just sort of extrapolate from existing data that we have, but actually teach us things that we don\u2019t know, and make new discoveries? Then the second thing is, do they have the capacity to self-improve and accelerate the process of deciding which hypotheses should be set, which ones should be pursued, how to generate training data for each of those, how to factor those into new runs, or even innovate on the actual architecture itself?<\/p>\n<\/div>\n So, I think both of those things need to be true to be able to see this compounding progress, but I think we\u2019re going to continue to get massive gains just from applying the next few orders of magnitude of compute. That probably does achieve parity with human performance on many, many more tasks, just as we\u2019ve seen that happen in the last six months on coding.<\/p>\n<\/div>\n Coding is really interesting, because it\u2019s easily validated, right? You write the code, you ask the computer to run it, it runs or fails. We\u2019ve seen some of the downsides, certainly around security, right? The downsides are obvious, and we\u2019re seeing that this sort of regulatory approach to coding security play out in lots of ways. I\u2019ve probably vibe coded some security disasters on my own phone and computer, and maybe that\u2019s a risk I\u2019m willing to take.<\/strong><\/p>\n<\/div>\n Every other function doesn\u2019t seem that easy. I always pick on law, because that\u2019s my background. But a judge doesn\u2019t validate legal writing the way a computer validates code. If you get it wrong, the judge can send you to jail, right? That is maybe the worst output validation error that you can probably run into.<\/strong><\/p>\n<\/div>\n How do you measure the effectiveness across domains as easily as you can measure the effectiveness in coding? Because this seems to me where the metaphor or the analogy from coding to other domains falls apart very quickly.<\/strong><\/p>\n<\/div>\n I\u2019m not so sure. Coding, obviously, you can verify the correct execution of code. It runs, or it crashes. But there\u2019s a ton of nuance in that. The quality of the code that gets written really matters: its extensibility, how reconfigurable it is, how useful it is in practice. It\u2019s not just that a piece of code runs, but it\u2019s also how a model actually uses it as a DevOps or an SRE in production to return to that piece of code that it\u2019s written, and then use it in a practical and useful way.<\/p>\n<\/div>\n And then, of course, you have to grade the quality of the output that has been produced. It may be high-quality, functioning code, but is it actually the app or the website that you wanted? And there are aesthetic judgments in that; there are commercial judgments in that. The challenge of internalizing non-verifiable rewards is present in code, even though code is still primarily a verifiable reward signal. I think the other thing to observe is that, like chat is also a non-verifiable space, and yet, we\u2019ve managed to climb that to basically human-level performance through interaction with real-world usage that provides a very strong-<\/p>\n<\/div>\n Wait. I\u2019m very curious. How do you measure chat at human-level performance?<\/strong><\/p>\n<\/div>\n Well, I think many people are having long, meaningful conversations with AIs at human-level performance. The quality is exceptionally good. It has very good emotional intelligence. It\u2019s broadly very accurate. We\u2019ve minimized the hallucinations. We don\u2019t talk so much about bias anymore. It\u2019s grounded in real-world observations. I think by most people\u2019s measures, we\u2019ve reached human-level performance in conversation for quite a wide range of tasks now.<\/p>\n<\/div>\n What are your measures, and actually, sure, most people\u2019s measures? I would disagree with almost all of this, but those are my measures. What are your measures?<\/strong><\/p>\n<\/div>\n My measure is like when I turn to my assistant and ask it to provide me with a daily briefing summarizing all the conversations that have happened on Teams and on email, the updates that have happened to documents, and I get basically a synthesized summary with a set of actions that I should take next. That is basically better than what my chief of staff can produce. I would say that\u2019s human-level performance in synthesis, analysis, proposed actions, and chat.<\/p>\n<\/div>\n There are many, many millions of people every day that are using it for emotional support, for counseling, for therapy, for coaching, for advice. I think it\u2019s one of the most popular use cases inside all of the chatbots. That\u2019s a pretty robust measure, I would say, to make the claim.<\/p>\n<\/div>\n I know you\u2019ve spent a lot of time thinking about this, particularly the emotional connection to some of these chatbots. These are products that you have built and deployed. I would draw a pretty big distinction between this thing is really, really good at summarizing my email, task list, and providing me a brief about what things to prioritize, and this thing is an emotional coach for somebody undergoing some kind of crisis. <\/strong><\/p>\n<\/div>\n Those are not similar tasks. Those are not necessarily similar kinds of intelligence, even in people. I know some people who are very good at making lists, and are very bad at emotional support. How do you put that all together in your brain and say, \u201cOkay, this is broadly human-level performance in chat?\u201d<\/strong><\/p>\n<\/div>\n I think if you define chat as an interactive exchange between two parties, one of which in this case is an AI, that broadly satisfies some goal, you\u2019re looking to learn the sports score, for advice on which restaurant to go to, for coaching and feedback on an essay that you\u2019ve written, for suggestions about which job to take next, or some tough conversation you\u2019re about to have with your manager. You get a response, you go back and forth, you have five or six exchanges, and you find that a useful output, which you might otherwise have to rely on an expert, friend, or even pay a coach.<\/p>\n<\/div>\n There are, just objectively, empirically speaking, hundreds of millions of people that get that experience every day from these chatbots. Maybe we could quibble over whether that technically represents human-level performance. I think it\u2019s a fairly reasonable thing to claim.<\/p>\n<\/div>\n There\u2019s no reason why that isn\u2019t going to continue climbing, right? The rate of climbing in the last three years is the thing that I think is most staggering. And so, what we\u2019re trying to do from this point is extrapolate: okay, what are the fundamental drivers of that climb \u2014 compute, data, interaction from real-world users \u2014 and those things look set to continue.<\/p>\n<\/div>\n I think that they apply to many other domains too, not just chat, emotional support, and productivity and that kind of thing, but also many other domains beyond that\/ Healthcare, live production deployments inside of education, assistants that are increasingly managing your home, looking at just everything that is in your everyday life basically to make you more productive. That is, I think, a trajectory that\u2019s likely to continue.<\/p>\n<\/div>\n You\u2019ve mentioned now that it\u2019s still the same fundamental architecture, transformers, and attention. We\u2019ve been applying compute to that for 15 years, and we\u2019re getting these big increases. You are in a fairly unique spot.<\/strong><\/p>\n<\/div>\n At Build, you announced your first flagship reasoning model, MAI-Thinking-1. You got to start from scratch. Is there anything you\u2019ve done differently now after 15 years of architecting and training this model, or is it just, yep, we\u2019re going to collect all the data and run the training just as we did, and we have more compute now, so it\u2019s going to be better?<\/strong><\/p>\n<\/div>\n No, actually, I think there are quite a lot of differences. The first thing to say is that the way that you curate the data\u2026 We start right from the top of the stack; we have basically paid for and acquired an extremely high-quality, very conservative set of data, and extracted a lot of the noisy, distracting, low-quality, potentially security-risk issues to do with that data. And the methods that you do for that, I think, are actually quite proprietary. We just shared a 109-page, very detailed, technical report, which was very well received on Twitter, and shares a lot of the details on how we do this. I think the second thing is, whilst I think it\u2019s important to be quite cautious with architectural choices, and we have been, there are also a number of pretty significant shifts that I think we\u2019ve made in how we put together our training runs.<\/p>\n<\/div>\n Our training runs have been incredibly stable, with very few crashes, and very few restarts. We shared a lot of those graphs to show infrastructure stability, and also MFU efficiency, so model FLOPS utilization, which basically shows that we can put a state-of-the-art number of FLOPS through each chip for every step in our training run. I think that this is extremely easy to get wrong, and we all hear lots of stories from different labs about how things do go wrong.<\/p>\n<\/div>\n It is actually pretty hard to make the very careful and deliberate choices to get things right, and take the right approach to make sure we produce high-quality models, because our job and our ambition is to try and build this hill-climbing machine. That means the integration of the silicon with the models, with the super high-quality data, with a stack of RLEs, reinforcement learning environments, that allow us to basically, systematically hill climb against any objective that we choose.<\/p>\n<\/div>\n And that\u2019s what MAI-Thinking-1 is. It\u2019s a general-purpose, fairly neutral, thinking model that is pretty good at coding. It\u2019s now roughly on par with Opus 4.6, at least on the benchmarks. We haven\u2019t deployed it at scale into production, so there\u2019s still lots more work to do there. But it\u2019s an extremely strong reasoner and scored 97 percent on AIME, which is the primary measure for its reasoning performance, at least on the benchmarks.<\/p>\n<\/div>\n It\u2019s very good at instruction following, and then the goal is basically to make that available to many, many developers and enterprises and allow them to climb on it for their use cases. Everybody has a sort of slightly different objective that they have in their company to try and build agents and so on that support their use case.<\/p>\n<\/div>\n One of the things that you\u2019ve noted in talking about MAI-Thinking-1 is that you didn\u2019t distill any existing models, which actually struck me as surprising, right? This is a thing you could do. You have access to OpenAI\u2019s IP. Everyone\u2019s distilling everything. We just found out in this trial that Grok was distilled from a number of models. Why not do distillation here? Why not jump ahead?<\/strong><\/p>\n<\/div>\n There\u2019s definitely lots of shortcuts to the frontier, and if you take a super high-quality model, and you polish your base model with high-quality instructions, or answers, or outputs from a superior model, then it\u2019s true that the model might quickly fit to that distribution. But it\u2019s very unclear that they would then be able to surpass that teacher.<\/p>\n<\/div>\n So, we\u2019ve been very deliberate for two reasons. The first is that we want to make sure that we can exceed the teacher in order to set the frontier ourselves over the next few years. And the second is that we really want to build one of the great labs, and it\u2019s going to take us many years to come, probably the next two or three years.<\/p>\n<\/div>\n But, in order to do that, we have to be able to show that we can actually build every component ourselves. We can hire the very best talent in the world. We can push the frontier with actual research, rather than just re-implementation, copying, or distillation from any other third party.<\/p>\n<\/div>\n We\u2019re in a great position where we\u2019re able to really carefully and meticulously pursue that objective, knowing that we have the resources to buy Anthropic models where they exceed the frontier. We have the resources to put 11,000 different models inside of Foundry, so every one of our developers gets pure optionality. And of course, we have the resources to continue to deploy OpenAI models, which are obviously outstanding and are at the frontier today.<\/p>\n<\/div>\n That\u2019s just a natural part of the self-sufficiency mission, and it\u2019ll take time for us to truly get to the absolute frontier on that. But I think we\u2019re in a great spot. We made a ton of progress. This is a very, very strong model, and it wasn\u2019t just that model that we released. We\u2019ve released seven new models simultaneously.<\/p>\n<\/div>\n Our transcribed model, for example, MAI-Transcribe-1.5 is literally the number one in the world. It\u2019s the most cost-effective of any of the hyperscalers. It\u2019s the highest on accuracy. Our image model is now number two. Our image editing model is number three right behind Google\u2019s and OpenAI\u2019s. I think we\u2019re well up there with our image and audio. Our code model, CodeFlash, is incredibly strong, optimized for VS Code. and is a really, really a great model that\u2019s on par with Sonnet 4.6. So it\u2019s really in a great spot this minute.<\/p>\n<\/div>\n Were there any legal or IP concerns with distillation? I know this is a live issue out in the world: Anthropic complains of other people distilling their models. There are concerns about Chinese companies distilling models, and whether our existing IP agreements can cover that. Did you have any of those concerns to keep you away from it?<\/strong><\/p>\n<\/div>\n Oh, we didn\u2019t, but I think I understand why a lot of people get frustrated. Anthropic has been very frustrated, and some of the rumors around xAI, and Meta, and obviously, the open source models, and so on, because essentially, that\u2019s basically taking the IP, and the knowledge that another team has put together, and then, literally force-feeding it into your own model. I think it\u2019s a bit of a short-term win, and like I said, really, we want to create a culture in the lab where we can come up with the next big thinking breakthrough, or the next big coding breakthrough, or the next big architectural push.<\/p>\n<\/div>\n Right now, we\u2019re experimenting with the looped transformer, which is a slightly different variant on the current transformer. Lots of people in the field are looking at it too. No one seems to have quite got into production yet. But, in order to create a culture and a team that can really push the frontier, they have to understand, own, and create the full stack as and when they need to, and also use things from third parties whenever we need to too. And like our paper, for example, has hundreds of citations grounded in the rest of the literature, so it\u2019s very much a contribution back to the field in return for everything that we\u2019ve learned over the years from all the great publications that have been out there.<\/p>\n<\/div>\n Can I ask you \u2014 if you understand that frustration from Anthropic and your peers in AI about distillation, do you also understand the frustration from creatives, publishers, and YouTubers about all the AI companies scraping their work as a collective to make these models? Because that frustration is only getting louder.<\/strong><\/p>\n<\/div>\n Yeah. No, I understand the frustration. The open web challenge is one we\u2019ve talked about before, and I get it, and I see that people are frustrated, and obviously, that\u2019s working its way through the conversation in the courts. And I see that people put things online, and they had different expectations about what the contract was with that being placed online, and it\u2019s a tricky one.<\/p>\n<\/div>\n You mentioned all your data was carefully curated. Did you pay for all the data that you\u2019re using to train the new models?<\/strong><\/p>\n<\/div>\n A lot of our data we obviously take from the open web in the normal way. Carefully curated means that it\u2019s extremely carefully filtered for security, for quality, for third-party dependencies from some of the open-source datasets, and keeping it away from a lot of the Chinese lineages, which I think are very different. Our enterprises want to make sure that when they put something into production, they can trust us that we\u2019ve really built it with their needs in mind. And I think this is one of the benefits of being very, very deliberate, patient, and being attentive to all the details.<\/p>\n<\/div>\n You mentioned enterprise. I think this is very interesting. Microsoft is all in on enterprise AI, in big ways, actually. I would even draw the line straight to Asha Sharma, the new head of Xbox, who is <\/strong>getting rid of AI in a bunch of places<\/strong><\/a>, and the gamers are happy, right? There\u2019s one reaction to AI in consumer space, but there\u2019s another in enterprise. I think AI has as close to product-market fit in enterprise as you can get with something changing as fast as AI. There are a bunch of databases that corporations control, and you can just go access them, because they control them. That\u2019s their data.<\/strong><\/p>\n<\/div>\n There\u2019s a bunch of repeatable processes and tasks, and old systems that maybe the models can just do more efficiently. There\u2019s something very important happening to enterprise. At the same time, <\/strong>the consumer antipathy towards AI<\/strong><\/a> is just increasing. And my argument is we have not built great consumer AI products. This industry has not produced them. It has not shifted them. It <\/strong>has not made it obvious that all of this is worth it<\/strong><\/a>, that using <\/strong>all the data from the open web<\/strong><\/a>, and changing the contract of publishing to a mass audience of people, so now, it\u2019s being used for training models that will deliver trillions of dollars of value to corporations. There isn\u2019t a product that says this is worth it. <\/strong><\/p>\n<\/div>\n Again, Satya Nadella recently gave an interview with Axios, and he said, \u201c<\/strong>We need social permission<\/strong><\/a> for this. And until we have it, until we deliver that value, people are going to feel this way.\u201d We\u2019ve seen college speakers get booed. We\u2019ve seen data centers get banned. Do you think that there\u2019s a consumer product that\u2019s worth it, that\u2019s worth the angst about training, that\u2019s worth the <\/strong>angst about<\/strong><\/a> data centers? <\/strong><\/p>\n<\/div>\n That was your focus; now your focus is enterprise. I would say that just on the face of it, it doesn\u2019t seem like Microsoft has interest in the consumer product anymore. But, do you see one that\u2019s worth it, or that could be built?<\/strong><\/p>\n<\/div>\n I\u2019m not sure I agree with you that there hasn\u2019t been any value for the consumer out of this. Across all of the chatbots, there are billions of people a month that are getting immense value out of it.<\/p>\n<\/div>\n Now, just for a moment, empathize a little bit with the small-scale business owner, or the kind of mom that\u2019s helping her kid with the homework, and can now just turn to a conversational AI, and get feedback, get instructions, get essay questions set. Just being able to ask questions like how do I generate revenue? How do I put together a cash flow forecast? Which college should I apply to?<\/p>\n<\/div>\n I mean, these are everyday tasks that are coming with some pretty high-quality factual advice and information. So I don\u2019t really buy that people are not getting benefit out of these things. I think they are.<\/p>\n<\/div>\n I think I can very clearly make the argument that they\u2019re not getting enough benefit, right?<\/strong><\/p>\n<\/div>\n They\u2019re the ones <\/strong>saying<\/strong><\/a> that we should not have more data centers. They are the ones <\/strong>booing AI<\/strong><\/a> at the graduation speeches. The <\/strong>polling<\/strong><\/a> is clear, <\/strong>particularly young people<\/strong><\/a>: the more they use AI, the more antipathy they have towards it. That\u2019s clear in every single poll. That\u2019s the argument I\u2019m making \u2014 not that there\u2019s no value, but the value exchange is not clear enough.<\/strong><\/p>\n<\/div>\n I\u2019m seeing Microsoft in particular pivot to enterprise, away from the big search product, the reinvention of Bing that would make Google dance. That\u2019s over, and we\u2019re all focused on enterprise, where the value is. I\u2019m just wondering if there\u2019s enough value for the consumer to make all of this worth it.<\/strong><\/p>\n<\/div>\n I think there\u2019s understandably a lot of anxiety. There\u2019s an enormous amount of speculation about what\u2019s going to happen in the next five to 10 years. Whether it\u2019s framed as the singularity or whether it\u2019s framed as the job apocalypse, these are not helpful framings. I think that people are scared because it\u2019s poorly defined and it\u2019s often framed as an inevitable, threatening gray cloud over people\u2019s heads.<\/p>\n<\/div>\n I think that what matters is what we do with technology.I think that I\u2019ve for a long time argued that we have to place the human first. Some people in the field have placed scientific discovery first or placed accelerating intelligences that can explore the galaxies and so on, and said that it\u2019s inevitable that we\u2019re going to have these AIs that are going to be more powerful than all of us combined. I mean, that\u2019s naturally scary to people.<\/p>\n<\/div>\n And I think that we have to basically flip it the other way around and say the purpose of science and technology is to make us all healthier and smarter and happier. That\u2019s been the quest that we\u2019ve been on as a species for thousands of years of invention, and it\u2019s the test that we should put superintelligence to again. And if it doesn\u2019t achieve that test, then I think people will reject it, and they\u2019ll be right to reject it.<\/p>\n<\/div>\n I think that everybody\u2019s focus is now going to turn in the next five years to, how is this making me healthier and happier, smarter, more capable, more productive? And if it\u2019s not doing that, then naturally people are going to be angry and resist and react. I don\u2019t think there is anything unexpected about that or anything wrong about that \u2014 I think that\u2019s inevitable.<\/p>\n<\/div>\n So that\u2019s why one of the things I\u2019ve been passionate about for many, many years is healthcare. And just a couple of days ago we announced a new partnership with Mayo Clinic. This is the number one hospital in the world, consistently reported. They have the highest quality longitudinal patient record dataset across all the modalities. They have the best clinical practice.<\/p>\n<\/div>\n They\u2019re also a nonprofit, which I think a lot of people don\u2019t realize, with 65 percent of their patient population on Medicaid. People often associate them with the international super elites flying in to get the best care in the world, but they actually have the majority on Medicaid. They\u2019re an amazing institution with an incredible mission to deliver the best healthcare everywhere. And we now have a very long-term partnership to co-train from scratch with their data, with our models, a brand new foundation model for health, deploy it in their hospitals, and hopefully take it around the world to deliver the best clinical care and healthcare that we possibly can to as many people as possible.<\/p>\n<\/div>\n That\u2019s why I got into the field. That\u2019s what I was originally motivated by, and it\u2019s what I\u2019m passionate about. And I can only focus on the things that I think are going to make a difference and that will help people and leave a good legacy for everybody, and that\u2019s what we\u2019re trying to do.<\/p>\n<\/div>\n I appreciate that. I appreciate the healthcare framing, and I understand why that\u2019s everyone\u2019s go-to, right? Healthcare in America in particular, if you could make it even 10 percent better, you will have affected a lot of people\u2019s lives in a particularly profound way.<\/strong><\/p>\n<\/div>\n The thing is, I know a very smart guy who has a very different and vastly more aggressive approach to all of this than you. That person is you, four months ago. This is what Mustafa Suleyman <\/strong>said to the <\/strong>Financial Times<\/em><\/strong><\/a> four months ago: \u201cWhite-collar work when you\u2019re sitting down at a computer, either being a lawyer or an accountant or a project manager, or a marketing person, most of those tasks will be fully automated by an AI within the next 12 to 18 months.\u201d<\/strong><\/p>\n<\/div>\n That\u2019s four months ago. That implies that a year from now, lawyers, accountants, project managers, and marketing people will not have jobs. Their jobs will be automated. Is that still your timeline?<\/strong><\/p>\n<\/div>\n