lumb63

This is really cool! I have a lot of short projects that are essentially “parse out 2 or 3 tags of HTML and convert that to CSV. This will be perfect for that; in the past I’ve done it by hand with vim. Next time I’ll give this a shot.

Others have given a lot of good suggestions about books for general software engineering. Some book on how to structure code and some book on how to be a professional and some book on the human aspect of the profession would all be beneficial. Pick one that covers each of these areas and they’ll provide you a good foothold.

I’m surprised I don’t see more advice to read a language- or domain-specific book. If you know what type of technology you want to work on and what language(s) you want to use, I’d read a book or two to gain expert-level understanding of those languages or domains. I cannot tell you how beneficial it has been to me to read books covering all the intricacies of C, C++, and Rust. Not to mention countless pages of documentation covering the Linux kernel, shell scripting, etc. Even if you have great general software engineering skill, you still have to be able to quickly read, write, understand, and modify the code you work on. You never know when you might find yourself 50 files deep in an exploration of why some open source software behaves a certain way that’s really boiling your blood, or when you need to parse the code to find an undocumented feature you’re certain exists (or, if not, to hack it in). Knowing the language is very beneficial for this, and getting eyes on open source code to recognize common patterns is very useful.

It would’ve been nice if this post compared the costs of various energy sources per kWh, duration to bring capacity online, externalities related to energy sources (e.g. mass land use for solar, risk of plant failure for nucleae, etc.). That could’ve made it a very informative piece that might have convinced someone who was not already convinced. Instead it reads more as an ad hominem hit piece against nuclear proponents. We should be aiming to compare policies, not people.

> There’s no magic number at which the debt load becomes a full-on crisis.

When the US loses its AAA borrowing status and countless funds are required to sell all US Bonds, you will see that statement is incorrect.

The “right answer” also depends on attack vector. If you are trying to protect against nation-state-level tampering or data leaks for two full generations, the enemy moves much quicker, and so you will need to advance much more quickly, to outpace what you think they might do to outpace you in the future.

If you are only trying to prevent your ISP from seeing your traffic, which they are not trying particularly hard to do, then that level of protection would be overkill.

I know the “AI label” drives me away. It means the product is unreliable and a black box.

The “AI” label also indicates the solution is way over complicated and simpler ways to improve the product have been ignored. For instance, Confluence now has an “AI” chatbot. Search is still substantially worse than grep.

I worry about the same thing. I will say, however, that I worry more about this in the context of news articles, conversation, and other shorter-duration activities. Books are a big commitment, and it is highly disappointing to find out someone stretched a 50 page book into 250 pages to get it published. Sometimes it helps to instill the idea, or provides useful elaboration, but sometimes it is filler with no discernible benefit to the reader. Our time is important and as such, I think there ought to be a referendum against wasting it. That is not to say all content should be short, but that it should be no longer than it needs to be, and that information density should be as high as is tolerable.

Furthermore, there’s a replication crisis in social sciences. The last thing we need is to accumulate less data and let an LLM tell us the “right” answer.

The definition they use is so incredibly broad as to mean anything. If you think man-made climate change is real but doesn’t take a higher priority than the activities causing climate change, you’re a climate denier. If you think man-made climate change is real but aren’t sure that natural disasters are increasing in frequency or severity because of that, you’re a climate denier. Both of these types of people believe in man-made climate change yet are said to deny it.

Does that even make sense? This article is trying to drive a wedge between people and foment hate, as is anyone using the term “climate denier”, whether they know it or not. It’s not productive.

The readers of this article could be talking to people, learning their nuanced views, exchanging ideas, and finding common ground with the so-called “climate deniers”, but instead they’re reading a hit piece that makes them think Republicans are going to cause the literal end of the world.

As an American, boy, do I have news for you…

This conversation makes me really happy I have successfully avoided C++ for the past two years.

Does anyone have any tips for similar wizardry-level SIMD optimization on ARM?

Can someone explain why proving and math problem solving is not a far easier problem for computers? Why does it require any “artificial intelligence” at all?

For example, suppose a computer is asked to prove the sum of two even numbers is an even number. It could pull up its list of “things it knows about even numbers”, namely that an even number modulo 2 is 0. Assuming the first number is “a” and the second is “b”, then it knows a=2x and b=2y for some x and y. It then knows via the distributive property that the sum is 2(x+y), which satisfies the definition of an even number.

What am I missing that makes this problem so much harder than applying a finite and known set of axioms and manipulations?

This is something I’ve been trying to find a way to account for in my fantasy drafting the last few years, after trying to maximize expected value has not gotten me the desired results (probably also due in part to the delta between expected and actual value). It’s occurred to me that a “boom or bust” lineup that scores +2 st dev one week and -2 the next is way better than one that performs +1 every week (hand waving exact values).

Something I don’t see talked about enough is the extent to which the rise of passive investing increases the extent to which passive investing is the best strategy. Passive investing is predicated entirely on either freeloading off of the decisions of actively managed portfolios, who will adjust the market prices of securities efficiently, therefore resulting in passively managed funds automatically owning the “best” stocks because the bad ones will fall out of the indices; or it is predicated on what is effectively a Ponzi scheme wherein if everyone passively invests the same, then that form of passive investing will yield the best returns.

For analogy, consider school students trying to get all the questions right on the test. The first scenario is equivalent to one student studying hard to find the right answers. The rest of the class copies his answers and benefits from his efforts. The other scenario is no student trying hard, them all failing, but succeeding because the teacher curves the grades.

Note that both of these scenarios, especially the second scenario, results in stock prices which become increasingly detached from the economic reality of companies in proportion to the extent to which passive investing becomes prevalent.

For instance, assume stock XYZ is a bad investment but is in the S&P 500. If 90% of funds are actively managed, maybe they can sell a sufficiently large amount of it to push it out of the S&P 500 and save the passive investors from owning it. But if 90% of funds are passively managed, even if XYZ is hot garbage, the 10% of passively managed funds cannot possibly sell enough to make up for the fact that 90% of the market participants are indiscriminately buying a terrible stock.

Passive investing breaks the market to some extent. The free market system is predicated on having rational participants, not zombie participants.

I understand that isn’t what’s being suggested. What I’m suggesting is that there is perhaps a distortion of the common idea of who is “responsible” for something. I think the idea that fault bubbles up to the highest level in the chain of command is silly. Fault is distributed across the entire chain, and if we want to address this issue, we can’t ignore that.

To draw an analogy, if someone’s 16-year-old child is texting while driving and gets in a car accident, is their parent to blame? Most people could see that there is some fault on the part of both the parent (for perhaps not emphasizing enough the importance of safety while driving), and the child (for doing something they know is unsafe). And this fault exists in a continuum; maybe the parent told their child every day to not text while driving, and the child did it anyway. Maybe the parent never told them anything about safe driving habits, so the child had never considered that texting while driving was unsafe.

My point is that pretending that the highest C-suite executive is wholly responsible for everything that goes on in the company is extreme. Everyone along the entire chain of command has to do their part to ensure secure products are shipped - the executive needs to prioritize it, hire the right people to develop a plan, ensure people are enforcing the plan, etc., all the way down to the software engineers, the cleaning staff, etc. If one link in that chain breaks, the entire system fails, and it could be because of a weakness anywhere along the chain.

The people “whose negligence made this possible” are probably just rank-and-file employees. Careful what you wish for. I know I sure wouldn’t want to be legally liable if my software were vulnerable to something I didn’t know about.

Maybe a reasonable first step is third-party standards, audits, and certifications around data security to make privacy- and security-conscious consumers aware of what a company is doing. If consumers really find value in that, then they will preferentially deal with that company, and other companies will follow suit.

This is another consequence of the surveillance state. The same data that can be used to surveil us by the government can be stolen by who-knows-who. We’d all (mostly) be far better off, IMO, if companies didn’t retain such records.

There is, as far as I am aware, nothing to stop Microsoft from crawling any other site’s code. Please correct me if I’m wrong. Things like “copyright” didn’t seem to stop OpenAI.

It seems to me that regardless of the outcome of this case, some developers do not want to have their code used to train LLMs. There may need to be a new license created to restrict this usage of software. Or, maybe developers will simply stop contributing open source. In today’s day and age, where open source code serves as a tool to pad Microsoft’s pockets, I certainly will not publish any of my software open source, despite how much I would like to (under GPL) in order to help fellow developers.

If I were Microsoft, I’d really be concerned that I’m going to kill my golden goose by causing a large-scale exodus from GitHub or open source development more generally. Another idea I’ve considered is publishing boatloads of useless or incorrect code to poison their training data.

As I see it, people should be able to restrict how people use something that they gave them. If some people prefer that their code is not used to train LLMs, there should be a way to enforce that.

Put another way, I think one of the author’s arguments is “what if we’re pouring more and more effort into research in an area, because our primary abstraction is wrong?” That would make it harder and harder to continue to roll the ball up an increasingly steep hill. And I think in some fields, they may be right. It’s akin to how in math some problems are very difficult in Cartesian coordinates but trivial in polar coordinates. For instance, it seems to me that the existence of the “wave-particle duality” in physics and our inability to explain quantum-scale phenomena in a way that makes intuitive or logical sense to people is the canary in the coal mine that we took a wrong turn a while back. That’s not to say it’s true in every field, but that’s one that has stood out to me for a while.

I think another cause could be bureaucratic. From what I understand, several hundred years ago, research was funded by rich patrons who wanted some credit in relation to discoveries. Nowadays there is a giant, lurking, centrally planned grant machine that distributes money to researchers. And as we know from economics, central planning becomes increasingly untenable as the system becomes larger. Results get worse as the complexity skyrockets.

Additionally, if we are going to blame the “burden of knowledge” in any capacity, we have to acknowledge the abysmal education system (in the US). An anecdote which I will never forget is, in 6th grade we had a student who had recently immigrated from India. By his own telling, he was an average student there. But compared to native students, he was light years ahead in math. We learned multiplication in third grade, division in fourth grade, long division and fractions in fifth grade, or something like that. He had learned multiplication in first grade, and all of division in second grade. While our smartest students in sixth grade were grappling with pre algebra, he was bored in classes with the eighth grade algebra students. Our education system has, since experiencing that, seemed to me deeply flawed; there really is no reason an efficient and effective public education system should take 12+ years to be able to have a child ready for college and then another 4+ in college to have them ready to contribute to an isolated field.

I made a Rust program that, given a database of football players and projections about how they will do during the season, can be used during a fantasy football draft live to provide an indication of who the best player to draft is. I have other fantasy/sports gambling applications I’ve written to solve knapsack-type problems, choose the best team to pick in a “survivor” pool, and convert from various Vegas odds to win probabilities.

Outside of sports I have written code to download the entire Jeopardy archive to categorize most likely questions to appear to help study, and code to interact with Fitbit APIs to pull down data, code to analyze certain aspects of my full genome sequence.

I have unfinished code (and a desire to finish it) for creating algorithms to play certain board games, a tool to automatically request books and audiobooks from the library based on a Goodreads reading list (or other source of books), a tool to optimize Universal Paperclips, and several different business ideas I’d like to eventually pursue.

Creating something you find useful is a great way to get something done, and make a good tool. If you want it, chances are there are a lot of others who also want it.

To elaborate on your statement, we all think in very different ways. Recently there was an academic test posted here that evaluated “how” a person thinks (internal monologue, use of images, how memories are recalled, etc.). After my girlfriend and I both took the test and I saw how differently we both think, I was shocked. Had we not taken this quiz I’d have assumed the inside of her mind fundamentally worked the same way as mine does. But that is seemingly very far from the truth.

As where I can visualize, use internal monologue, vividly recall my memories, etc. at will, by default I do none of the above, and my thoughts are opaque to me. For her, she almost exclusively uses her internal monologue when thinking, and her entire thought process is consciously visible to her. It’s entirely conceivable that other people might not have an experience of “consciousness” resembling anything like what my idea of consciousness is.

A couple interesting points here. One is the supposition that something which is more “justified” is better. We humans do a lot of things which cannot be justified and are good, so this seems an odd position to me.

The other is that “something more free” is coupled with something “more just”, and that that is also better than something which is less justified. The implication here would be that adding more freedom is always more just than something “unjustified”. I don’t think this logically holds, either.

As someone who formerly identified as a libertarian, this is the logic that made me question the position. A philosophy which when empowered aims to minimize its sphere of influence is self-defeating, unfortunately.

I think and hope that the results of this ruling will be less extreme than the dissent warns. It could potentially be very bad, though. The word “official” is doing a lot of lifting in this ruling. If it is not interpreted too liberally, then perhaps not that much is “official” and thus immune. If it is interpreted the other way, then this could be bad.

If you were running the business, what would you do? Would you intentionally damage your odds of producing the best outcome for your company to support someone because of their race or gender?

I’ll go ahead and make an assumption that the Chinese government was involved. Countries badly need to figure out a way to punish bad actors in cybersecurity realms. It seems that this type of attack, along with many others, are quickly ramping up. If there isn’t competent policy in this area, it could become very dangerous.

This test really threw me for a loop because it didn’t cover my main mode of thought. For me, I can consciously use my internal voice or visualize things, but if I am not actively trying to, they never happen. Typically, the results of my thoughts simply enter my conscious mind, and what brought them about is opaque to me.