More from the singularity is nearer
AMD is sending us the two MI300X boxes we asked for. They are in the mail. It took a bit, but AMD passed my cultural test. I now believe they aren’t going to shoot themselves in the foot on software, and if that’s true, there’s absolutely no reason they should be worth 1/16th of NVIDIA. CUDA isn’t really the moat people think it is, it was just an early ecosystem. tiny corp has a fully sovereign AMD stack, and soon we’ll port it to the MI300X. You won’t even have to use tinygrad proper, tinygrad has a torch frontend now. Either NVIDIA is super overvalued or AMD is undervalued. If the petaflop gets commoditized (tiny corp’s mission), the current situation doesn’t make any sense. The hardware is similar, AMD even got the double throughput Tensor Cores on RDNA4 (NVIDIA artificially halves this on their cards, soon they won’t be able to). I’m betting on AMD being undervalued, and that the demand for AI has barely started. With good software, the MI300X should outperform the H100. In for a quarter million. Long term. It can always dip short term, but check back in 5 years.
This is a map of primary trading partners, US vs China, and how it has evolved over the last 20 years. Think about it, and realize this probably reflects your experience. I know there was a similar panic about Japan in the 80s, but Japan by population has always been 3x smaller than the US, whereas China is 3x larger. In addition, we had and have military bases in Japan. This is not the same situation. The US, since I have been born, has been coasting. The main product made by the US is the dollar, and it used those manufactured dollars to outsource everything. Most jobs in the US are now basically fake. It’s basically an economy in which five people stick a pipe in the ground, but that pipe is the fed and the oil was the good will built up over 1870-1970. In 2008, with the bailouts, it was made clear that the US has no interest in reform. The next decade, in perhaps a spitting in your face move, the fed made the interest rate 0. Known as ZIRP, this had never been done before. This led to insane perversions. When I got into business, I didn’t understand that business in America was mostly a total scam. Sure, you might look at a single business, and be like, oh, that sounds reasonable, but then you zoom out and look at the entire system, and it doesn’t really make sense. It’s scams feeding other scams. Wanna each start a business, pass dollars back and forth over and over again, and drive both our revenues super high? Sure, we don’t produce anything, but we have companies with high revenues and we can raise money based on those revenues. We’ll both be rich! Let’s do it with a bunch of extra steps so people don’t catch on though. They’ll only see it reflected in the lack of movement of real macro metrics. You see, the US is a “developed” country, which means real growth is over? You do understand that guns and boats are made of steel, right? Oh, airplanes aren’t, they are made of aluminum. Oh…right, yea, it’s not just steel it is absolutely everything. The future is chips you say? All the good chips are made in the Republic of China you say? This 2021 article lays it out clearly, and it also explains why nothing I saw in Silicon Valley made any sense. I’m not going to go into the personal stories, but I just had an underlying assumption that the goal was growth and value production. It isn’t. It’s self licking ice cream cone scams, and any growth or value is incidental to that. It isn’t until you understand this that people’s behavior starts to make sense. America really is at a fork in the road. In one world, they abandon all hopes of being an empire, becoming a regional power with highly protectionist economics. This happened before, and it’s called Europe. I know it’s hard to believe now, but Europe used to be the seat of power for the whole world. The sun never set on the British empire. Now they put you in jail for memes. Protectionist America is a boring place and not somewhere I want to be. It kicks the can further down the road of poverty, basically embraces socialism, is stagnant, is stale, is a museum…etc, again there’s a contemporary example of this. When I said on Lex they were gonna nationalize NVIDIA, look at the AI Diffusion Framework, and notice how Trump hasn’t repealed it. It allows export of GPUs to only 18 countries. Nationalization with American characteristics. It tells the other 177 countries that they should plan on purchasing their AI infrastructure from China. The other path, which is the exciting path, is the attempt to maintain an empire. An empire has to compete on its merits. There’s two simple steps to restore American greatness: 1) Brain drain the world. Work visas for every person who can produce more than they consume. I’m talking doubling the US population, bringing in all the factory workers, farmers, miners, engineers, literally anyone who produces value. Can we raise the average IQ of America to be higher than China? 2) Back the dollar by gold (not socially constructed crypto), and bring major crackdowns to finance to tie it to real world value. Trading is not a job. Passive income is not a thing. Instead, go produce something real and exchange it for gold. The first will bring the value of “American” labor in line with its global market value. It is a particularly unique advantage of the US over China, the US has a potentially much larger pool of talent. Non ironically, diversity is our strength. Unfortunately, there’s a lot of resistance to American labor finding its market value. The second will prevent a lot of the scams. The reason the banking industry is so big is that it is close to the source of the made up dollars. If currency is gold backed, you could imagine something similar happening to the mining industry instead. However, the mining industry is real! It uses steel and aluminum to build physical things. And imagine when we start to mine space. That’s a way better reward function than scamming politicians out of fake dollars. Unfortunately, I doubt either will happen. They very much both can, but people haven’t been demoralized enough yet.
A lot of smooth brains on Hacker News about the last post. I’m sorry if you spent your whole life worshipping money, but hey, the Bible warned you about false idols, don’t shoot the messenger. “It’s easier to imagine the end of the world than the end of capitalism” – Mark Fisher It’s actually very easy to imagine the end of capitalism. Imagine capitalism as a game of sharks, where eventually the biggest shark ends up gobbling up all the fish, and that one shark is the last player left standing with all the money. When one person (or company) has all the money, do you see how the money would be worthless? I’ll spell this out clearly. Money is a map, it is not a territory. Please understand what I mean by this before continuing to read. You can erase the mountains from the map, but you still have to climb over them in real life, and even worse, now you don’t have a map! “Everything around you that you call ‘life’ was made up by people who were no smarter than you” – Steve Jobs So, if money is the map, what territory is it attempting to capture? Presumably something having to do with value, but increasingly, as we are buying and selling baskets of derivatives of memecoins, nothing. A map that doesn’t accurately capture a territory is not a Schelling point. It’s not a useful map. And maps are only as good as their usefulness. Useless maps die out. Do you agree or disagree that money is supposed to be a map of value? If you disagree, that’s an ought and I can’t use logic to convince you otherwise, I can just call you a moron who refuses to burn paper $100 bills for warmth on a deserted island. Many capitalists I meet are as stupid as communists, trying to give a moral justification for their system. This is my money, I deserve it. I should be able to passively deploy my capital into the markets and live off the returns. “Moral victories are for minor league coaches.” – JAY-Z A economic system is only good in so much as it effectively deploys capital for real growth. If real economic growth is only 3 percent, any time you are earning beyond that, somebody else is losing. And yet somehow, today, you can put your money in money market accounts and earn a “risk-free” 5 percent…hmm something doesn’t make sense. Who is losing? You will eventually be unable to squeeze the productive people any further. The worst was an e-mail I got with someone who supposedly agreed with me. “Value creation (for all stakeholders) is at the core of the organization/ business model I am putting together…Anyway I wanted to let you know others out there who share your vision.” – anon email Fuck your stakeholders. Fuck your business model. You don’t understand me at all. Stop worrying so much about the distribution of the pie. Start thinking about how to make the pie bigger. With exponential (what 3 percent year over year is) growth, the latter outstrips the former by so much. The right distribution is simply: From each according to his ability, to each according to his ability to effectively deploy capital to achieve real economic growth. Communism is dumb cause it goes to the poor (who routinely demonstrate that they poorly deploy capital). Capitalism is dumb cause it goes to the rent-seekers (who frequently deploy capital to increase their moat). Acceleration is the way.
Intellectual property is a really dumb idea. “But piracy is theft. Clean and simple. It’s smash and grab. It ain’t no different than smashing a window at Tiffany’s and grabbing merchandise.” - Joe Biden, 46th president of the USA Except it isn’t and Joe Biden is a senile moron. Because when you smash the windows and grab the stuff, Tiffany’s no longer has the stuff. With piracy, everyone has the stuff. It’s a lot more like taking a picture, which Tiffany’s probably encourages. Win-win cooperation. Wealth is being increasingly concentrated. What’s shocking to me is how much everyone still cares about money. Even the die-hard complain about capitalism type deeply cares, because the opposite of love isn’t hate, it’s indifference. I hate scammers, but I’m pretty indifferent to money. The best outcome of AI is if it delivers huge amounts of value to society but no profit to anyone. The old days of the Internet were this goldmine. The Internet delivered huge value but no profit, and that’s why it was good. Suddenly we had all these new powers. Then people figured out how to monetize it. It was a race to extract every tiny bit of value, and now we have today’s Internet. Can this play out differently with AI? Let’s build technology and open source software that market breaks everything. Let’s demoralize the scammers so hard that they don’t even try. Every loser and grifter will be gone from technology because there’s nothing to be gained there. They can play golf all day or something. If I ever figure out how to channel power like Elon, I will do this. Spin up open source projects in every sector to eliminate all the capturable value. This is what I’m trying to do with comma.ai and tinygrad. I dream of a day when company valuations halve when I create a GitHub repo. Someday.
Pulled up at a stop light Imagine flying an x-wing down a corridor, having to turn the plane sideways to fit, a missile on your tail and closing, hitting the turbo, feeling the g force, coming up on the end of the corridor, pulling back hard on the stick the second the corridor opens, turning 90 degrees and watching the missile continue straight. Tingles. Adrenaline. Release. Or if you don’t want sci-fi, imagine winter circa 1645 in America. Several of your group almost dead from lack of food, tracking a deer, spotting it, shooting it with your bow, hitting but the deer is trying to run, fast twitch muscles charging and leaping, plunging a knife into its heart and knowing at that moment everyone is going to be okay. Heart rate calming laying on the warm deer. The modern world doesn’t have any real experiences like this any more. Survival has become a technocratic plod, making the right boring and careful decisions. There’s only fake experiences like the above, video games, sports, and drugs. And things like reckless driving, which are just kind of stupid. As we march toward ASI, this will only get worse. What the unabomber describes as Type 2 experiences, ones where you can achieve results with serious effort, will vanish. All that will be left are things you can have for no effort (like food) and things you can never have (like world peace). Even when humanity goes to Mars, we will be going as cargo. I was told recently I’m not engaged in my life, and it’s pretty true. Until I see a solution to this problem, even a sketch of a solution, what’s the point? Why sprint if you aren’t sure where you are going? I’m trying my best with comma and tiny corp, how do you make technology itself more accessible, not a fucking packaged product like when the default world talks about making technology more accessible. That’s just hiding complexity. But it’s so hard. Companies don’t work like how I thought they did, they just…exist. Which I guess in retrospect is obvious, there’s no adults in the room. Knowing the future doesn’t help you change it. I am continually shocked at how little people understand about anything, they don’t even understand that they don’t understand. Am I the same way? I try extremely hard to constantly test myself, if my predictions are wrong it’s clear I don’t understand. If I can’t build it I don’t understand. I’ll frequently read comments saying I don’t understand, but when I engage with these people they can’t explain what my world model gets wrong. A different meta world model? Or are they just idiots? To anyone who wants to supersede rationality, you better understand how to steelman every rationality argument. If I want to succeed, I believe I have to change who I am, and I’m not sure if that’s possible. I believe I’ve been making efforts in that direction, but I haven’t seen results yet. Working on AI is both the only thing that matters and also so demoralizing because of the above. I believe you have to give individuals control over the technology. And not by setting permissions in AWS that can be revoked, I mean in a nature sense. The ghost gunner is the real second amendment. This ideology holds me back so much in business, to the point I struggle to be competitive. But if you abandon that ideology, what’s the point to doing it at all? I have to win with a hand tied behind my back. We only make products for spiritual tops, not the majority of the world which is spiritual bottoms. Now, perhaps I have an ace in the hole. With the rise of AI, the spiritual bottoms will soon have no cash, because AI is the ultimate spiritual bottom. It would take a highly skilled terrorist to build spiritual top AI and even I’m not that crazy. So we’ll only have bottom AI, and it will outcompete all the human bottoms. Advertising will vanish once the hypnodrones have been released. Those humans will likely wirehead themselves out of the picture. This is the world I’m building for. Have you ever unconstrained your mind and thought about where the world is going? This won’t be like the steam engine replacing the horse, because all horses were bottoms. Ever seen a horse riding a human? Humanity bifurcates. Humans will retain control for the foreseeable future, the only question is, how many humans? If it’s 10, I’m out. If it’s 10k, 50/50 I’m in. If it’s 10M, I’m definitely in. My goal is to make this number as large as possible, it’s my best chance of survival. Give control of the technology to as many people as possible in a deep nature sense, not a permissions sense. In my opinion, this is what Elon gets wrong. Of course, he’s likely to be one of the 10, so maybe that’s why he doesn’t care. But what if he isn’t? Tesla and SpaceX are huge silos begging to be co-opted. I don’t think building silos like this is a good idea, compare the fate of the Telegram founder to the Signal founder. Build technology and structures that are inseparable from the narrative you want, as opposed to ones you think you can wield for good. On a long enough timeline, it will always end up in your enemy’s hands. Imagine if the only thing they could do with it furthers your goals.
More in programming
This is a re-publishing of a blog post I originally wrote for work, but wanted on my own blog as well. AI is everywhere, and its impressive claims are leading to rapid adoption. At this stage, I’d qualify it as charismatic technology—something that under-delivers on what it promises, but promises so much that the industry still leverages it because we believe it will eventually deliver on these claims. This is a known pattern. In this post, I’ll use the example of automation deployments to go over known patterns and risks in order to provide you with a list of questions to ask about potential AI solutions. I’ll first cover a short list of base assumptions, and then borrow from scholars of cognitive systems engineering and resilience engineering to list said criteria. At the core of it is the idea that when we say we want humans in the loop, it really matters where in the loop they are. My base assumptions The first thing I’m going to say is that we currently do not have Artificial General Intelligence (AGI). I don’t care whether we have it in 2 years or 40 years or never; if I’m looking to deploy a tool (or an agent) that is supposed to do stuff to my production environments, it has to be able to do it now. I am not looking to be impressed, I am looking to make my life and the system better. Another mechanism I want you to keep in mind is something called the context gap. In a nutshell, any model or automation is constructed from a narrow definition of a controlled environment, which can expand as it gains autonomy, but remains limited. By comparison, people in a system start from a broad situation and narrow definitions down and add constraints to make problem-solving tractable. One side starts from a narrow context, and one starts from a wide one—so in practice, with humans and machines, you end up seeing a type of teamwork where one constantly updates the other: The optimal solution of a model is not an optimal solution of a problem unless the model is a perfect representation of the problem, which it never is. — Ackoff (1979, p. 97) Because of that mindset, I will disregard all arguments of “it’s coming soon” and “it’s getting better real fast” and instead frame what current LLM solutions are shaped like: tools and automation. As it turns out, there are lots of studies about ergonomics, tool design, collaborative design, where semi-autonomous components fit into sociotechnical systems, and how they tend to fail. Additionally, I’ll borrow from the framing used by people who study joint cognitive systems: rather than looking only at the abilities of what a single person or tool can do, we’re going to look at the overall performance of the joint system. This is important because if you have a tool that is built to be operated like an autonomous agent, you can get weird results in your integration. You’re essentially building an interface for the wrong kind of component—like using a joystick to ride a bicycle. This lens will assist us in establishing general criteria about where the problems will likely be without having to test for every single one and evaluate them on benchmarks against each other. Questions you'll want to ask The following list of questions is meant to act as reminders—abstracting away all the theory from research papers you’d need to read—to let you think through some of the important stuff your teams should track, whether they are engineers using code generation, SREs using AIOps, or managers and execs making the call to adopt new tooling. Are you better even after the tool is taken away? An interesting warning comes from studying how LLMs function as learning aides. The researchers found that people who trained using LLMs tended to fail tests more when the LLMs were taken away compared to people who never studied with them, except if the prompts were specifically (and successfully) designed to help people learn. Likewise, it’s been known for decades that when automation handles standard challenges, the operators expected to take over when they reach their limits end up worse off and generally require more training to keep the overall system performant. While people can feel like they’re getting better and more productive with tool assistance, it doesn’t necessarily follow that they are learning or improving. Over time, there’s a serious risk that your overall system’s performance will be limited to what the automation can do—because without proper design, people keeping the automation in check will gradually lose the skills they had developed prior. Are you augmenting the person or the computer? Traditionally successful tools tend to work on the principle that they improve the physical or mental abilities of their operator: search tools let you go through more data than you could on your own and shift demands to external memory, a bicycle more effectively transmits force for locomotion, a blind spot alert on your car can extend your ability to pay attention to your surroundings, and so on. Automation that augments users therefore tends to be easier to direct, and sort of extends the person’s abilities, rather than acting based on preset goals and framing. Automation that augments a machine tends to broaden the device’s scope and control by leveraging some known effects of their environment and successfully hiding them away. For software folks, an autoscaling controller is a good example of the latter. Neither is fundamentally better nor worse than the other—but you should figure out what kind of automation you’re getting, because they fail differently. Augmenting the user implies that they can tackle a broader variety of challenges effectively. Augmenting the computers tends to mean that when the component reaches its limits, the challenges are worse for the operator. Is it turning you into a monitor rather than helping build an understanding? If your job is to look at the tool go and then say whether it was doing a good or bad job (and maybe take over if it does a bad job), you’re going to have problems. It has long been known that people adapt to their tools, and automation can create complacency. Self-driving cars that generally self-drive themselves well but still require a monitor are not effectively monitored. Instead, having AI that supports people or adds perspectives to the work an operator is already doing tends to yield better long-term results than patterns where the human learns to mostly delegate and focus elsewhere. (As a side note, this is why I tend to dislike incident summarizers. Don’t make it so people stop trying to piece together what happened! Instead, I prefer seeing tools that look at your summaries to remind you of items you may have forgotten, or that look for linguistic cues that point to biases or reductive points of view.) Does it pigeonhole what you can look at? When evaluating a tool, you should ask questions about where the automation lands: Does it let you look at the world more effectively? Does it tell you where to look in the world? Does it force you to look somewhere specific? Does it tell you to do something specific? Does it force you to do something? This is a bit of a hybrid between “Does it extend you?” and “Is it turning you into a monitor?” The five questions above let you figure that out. As the tool becomes a source of assertions or constraints (rather than a source of information and options), the operator becomes someone who interacts with the world from inside the tool rather than someone who interacts with the world with the tool’s help. The tool stops being a tool and becomes a representation of the whole system, which means whatever limitations and internal constraints it has are then transmitted to your users. Is it a built-in distraction? People tend to do multiple tasks over many contexts. Some automated systems are built with alarms or alerts that require stealing someone’s focus, and unless they truly are the most critical thing their users could give attention to, they are going to be an annoyance that can lower the effectiveness of the overall system. What perspectives does it bake in? Tools tend to embody a given perspective. For example, AIOps tools that are built to find a root cause will likely carry the conceptual framework behind root causes in their design. More subtly, these perspectives are sometimes hidden in the type of data you get: if your AIOps agent can only see alerts, your telemetry data, and maybe your code, it will rarely be a source of suggestions on how to improve your workflows because that isn’t part of its world. In roles that are inherently about pulling context from many disconnected sources, how on earth is automation going to make the right decisions? And moreover, who’s accountable for when it makes a poor decision on incomplete data? Surely not the buyer who installed it! This is also one of the many ways in which automation can reinforce biases—not just based on what is in its training data, but also based on its own structure and what inputs were considered most important at design time. The tool can itself become a keyhole through which your conclusions are guided. Is it going to become a hero? A common trope in incident response is heroes—the few people who know everything inside and out, and who end up being necessary bottlenecks to all emergencies. They can’t go away for vacation, they’re too busy to train others, they develop blind spots that nobody can fix, and they can’t be replaced. To avoid this, you have to maintain a continuous awareness of who knows what, and crosstrain each other to always have enough redundancy. If you have a team of multiple engineers and you add AI to it, having it do all of the tasks of a specific kind means it becomes a de facto hero to your team. If that’s okay, be aware that any outages or dysfunction in the AI agent would likely have no practical workaround. You will essentially have offshored part of your ops. Do you need it to be perfect? What a thing promises to be is never what it is—otherwise AWS would be enough, and Kubernetes would be enough, and JIRA would be enough, and the software would work fine with no one needing to fix things. That just doesn’t happen. Ever. Even if it’s really, really good, it’s gonna have outages and surprises, and it’ll mess up here and there, no matter what it is. We aren’t building an omnipotent computer god, we’re building imperfect software. You’ll want to seriously consider whether the tradeoffs you’d make in terms of quality and cost are worth it, and this is going to be a case-by-case basis. Just be careful not to fix the problem by adding a human in the loop that acts as a monitor! Is it doing the whole job or a fraction of it? We don’t notice major parts of our own jobs because they feel natural. A classic pattern here is one of AIs getting better at diagnosing patients, except the benchmarks are usually run on a patient chart where most of the relevant observations have already been made by someone else. Similarly, we often see AI pass a test with flying colors while it still can’t be productive at the job the test represents. People in general have adopted a model of cognition based on information processing that’s very similar to how computers work (get data in, think, output stuff, rinse and repeat), but for decades, there have been multiple disciplines that looked harder at situated work and cognition, moving past that model. Key patterns of cognition are not just in the mind, but are also embedded in the environment and in the interactions we have with each other. Be wary of acquiring a solution that solves what you think the problem is rather than what it actually is. We routinely show we don’t accurately know the latter. What if we have more than one? You probably know how straightforward it can be to write a toy project on your own, with full control of every refactor. You probably also know how this stops being true as your team grows. As it stands today, a lot of AI agents are built within a snapshot of the current world: one or few AI tools added to teams that are mostly made up of people. By analogy, this would be like everyone selling you a computer assuming it were the first and only electronic device inside your household. Problems arise when you go beyond these assumptions: maybe AI that writes code has to go through a code review process, but what if that code review is done by another unrelated AI agent? What happens when you get to operations and common mode failures impact components from various teams that all have agents empowered to go fix things to the best of their ability with the available data? Are they going to clash with people, or even with each other? Humans also have that ability and tend to solve it via processes and procedures, explicit coordination, announcing what they’ll do before they do it, and calling upon each other when they need help. Will multiple agents require something equivalent, and if so, do you have it in place? How do they cope with limited context? Some changes that cause issues might be safe to roll back, some not (maybe they include database migrations, maybe it is better to be down than corrupting data), and some may contain changes that rolling back wouldn’t fix (maybe the workload is controlled by one or more feature flags). Knowing what to do in these situations can sometimes be understood from code or release notes, but some situations can require different workflows involving broader parts of the organization. A risk of automation without context is that if you have situations where waiting or doing little is the best option, then you’ll need to either have automation that requires input to act, or a set of actions to quickly disable multiple types of automation as fast as possible. Many of these may exist at the same time, and it becomes the operators’ jobs to not only maintain their own context, but also maintain a mental model of the context each of these pieces of automation has access to. The fancier your agents, the fancier your operators’ understanding and abilities must be to properly orchestrate them. The more surprising your landscape is, the harder it can become to manage with semi-autonomous elements roaming around. After an outage or incident, who does the learning and who does the fixing? One way to track accountability in a system is to figure out who ends up having to learn lessons and change how things are done. It’s not always the same people or teams, and generally, learning will happen whether you want it or not. This is more of a rhetorical question right now, because I expect that in most cases, when things go wrong, whoever is expected to monitor the AI tool is going to have to steer it in a better direction and fix it (if they can); if it can’t be fixed, then the expectation will be that the automation, as a tool, will be used more judiciously in the future. In a nutshell, if the expectation is that your engineers are going to be doing the learning and tweaking, your AI isn’t an independent agent—it’s a tool that cosplays as an independent agent. Do what you will—just be mindful All in all, none of the above questions flat out say you should not use AI, nor where exactly in the loop you should put people. The key point is that you should ask that question and be aware that just adding whatever to your system is not going to substitute workers away. It will, instead, transform work and create new patterns and weaknesses. Some of these patterns are known and well-studied. We don’t have to go rushing to rediscover them all through failures as if we were the first to ever automate something. If AI ever gets so good and so smart that it’s better than all your engineers, it won’t make a difference whether you adopt it only once it’s good. In the meanwhile, these things do matter and have real impacts, so please design your systems responsibly. If you’re interested to know more about the theoretical elements underpinning this post, the following references—on top of whatever was already linked in the text—might be of interest: Books: Joint Cognitive Systems: Foundations of Cognitive Systems Engineering by Erik Hollnagel Joint Cognitive Systems: Patterns in Cognitive Systems Engineering by David D. Woods Cognition in the Wild by Edwin Hutchins Behind Human Error by David D. Woods, Sydney Dekker, Richard Cook, Leila Johannesen, Nadine Sarter Papers: Ironies of Automation by Lisanne Bainbridge The French-Speaking Ergonomists’ Approach to Work Activity by Daniellou How in the World Did We Ever Get into That Mode? Mode Error and Awareness in Supervisory Control by Nadine Sarter Can We Ever Escape from Data Overload? A Cognitive Systems Diagnosis by David D. Woods Ten Challenges for Making Automation a “Team Player” in Joint Human-Agent Activity by Gary Klein and David D. Woods MABA-MABA or Abracadabra? Progress on Human–Automation Co-ordination by Sidney Dekker Managing the Hidden Costs of Coordination by Laura Maguire Designing for Expertise by David D. Woods The Impact of Generative AI on Critical Thinking by Lee et al.
AMD is sending us the two MI300X boxes we asked for. They are in the mail. It took a bit, but AMD passed my cultural test. I now believe they aren’t going to shoot themselves in the foot on software, and if that’s true, there’s absolutely no reason they should be worth 1/16th of NVIDIA. CUDA isn’t really the moat people think it is, it was just an early ecosystem. tiny corp has a fully sovereign AMD stack, and soon we’ll port it to the MI300X. You won’t even have to use tinygrad proper, tinygrad has a torch frontend now. Either NVIDIA is super overvalued or AMD is undervalued. If the petaflop gets commoditized (tiny corp’s mission), the current situation doesn’t make any sense. The hardware is similar, AMD even got the double throughput Tensor Cores on RDNA4 (NVIDIA artificially halves this on their cards, soon they won’t be able to). I’m betting on AMD being undervalued, and that the demand for AI has barely started. With good software, the MI300X should outperform the H100. In for a quarter million. Long term. It can always dip short term, but check back in 5 years.
Earlier this weekGuileWhippet But now I do! Today’s note is about how we can support untagged allocations of a few different kinds in Whippet’s .mostly-marking collector Why bother supporting untagged allocations at all? Well, if I had my way, I wouldn’t; I would just slog through Guile and fix all uses to be tagged. There are only a finite number of use sites and I could get to them all in a month or so. The problem comes for uses of from outside itself, in C extensions and embedding programs. These users are loathe to adapt to any kind of change, and garbage-collection-related changes are the worst. So, somehow, we need to support these users if we are not to break the Guile community.scm_gc_malloclibguile The problem with , though, is that it is missing an expression of intent, notably as regards tagging. You can use it to allocate an object that has a tag and thus can be traced precisely, or you can use it to allocate, well, anything else. I think we will have to add an API for the tagged case and assume that anything that goes through is requesting an untagged, conservatively-scanned block of memory. Similarly for : you could be allocating a tagged object that happens to not contain pointers, or you could be allocating an untagged array of whatever. A new API is needed there too for pointerless untagged allocations.scm_gc_mallocscm_gc_mallocscm_gc_malloc_pointerless Recall that the mostly-marking collector can be built in a number of different ways: it can support conservative and/or precise roots, it can trace the heap precisely or conservatively, it can be generational or not, and the collector can use multiple threads during pauses or not. Consider a basic configuration with precise roots. You can make tagged pointerless allocations just fine: the trace function for that tag is just trivial. You would like to extend the collector with the ability to make pointerless allocations, for raw data. How to do this?untagged Consider first that when the collector goes to trace an object, it can’t use bits inside the object to discriminate between the tagged and untagged cases. Fortunately though . Of those 8 bits, 3 are used for the mark (five different states, allowing for future concurrent tracing), two for the , one to indicate whether the object is pinned or not, and one to indicate the end of the object, so that we can determine object bounds just by scanning the metadata byte array. That leaves 1 bit, and we can use it to indicate untagged pointerless allocations. Hooray!the main space of the mostly-marking collector has one metadata byte for each 16 bytes of payloadprecise field-logging write barrier However there is a wrinkle: when Whippet decides the it should evacuate an object, it tracks the evacuation state in the object itself; the embedder has to provide an implementation of a , allowing the collector to detect whether an object is forwarded or not, to claim an object for forwarding, to commit a forwarding pointer, and so on. We can’t do that for raw data, because all bit states belong to the object, not the collector or the embedder. So, we have to set the “pinned” bit on the object, indicating that these objects can’t move.little state machine We could in theory manage the forwarding state in the metadata byte, but we don’t have the bits to do that currently; maybe some day. For now, untagged pointerless allocations are pinned. You might also want to support untagged allocations that contain pointers to other GC-managed objects. In this case you would want these untagged allocations to be scanned conservatively. We can do this, but if we do, it will pin all objects. Thing is, conservative stack roots is a kind of a sweet spot in language run-time design. You get to avoid constraining your compiler, you avoid a class of bugs related to rooting, but you can still support compaction of the heap. How is this, you ask? Well, consider that you can move any object for which we can precisely enumerate the incoming references. This is trivially the case for precise roots and precise tracing. For conservative roots, we don’t know whether a given edge is really an object reference or not, so we have to conservatively avoid moving those objects. But once you are done tracing conservative edges, any live object that hasn’t yet been traced is fair game for evacuation, because none of its predecessors have yet been visited. But once you add conservatively-traced objects back into the mix, you don’t know when you are done tracing conservative edges; you could always discover another conservatively-traced object later in the trace, so you have to pin everything. The good news, though, is that we have gained an easier migration path. I can now shove Whippet into Guile and get it running even before I have removed untagged allocations. Once I have done so, I will be able to allow for compaction / evacuation; things only get better from here. Also as a side benefit, the mostly-marking collector’s heap-conservative configurations are now faster, because we have metadata attached to objects which allows tracing to skip known-pointerless objects. This regains an optimization that BDW has long had via its , used in Guile since time out of mind.GC_malloc_atomic With support for untagged allocations, I think I am finally ready to start getting Whippet into Guile itself. Happy hacking, and see you on the other side! inside and outside on intent on data on slop fin
I’ve been working on a project where I need to plot points on a map. I don’t need an interactive or dynamic visualisation – just a static map with coloured dots for each coordinate. I’ve created maps on the web using Leaflet.js, which load map data from OpenStreetMap (OSM) and support zooming and panning – but for this project, I want a standalone image rather than something I embed in a web page. I want to put in coordinates, and get a PNG image back. This feels like it should be straightforward. There are lots of Python libraries for data visualisation, but it’s not an area I’ve ever explored in detail. I don’t know how to use these libraries, and despite trying I couldn’t work out how to accomplish this seemingly simple task. I made several attempts with libraries like matplotlib and plotly, but I felt like I was fighting the tools. Rather than persist, I wrote my own solution with “lower level” tools. The key was a page on the OpenStreetMap wiki explaining how to convert lat/lon coordinates into the pixel system used by OSM tiles. In particular, it allowed me to break the process into two steps: Get a “base map” image that covers the entire world Convert lat/lon coordinates into xy coordinates that can be overlaid on this image Let’s go through those steps. Get a “base map” image that covers the entire world Let’s talk about how OpenStreetMap works, and in particular their image tiles. If you start at the most zoomed-out level, OSM represents the entire world with a single 256×256 pixel square. This is the Web Mercator projection, and you don’t get much detail – just a rough outline of the world. We can zoom in, and this tile splits into four new tiles of the same size. There are twice as many pixels along each edge, and each tile has more detail. Notice that country boundaries are visible now, but we can’t see any names yet. We can zoom in even further, and each of these tiles split again. There still aren’t any text labels, but the map is getting more detailed and we can see small features that weren’t visible before. You get the idea – we could keep zooming, and we’d get more and more tiles, each with more detail. This tile system means you can get detailed information for a specific area, without loading the entire world. For example, if I’m looking at street information in Britain, I only need the detailed tiles for that part of the world. I don’t need the detailed tiles for Bolivia at the same time. OpenStreetMap will only give you 256×256 pixels at a time, but we can download every tile and stitch them together, one-by-one. Here’s a Python script that enumerates all the tiles at a particular zoom level, downloads them, and uses the Pillow library to combine them into a single large image: #!/usr/bin/env python3 """ Download all the map tiles for a particular zoom level from OpenStreetMap, and stitch them into a single image. """ import io import itertools import httpx from PIL import Image zoom_level = 2 width = 256 * 2**zoom_level height = 256 * (2**zoom_level) im = Image.new("RGB", (width, height)) for x, y in itertools.product(range(2**zoom_level), range(2**zoom_level)): resp = httpx.get(f"https://tile.openstreetmap.org/{zoom_level}/{x}/{y}.png", timeout=50) resp.raise_for_status() im_buffer = Image.open(io.BytesIO(resp.content)) im.paste(im_buffer, (x * 256, y * 256)) out_path = f"map_{zoom_level}.png" im.save(out_path) print(out_path) The higher the zoom level, the more tiles you need to download, and the larger the final image will be. I ran this script up to zoom level 6, and this is the data involved: Zoom level Number of tiles Pixels File size 0 1 256×256 17.1 kB 1 4 512×512 56.3 kB 2 16 1024×1024 155.2 kB 3 64 2048×2048 506.4 kB 4 256 4096×4096 2.7 MB 5 1,024 8192×8192 13.9 MB 6 4,096 16384×16384 46.1 MB I can just about open that zoom level 6 image on my computer, but it’s struggling. I didn’t try opening zoom level 7 – that includes 16,384 tiles, and I’d probably run out of memory. For most static images, zoom level 3 or 4 should be sufficient – I ended up a base map from zoom level 4 for my project. It takes a minute or so to download all the tiles from OpenStreetMap, but you only need to request it once, and then you have a static image you can use again and again. This is a particularly good approach if you want to draw a lot of maps. OpenStreetMap is provided for free, and we want to be a respectful user of the service. Downloading all the map tiles once is more efficient than making repeated requests for the same data. Overlay lat/lon coordinates on this base map Now we have an image with a map of the whole world, we need to overlay our lat/lon coordinates as points on this map. I found instructions on the OpenStreetMap wiki which explain how to convert GPS coordinates into a position on the unit square, which we can in turn add to our map. They outline a straightforward algorithm, which I implemented in Python: import math def convert_gps_coordinates_to_unit_xy( *, latitude: float, longitude: float ) -> tuple[float, float]: """ Convert GPS coordinates to positions on the unit square, which can be plotted on a Web Mercator projection of the world. This expects the coordinates to be specified in **degrees**. The result will be (x, y) coordinates: - x will fall in the range (0, 1). x=0 is the left (180° west) edge of the map. x=1 is the right (180° east) edge of the map. x=0.5 is the middle, the prime meridian. - y will fall in the range (0, 1). y=0 is the top (north) edge of the map, at 85.0511 °N. y=1 is the bottom (south) edge of the map, at 85.0511 °S. y=0.5 is the middle, the equator. """ # This is based on instructions from the OpenStreetMap Wiki: # https://wiki.openstreetmap.org/wiki/Slippy_map_tilenames#Example:_Convert_a_GPS_coordinate_to_a_pixel_position_in_a_Web_Mercator_tile # (Retrieved 16 January 2025) # Convert the coordinate to the Web Mercator projection # (https://epsg.io/3857) # # x = longitude # y = arsinh(tan(latitude)) # x_webm = longitude y_webm = math.asinh(math.tan(math.radians(latitude))) # Transform the projected point onto the unit square # # x = 0.5 + x / 360 # y = 0.5 - y / 2π # x_unit = 0.5 + x_webm / 360 y_unit = 0.5 - y_webm / (2 * math.pi) return x_unit, y_unit Their documentation includes a worked example using the coordinates of the Hachiko Statue. We can run our code, and check we get the same results: >>> convert_gps_coordinates_to_unit_xy(latitude=35.6590699, longitude=139.7006793) (0.8880574425, 0.39385379958274735) Most users of OpenStreetMap tiles will use these unit positions to select the tiles they need, and then dowload those images – but we can also position these points directly on the global map. I wrote some more Pillow code that converts GPS coordinates to these unit positions, scales those unit positions to the size of the entire map, then draws a coloured circle at each point on the map. Here’s the code: from PIL import Image, ImageDraw gps_coordinates = [ # Hachiko Memorial Statue in Tokyo {"latitude": 35.6590699, "longitude": 139.7006793}, # Greyfriars Bobby in Edinburgh {"latitude": 55.9469224, "longitude": -3.1913043}, # Fido Statue in Tuscany {"latitude": 43.955101, "longitude": 11.388186}, ] im = Image.open("base_map.png") draw = ImageDraw.Draw(im) for coord in gps_coordinates: x, y = convert_gps_coordinates_to_unit_xy(**coord) radius = 32 draw.ellipse( [ x * im.width - radius, y * im.height - radius, x * im.width + radius, y * im.height + radius, ], fill="red", ) im.save("map_with_dots.png") and here’s the map it produces: The nice thing about writing this code in Pillow is that it’s a library I already know how to use, and so I can customise it if I need to. I can change the shape and colour of the points, or crop to specific regions, or add text to the image. I’m sure more sophisticated data visualisation libraries can do all this, and more – but I wouldn’t know how. The downside is that if I need more advanced features, I’ll have to write them myself. I’m okay with that – trading sophistication for simplicity. I didn’t need to learn a complex visualization library – I was able to write code I can read and understand. In a world full of AI-generating code, writing something I know I understand feels more important than ever. [If the formatting of this post looks odd in your feed reader, visit the original article]
This website has a new section: blogroll.opml! A blogroll is a list of blogs - a lightweight way of people recommending other people’s writing on the indieweb. What it includes The blogs that I included are just sampled from my many RSS subscriptions that I keep in my Feedbin reader. I’m subscribed to about 200 RSS feeds, the majority of which are dead or only publish once a year. I like that about blogs, that there’s no expectation of getting a post out every single day, like there is in more algorithmically-driven media. If someone who I interacted with on the internet years ago decides to restart their writing, that’s great! There’s no reason to prune all the quiet feeds. The picks are oriented toward what I’m into: niches, blogs that have a loose topic but don’t try to be general-interest, people with distinctive writing. If you import all of the feeds into your RSS reader, you’ll probably end up unsubscribing from some of them because some of the experimental electric guitar design or bonsai news is not what you’re into. Seems fine, or you’ll discover a new interest! How it works Ruben Schade figured out a brilliant way to show blogrolls and I copied him. Check out his post on styling OPML and RSS with XSLT to XHTML for how it works. My only additions to that scheme were making the blogroll page blend into the rest of the website by using an include tag with Jekyll to add the basic site skeleton, and adding a link with the download attribute to provide a simple way to download the OPML file. Oddly, if you try to save the OPML page using Save as… in Firefox, Firefox will save the transformed output via the XSLT, rather than the raw source code. XSLT is such an odd and rare part of the web ecosystem, I had to use it.