The Future of AI and White Collar Work

So one definition of intelligence is sample efficiency. That is to say how much data do you need in a given domain to operate fluently and competently And it's actually not clear that we've made that much progress in training sample efficiency over the last few years It seems like more so, we've just dramatically widened and improved the data distribution The main way that AIs have been getting better is from adding more and better data and scaling the compute required to develop that data in the first place Obviously, RL is the main way that this has happened. You can think of RL as basically a kind of synthetic data generation where you dump a ton of compute against a verifier. or a rubric if you have another one as a judge. And you do this in to find out what the good data is in the first place And then you train your model to predict these correct rollouts much in the same way that you might train that model to predict the next word in internet text For this process to work, the model must have at least some prior probability to anticipate the correct solution in the first place, which is why you need mind stretching amounts of human expert trajectories in every single field and skill that you want the model to eventually be competent in. It's hard to overstate how task specific and bespoke this human expert data it is If you want some intuition, I recommend checking out the job descriptions on Mercore or Surg' websites There are listings for word specialists who will convert legacy documents into polished word files And legal experts will write realistic M andA diligences or securities filings and management consultants who will write up template market research And's not only did the data have to be so domain specific But there has to be so much of it Each skill corresponds to at least hundreds of human experts who are generating example completions, writing rubrics, and explaining their chain of thought There's a reason that the data industry that is producing these expert labels and the areL environments in which these meticulously catalog skills can congeal is earning billions a year in revenue, soon to be dec of billions. Now imagine if it took a couple decades worth of courses with hundreds of concurrent professors. and millions of practice tasks for you to learn how to plish a word file Even the task count difference here underersates the gap because models have to grind their far more numerous tasks Ease for her Whereas a human student might practice a textbook problem once or twice. With GRPO, these models are generating hundreds to thousands of rollouts per task and they need to solve the credit assignment problem The correct way to think about these models is not like a human who has learned all these different skills that you see these models displaying. It's more like a Frankenstein's monster, which has been built out of a billion graphs of carefully constructed examples all sewn together Epoch recently reported that open models lag state of the art frontier models by four months. I think the reason it is relatively easy for open source and previous Lagards to catch up to within months of the frontier is that data is the real driver of progress. And data can be easily distilled from public APIs, whereas hyperparameters and training tricks and architectural optimizations cannot And if the latter were driving most of the progress, then catching up would be far harder than we are observing it to be It is easy to forget how much data these models are trained on And how much more it is than what we humans see in our lietpes We see these AIs as a galaxy glittering with capabilities But at their center, invisible to the naked eye holding all the constellations together. is an unimaginably massive black hole of data. Just a couple of points of comparison to help drive home how big this difference is Here's one If a person sees and hears on average, let's say generously wo thousand words an hour then between the time they're born and the time they're an adult They'll see about two hundred million tokens Now by contrast, these frontier models are trained on somewhere between tens to hundreds of Trillions of tokens That is close to a million full difference Here's another point of comparison If you wanted to, you could learn toler operate any random humanoid or robot arm within hours. And if we could get AI to learn just as fast, robotics would be a decade trillion dollar industry, and you'd have an endless army of Unit three G Oes doing all kinds of useful work in the world But the reason we can't do this is that our AIs learn much less efficiently than we do Even with the millions of hours of demonstrations that we've collected, this is not enough to allow them to perform complex open ended tasks. And a final point of comparison, a teenager can learn to drive a car with about twenty hours of practice. And even if it include their sixteen years of growing up and understanding how the world works and building physical intuition There's still three to four orders of magnitude less data than WMo and Tesla are using to train their self driving car models. Now I want to deal with a couple of common responses and objections that people have to these kinds of comparisons One thing people will say and I think Ill probably said this when it came out of my podcast. is that For humans, many billions of years of evolution had to go into basically pre training us And so we're being unfair when we're comparing how little data we see within our lifetimes. to why these cold started LLMs were're just starting off with a totally random initialization have to learn from I think this is not the right way to think about it. Our genome is only three gigabytes big and only one to two percent of it is protein codating. And there's simply not enough space to store the parameters of this network that supposedly evolution has pre trained I think the closer analogy is more that evolution found the right hyperparameters and the right loss functions and that within our lifetime, we are still from scratch building up the connect home in our brain. That is to say the analogous thing to the weights and parameters of the neural network itself. And even if you granted this comparison and you said, yes, the hundreds of trillions of tokens that these models see to get pre trained is similar to just catching up to evolion That sort doesn't explain why any new marginal capability that you want to give these models takes so much data. So once you have been educated, again, you don't need a hundred different professors to teach you how to learn a new programming language But these AI's, even once they're pretty trained, still require enormous amounts of data to learn the next marginal skill and the next marginal skill after that Another objection to this kind of comparison is that we're not including multim modal data that we're seeing in our lifetimes. So to include all this sensor information that we see from birth to adulthood. That's probably tens to hundreds of billions of tokens of data And my response to this objection is simply that blind and deaf people who have been cut off from all this in information still have general intelligence. And that suggests to me that all these billions of sensory tokens are not really the thing that is making humans smart And in fact, deaf people who don't have the ability to hear any tokens, who just have to consume them via sign language and reading are probably ingesting far less than the two hundred million language tokens that we ballparked earlier which suggests that even the millionfold difference that we calculated earlier might be an understatement Okay, the third common objection people make is that we just haven't scaled it We have these scaling laws, they tell us that bigger models are more sample efficient The human brain we know is about a one hundred trillan synapses. And we have frontier models that are currently around five shllion parameters And so maybe we could just achieve human label sample efficiency if we made these models one to two ordderers of magnitude bigger. The reason this objection is off mark is actually quite interesting. So if you look at the way the scaling loss equations work, they tell you that the parameter and data terms or added to the loss independently. So suppose you have a model and you've trained it compute optimally And you say, I want to be sample efficient. I want to use as little data as possible. and I'll throw in as many parameters as is necessary to make that happen. Take the constants from the Chinilla scaling law paper. evenven if you increase the number of parameters by infinity, only decrease by a factor of ten the amount of data that you need in order to keep the same loss Humans are somewhere between thousands to millions of times more sample efficient than these models. So scaling the size of curnt models simply can't make up for that discrepancy. And this really does suggest that humans are in a different scaling curve altogether As soon as I earn money, I want to put it to work. But I also need to save for things like upcoming expenses and estimated taxes. So to figure out exactly how much I need to set aside, I ask command. Command is AI that is built into Mercury, which is my banaking platform And since I already use Mercury to run my entire business, Command has access to all the information it needs to get worked done. I just now command the date I'm interested in, and it does the rest. It takes my current balance and as whatever invoices will we do by the cut offff. Then it reviews my last six months of transaction history, so it can subtract out my monthly average expenses along with any scheduled payments. And if there's anything relevant coming up that's not in Mercury yet, I can just flag it Things like heads up, there's a twelve thousand dollars contractor payment that's slated for July. And that gets included in the final output. Because this is all happening in chat and every answer has links to the underlying data, I can easily double check command's work And once I'm convinced I can just tell command, all right, that looks good. Just transfer the surplus to my personal account. And you will immediately draft a transfer for me to approve. Command is live now. Visit mercury d. com slash command to learn more. Mercury is a Finte company, not an FAIC insured bank. banking serv provided through Choice Fancial Group and column A members FPIC. EI generated responses and suggested actions may vary and are not guaranteed Okay, all these nerdy comparisons aside. you might ask Why do we even care about sample efficiency? Is this actually necessary for the labs to achieve the two overarching objectives they have, which are one, automate white colorwk and two, automate AI research itself? The bet that the labs are making with white collar work is that the common tasks that our software engineer or analyst or accountant needs to do or comment And as a result, you can bring them into the training distribution Quite easily If you look at the revenue curves of these labs over the last few months It does suggest that there's an enormous amount of value from bringing into distribution these kinds common tasks, even if we can't replicate whatever is making human learning so special And it might be more inefficient to train AIs to do these kinds of tasks than it is to train humans. But so what The human lifespan simply does not allow for the quantity and the breadth of trading that these models experience If you, as a human, had some weird learning disability where you needed to read through every public repository on GitHub before you could be a competent software engineer And it would simply not make sense to train you up You'd be on social security By the early stages of your education, And even once you were trained onlyn be able to work on one project at a time AIs can learn these skills by fire hosing gigawatts of training at a time. And what they learn can be amortized across billions of sessions at once So we can be ludicrously inefficient in trading them up and still be wildly in the green And then there's a question of, well, how much outer distribution thinking do white collor employees need to do, that you simply cant train for in advance? This is more a question about the nature of different jobs than it is a question about AI research. And it also depends on which job you're talking about Some jobs are so mechanical and predictable that we were able to automate them long before the modern era ABI For example, bank tellers or travel agents But there are other jobs which require dealing on a daily basis with problems that are quite distant from the data distribution I think software engineering is probably one such. This is the job that A guys are supposed to take first But I would be willing to bet that there's overall more demand for human software engineers in twenty twenty seven than there is right now largely due to the complementary input of AI The lab's plans for this latter category of jobs is first to automate AI research and then have the automated AI researchers solve the sample efficiency problem. So then the question is, can AIs, which do not have human level sample efficiency Nonetheless solve the remaining research problems that stand on the way of human like intelligence and learning. This is a very complicated question and I'll have to address it in a much longer future blog post. Butre just to tease it a bit I think that the way that people currently think about an intelligence exlution is very clumsy because either people dismiss the possibility of AI speeding up AI progress altogher where they assume that some kind of go pops out the other end They don't reason carefully about what it looks like to have a period where AI progress is much faster than usual haveave that happen LLMs and the particular kinds of intelligenceces that LLMs are I'll save that for next time In the meanwhile, if you want to read this blog post or all the other blog posts I write or be alerted when I write a future blog post Go sign that for my newsletter at my website. arcash. com I'll see you later