We could not possibly know that such a model is aligned. If that is all it takes to get a generation ahead of Astra, and we are willing to move this fast, these pauses in training are not going to end up meaning very much, time is even shorter than we knew
there exists a wisdom in the involuntary, and that our attempts to shape reality, however intentional, are themselves a form of unintentional expression.
But given the Chinese Communist Party’s power-seeking nature, it seems much more likely that China would agree to cooperate on AI safety if U.S. capabilities were comfortably ahead.
AIs don’t just repeat the same sentence patterns but also the same themes (memory is a favorite), names (Elara Voss, Marcus Chen), and underlying ideas.
In verifiable domains, model capability scaling should remain unbounded. Models will simply keep improving by "absorbing more and more of the computational universe", which is infinite by construction.
Auto instrumentation has gotten so good in recent years. If you're using Open Telemetry and everyone should be using Open Telemetry. All of the common patterns like all of the models are trained on them. So, it is literally faster and easier to build with instrumentation than to than not to.
And so basically everything that we can verify reasonably well with some feedback loop, the AIS are doing pretty well on. And that's sufficient to make AR and D go quite fast and to continue. But there's some parts of of developing uh aligned and safe AIs that are more subtle, hard to check, depend on, you know, detailed in the weeds things.
Structure and Interpretation of Computer Programs (SICP) by Gerald Jay Sussman and Hal Abelson. SICP is my all-time favourite programming book. It's also where I learned about the power of wishful thinking when coding. SICP uses wishful thinking as a design tool: write code as if the abstraction already existed. Pretend. Then, once the ideal abstraction has taken shape, you go ahead and implement those functions. This meta-level is what made SICP so valuable. Ultimately, it's a book that teaches how to think about code and problem solving. I'm tempted to even say that SICP is more a work of art than a pure coding book, but I won't go there. A beautiful book.
Smalltalk Best Practice Patterns. True, Kent Beck is better known for his later work, which is excellent too. But Smalltalk Best Practice Patterns is particularly strong on coding style. I learned a lot by just reading the code examples. Small tweaks to names and abstractions add up. No one captures that better than Kent.
So this speaks of a very important fact that um AI learns from patterns. When the patterns are not abundant, then we have to be careful. We have to know how to use AI or how not to use AI.
However, looking ahead, I still do not believe that future AI (say, in 15 years) will be based on the LLM stack. I believe it will necessarily have to move closer to its optimal, final form -- symbolic learning. Obviously this is a risky and contrarian belief -- the safe bet would be LRMs. But let's see.
However, there's an increasing body of literature that shows that by engaging older adults in novel cognitive activities, you see improvements in function. So that can be language learning. Uh that can be uh complex like coordinative movement or exercise. You see the same things with um like musical training like people learning a new musical instrument or learning um musical theory trying to identify different patterns in music. And you see in randomized control trials improvements particularly in executive function that seems to be most common across those different interventions.
Some people think CI means you reject grammar books. I disagree. I often reference grammar materials and find these explanations useful. However, the focus of my learning is not to “master the rules”. Grammar resources simply help me notice and make sense of patterns in the language.
there's a different kind of intuition that you that you develop over years as a software engineer and uh there's many categories of it but the one I'll I'll call attention to that is like a thing that you cannot teach you cannot do you cannot learn in a textbook. The only way to learn it is like I know bad patterns in software because I have debugged them at three in the morning.
In simple, crystalline prose, Clements shares a lifetime of wisdom on using money to buy happiness, counteracting the behavioral quirks that bedevil our financial decisions, and making family and friends an integral part of our wealth.
Know where the money is coming from. Know where you plug into that. And like if you're just asking those questions, you're actually already like steps ahead of all of the other like fledgling perspective mathematicians.
And then I was like, okay, can you come up with an algorithm that is better than the algorithms that I came up with or that anybody else came up with and go ahead and like look at all the published work and synthesize that and then try to come up with something novel and it it's not able to do it. And I can give it a lot of time and it's it's still not able to do it.
It shouldn't It shouldn't be shelved next to How to Win Friends and Influence People because it's a manual not of how to gain power, but of how to keep power.
We're now doing it in a much heavier way because we find that these like kind of bory enterprisey uh patterns end up being pretty useful because you have a bunch of idiots on your team now. The coding agents are a bunch of idiots and they are going to work 24/7 and they're going to like ship a lot of stuff. So you need way more guardrails than you used to.
The thing that I find interesting is that's not novel. This has been the thing we've always been trying to do forever. How do we get a junior engineer to ship code safely without breaking stuff? Right? How do we make patterns in the codebase? How do we make tests? Like it's it's all the old stuff that we've always wanted to do.
It may still be the case, and I guess it is, that immune traits are the most selected category, but is not at all the case, and in fact, we can prove it's not the case that behavioral traits are not selected.
That's why I don't think accelerationist steamrolling will work. The only way this conflict resolves peacefully is some kind of grand bargain: a flagship policy plan that directly addresses the public's top fears about AI and meaningfully redistributes the gains.
For years, it was faster to mock up software than to ship it. Designers stayed "ahead" of engineering with prototypes. Now AI coding agents make development so much faster that the loop has flipped.
one of the magic tricks about these things is that they they're incredibly consistent. If you've got a code base with a bunch of patterns in, they will follow those patterns almost to a T.
I actually think profitability happens when you underestimated the amount of demand you were going to get and loss happens when you overestimated the amount of demand you were going to get because you're buying the data centers ahead of time.
Write notes in your own words, connect them to existing notes, and let ideas emerge from the connections. I wrote a whole blog post about this, so go ahead and read it. Highly recommended.
You know, interestingly, LM themselves are quite bad at playing chess. Like, they hallucinate moves. They look at patterns, right? They're they're very good at pattern recognition, but not so good at going super super super deep on a specific chess thing.
Decades of grading data; standardized test scores; cross-sectional, longitudinal, observational, and experimental studies; along with many other types of ancillary and convergent evidence, ultimately tell the same story: education can raise the absolute performance of most students modestly, but it almost never meaningfully reshuffles the relative distribution of ability and achievement
And this is why I say that history doesn’t repeat itself. Of course it doesn’t. It can’t possibly repeat itself because we’re always living in a constantly evolving time. But patterns of human behaviour do. And what you always get after economic crisis is political upheaval, always, always, always.
You Need a Budget (YNAB) is a simple to use (but sophisticated under the hood) bit of budgeting software that uses the principles of human habits and behavioral finance to make you more conscious and efficient in your spending.
I mean, I've been doing this for over 15 years. If there were any shortcuts, I'd be all over them. There are patterns and there are frameworks, but there's no shortcuts.
An eye-opening view on considerations going into building a widely used public API or reusable library. While the book focuses on the .NET framework, many of the conventions apply to maintainable and reusable components, in general. This book had an outsized impact on me as I read it when I was a mid-level .NET developer.
I do not believe that analytical skills are the missing ingredient in thinking about the future. Rather, I believe that imagination about the future and a theory of change that helps to describe it clearly, are what is needed to look ahead in a more compelling way.
This fits with some of the studies of chess experts and so forth that it’s not so much that you learn the patterns passively. You learn what to look for. You learn what’s important and what’s not.
So Tesla hasn’t found a different, better way to bring driverless technology to market. Waymo is just so far ahead that it’s dealing with challenges Tesla hasn’t started thinking about.
Installed generators cost ~$800/kW, and the data center capital cost, including servers, is ~$40,000/kW, so adding 1% more makeup compute capacity is money ahead of purchasing generators.
I expect that the delta between 5 and 4 will be the same as between 4 and 3 and I think it is our job to live a few years in the future and remember that the tools we have now are going to kind of suck looking backwards at them and that’s how we make sure the future is better.
Take the free Attachment Style Quiz and get personalised insight into how you connect, where your patterns come from, and your next step toward secure love.
Obviousness comes from conforming to people’s existing mental models. Don’t waste time reinventing common UI patterns or paradigms unless they are at least 2x better, or you have some critical brand reason to do so.
What we could we could train people by saying, "Here's the recognition prime decision model. Now, you know, follow this." And that would be useless. Because of uh telling people the strategy isn't going to buy them anything. The intuition part is a reflection of the patterns they've built up through experience. And so there's no shortcut for that.
But their lives became miserable. It would take them 45 minutes to decide what restaurant to go through. They got divorced. They lost their jobs because they could not use their emotions because our emotions are a way of drawing on on the patterns we've built and on our experiences.
The Sprint 225 uses one TriplePower LED bulb. Its light pattern is wonderfully smooth, with no distracting rings; it seems almost perfectly optimized for night-running and night-hiking — it focuses most of the light ahead, but still manages to illuminate the periphery.
I register all of my domain names through Hover. I think it's a little more expensive than some of the other options, but the UI is simple and reliable and not loaded with dark patterns like some of the cheaper competitors.
Rory Sutherland, Alchemy: The Dark Art and Curious Science of Creating Magic in Brands, Business, and Life: Rory Sutherland is the Vice Chairman of advertising giant Ogilvy; he describes his “attractively vague job title” as allowing him the freedom to create a behavioral science practice within the ad agency. This looks intriguing.
In some cases, this has led to once-held "truths" about how we create and manage accounts to be totally flipped on their head, yet we still see modern organisations applying the patterns of yesterday to the threats of today.
Rough Notes. This book identifies common patterns among people struggling to stay focused, and offers some coping strategies. Haven’t found it as inspiring as I’d hoped, but has some good stuff.
But to date, the integration of behavioral science into public policy has proceeded from developing a set of tools to then searching for problems these tools can help solve.
Behavioral science can play an even more important role in the policymaking process in coming years if practitioners instead begin with some of the large-scale questions that economic policymakers face and then develop insights that, often as a complement to more traditional policy tools, can help solve them.
A korrent is a belief a person has stated in their own words: one
sentence stating the claim, backed by a quote and a source, kept at
korrents.com.
Under a name here, the quoted block is what they actually said.
The korrent beneath it is the claim those words support, in
korrents' wording — tap it to see the record, its source, and who
else holds it.
Nobody here wrote their own korrents. They are compiled from public
statements, and a person can change their mind, which is recorded too.
About the English under a post
Some people here publish in a language other than English. Where they
do, this site shows a machine translation beneath the post, in
this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the
serif above is exactly what the person published, and it is what to
quote them on. A translation can be wrong in ways that matter,
especially about tone.
Only the post's own words are translated. A quoted post, a linked
article and a belief on korrents.com
are left in their original language.