mirrorlake

LLM text and images have crept into so many places, and I've experienced quite a lot of existential dread this year from it. I've definitely felt what the author wrote about here.

Another trend, starting sometime last year, has been witnessing people reading scripts on camera: YouTube, public speeches, advertisements, and podcasts where a real human is reading from a script that is LLM-generated.

In the case of smaller video creators and podcasters, I then start wondering if they really think they can pretend that this is organic human speech? Or do they think I'm stupid and will continue to consume this slop without any critical thought..? Or do they think that their audience must know, and therefore consent to it, and so it's okay? In some cases, the person reading the LLM script is literally unaware that they're reading LLM text because a human scriptwriter handed them the page. I see this with advertising copy and a few YouTube channels.

Either way, it has led to quite a few instances where I got the LLM "ick", like I'm witnessing an LLM wearing a human skin, and it grosses me out so much.

As one final note: some of us assumed in ~2023 that LLMs would help to improve technical documentation. Thus far, 3 years after that prediction, I can say definitively that we've not ushered in an era of great documentation. We've got a steadily increasing percentage of documents that cause my eyes to glaze over because I don't know if a human expert ever once vetted it.

Looking forward to reading this, I am one of the people who preordered. I've often heard people with degrees say that they regret not being better at the material in this book, so for many peope this will be a chance to practice these skills in a way that their CS/math/physics/engineering degrees didn't really enable.

Also, Hillel (author) has a blog that's absolutely worth checking out as well as some great conference talks which are on YouTube, he's on the short list of speakers who I automatically watch any of their talks.

I always bind a side button on my gaming mice to ctrl+click so that my thumb can open links in a new tab. Slight ergonomic improvement for me.

Be sure to drink your Ovaltine!

I feel like I'm being rickrolled over and over again by infomercial-grade slop.

It's also TV in 1950.

Perhaps consider making some tutorials, then, and share your wealth of knowledge rather than calling people stupid.

It's a social-media-level of fact checking, that is to say, you feel something is right but have no clue if it actually is. If you had a better source for a fact, you'd quote that source rather than the LLM.

Just do the research, and you don't have to qualify it. "GPT said that Don Knuth said..." Just verify that Don said it, and report the real fact! And if something turns out to be too difficult to fact check, that's still valuable information.

Yeah, the Steam HW survey shows that 16:9 resolutions form a majority (60%+) of their users with 1080p + 4K, so it makes sense as a default design choice for a company that only wants to target one ratio.

As a former user of 16:10, I feel your pain, though.

I quite like the 2x3 grid of videos. No complaints, actually.

I saw profiles across different searches which had no contact info listed, seems like it isn't really designed to be a hiring tool.

I'm very skeptical of the claim that you'll be able to identify people by "usefulness of code", whatever that means.

"There are no shortcuts, you have to put in the work." Spoken like someone who doesn't use an SRS system, then. They're actually extremely hard to use, because the focus is on feeding you the toughest possible version of every recalled card. Part of why people quit using them is because it's mentally exhausting!

Anecdotally, English/History/Communications professors are confirming cheaters with them because they find it easy to identify false information. The red flags are so obvious that the checker tools are just a formality: student papers now have fake URLs and fake citations. Students will boldly submit college papers which have paragraphs about nonexistent characters, or make false claims about what characters did in a story.

The e-mail correspondence goes like this: "Hello Professor, I'd like to meet to discuss my failing grade. I didn't know that using ChatGPT was bad, can I have some points back or rewrite my essay?"

For longer beer drinking sessions, store in a double walled vacuum-sealed insulated bottle, of course.

I quite like the sociological definition for several reasons. Rather than trying to pin down precise criteria, you simple can ask people "Are you a mathematician? Are you a theoretical computer scientist?" And once someone has gone through that filter, everything that follows is opinion and also a historical snapshot of what people felt the field contained at the time. It provides future theoreticians and students a way to orient themselves on a map which is only partially drawn.

The definition of "theoretical physics" might have rapidly changed between 1900, 1920, 1940, and 1950--but certainly people who called themselves theoreticians remained mostly unchanged. Analyzing how everyone's definitions were changing gives a wealth of information about when and where breakthroughs were happening. 1919 and 1945 come to mind as such examples of when a theoretical field changed as a result of experiments [1][2].

Back to computing: Dijkstra told the story of attempting to put "Programmer" as his profession on his marriage certificate in 1957, and was rejected [3]. Clearly there are both pros and cons of using the sociological definition of a field. We all know programming existed before 1957, but the perception of it as a profession was so foreign that it wasn't allowed on an official document. It would've been impossible, apparently, where he lived to ponder about "What is programming?" if no one could BE a programmer. For that reason, we should probably be flexible and always be willing to discuss different definitions for every field so that we gain the benefits from multiple lines of reasoning.

[1] https://en.wikipedia.org/wiki/Eddington_experiment

[2] https://en.wikipedia.org/wiki/Trinity_(nuclear_test)

[3] https://amturing.acm.org/award_winners/dijkstra_1053701.cfm

I've seen math PhDs mess up addition and subtraction on a whiteboard, though.

Beating the 99th percentile human at any subject should not be difficult when the LLM training is equivalent to living thousands of lifetimes spent reading and nearly memorizing every book ever written on every university subject.

The fact that it only just barely beats humans feels hollow to me.

For those who've seen it, imagine if at end of Groundhog Day everyone in the crowd went, "Wow, he's slightly better than average at piano!"

Reminds me of this talk [0] led by CB Bailey, a top answerer on StackOverflow for the tag 'git' [1].

They create commits from scratch from the command line--manually creating each /.git/ file with shell commands and a text editor. Really fun talk. Would highly recommend it for people who were planning on learning about git internals at some point.

[0] "How does Git actually work? - CB Bailey & Andy Balaam [ACCU 2019]"

https://www.youtube.com/watch?v=N0m42TKk_dc

[1] https://stackoverflow.com/tags/git/topusers

In the event that this is added to the standard library, I'm going to be really curious to see what a "hello world" project/example would look like.

I went so far as to find the commit where David Chase added for loopvar on Mar 6, 2023 (github: golang/go/commit/c20d959) to try to design my own hello world with x/tools/cmd/bisect, but I'm out of my depth.

The hash tree is a great visualization. I wouldn't have grasped the importance of the hash suffix until I saw the tree. Awesome stuff.

When simulating or modeling how long something will take, (0 hours, 1 hours, 2 hours...) you never use a normal distribution because of the potential for negative values in the left tail. You would rule that out by default. The title itself implies that the author has this misconception.

> exactly the opposite of what we should expect out of programmers.

I mean this only jokingly: Who gets to decide what programmers are supposed to be like? You?

As someone who tries on a new language every year or two, I love using languages with formatters. Worrying about how many spaces to indent with or other arbitrary things is the opposite of what this particular programmer likes to worry about--I want to jump right in and solve problems with a language, and let the formatter make the code look like it came from a textbook.

I was reading over a student's code and he couldn't tell me his own variable names, couldn't even find one when I referred to it conceptually ("Where is the variable that holds your cache?".... no answer at all, not even a guess. Just, "I don't know.")

Having used GPT all year myself, I quit using it to generate new code for me for the most part. Back to StackOverflow, books, and of course reading boring documentation/manuals. I'm not closed off to the idea, just that I'm very worried it could atrophy certain skills.

When I was first learning Go last year, Penn mentioned on his podcast that he had a friend Rob who worked at Google.

I realized that he was talking about Rob Pike who helped create Go. Truly felt like a "small world" moment.

For existing homes, owners should pay a plumber to install a "drain water heat recovery unit", assuming the price of copper is reasonable. It's one of those low hanging fruits that will improve the energy efficiency of every home with minimal effort.

Likewise, insulation can be added to hot water pipes to increase efficiency even further.

The best MPG you can get on the highway is just the slowest you can safely go without confusing other drivers, making people angry, or getting pulled over by the police. Drag is proportional to velocity squared, so every increase in mi/hr wastes more energy pushing air out of the way of the car.

To answer your regenerative braking question: braking is inefficient by default, so just avoid tapping the brakes in situations where you could coast.

With regenerative braking, you'll never get back 100% the energy because of thermodynamics. There's always wasted energy any time energy is moved, transformed, stored, etc.

I think you'd do well doing a podcast interview with some financial independence folks, if that's in your comfort zone. Understandable if it's not.

Paula Pant gives a great interview, as does Brad Barrett at ChooseFI.

This site helps you run financial simulations, though. It's not supposed to be a product you use to track your account balances.

If you don't mind me asking, what did you to do reduce cost 90%?

A much easier solution: don't use their product. This would serve as a fantastic article about the reasons why it's worthwhile to stop giving them money. Except, it's the exact opposite of that--it teaches (or encourages?) everyone to learn the tricks required to ignore a company's unethical behavior, or install the extension to do it for you.

To get the more traditional results you have to go into filters and choose "video", but I'm sure they'll eventually remove the ability to search for only videos eventually, too (ugh).

It's the "news feed-ification" of search.

Should NASA's leadership be given to a monkey?

Debater A

While adaptability and resourcefulness are valuable traits, they are not unique to monkeys and can be found in human leaders as well. In addition, the consequences of a mistake made by a monkey leader in a critical situation could be catastrophic. We cannot risk such a gamble with the future of NASA and the safety of its astronauts.

Debater B

I understand your concern about the risks involved, but let's not underestimate the capability of monkeys. With proper screening and training, a capable monkey leader could emerge who could handle the immense responsibility of leading NASA. It's important to explore all options and not limit ourselves based on species.