username223

There's no way I could say what percentage of my schooling was "actual learning" versus assessment or "busywork." The modern obsession with standardized testing and constant measurement has certainly made things worse.

I disagree that "ability to grade said learning is never used again." Stack-ranking is very much a thing; the grading just gets fuzzier.

What the article points to is that, when a teacher gives an assignment meant to encourage students to think and learn, e.g. picking themes out of a novel, most students completely miss the point, instead getting an LLM to generate words in the shape of a student essay. They're crippling their future selves to save some time.

What would I do if I were a teacher? My impulse would be to make all assignments ungraded or pass/fail. Students could choose to learn or cheat as much as they liked. But then their grades would depend upon a midterm and a final, either oral or written in-person. For the ones who had been cheating their way along, the midterm would hopefully be a wake-up call, and they could redeem themselves on the final.

> Importantly, the recording should indicate whether it was human or AI generated.

This is all that's necessary. Sometimes I'm fine with mediocre TTS; sometimes I want an actual professional; librivox is somewhere in between, but should clearly specify whether I will be getting an amateur human or a robot.

Yep. It won't be long before the web is flooded with pages full of AI-generated content repeatedly mentioning brands near keywords, and "search engines" have been replaced by "summaries" monetized by prompt-stuffing. That's pretty much the extent of these people's genius.

EDIT: I guess the final step is for an "AI agent" to enter your credit card number based on this bot-chat.

> To: FDA Office of Drug Safety > > URGENT SAFETY ALERT—EVIDENCE OF CLINICAL TRIAL FRAUD

I don't think overwhelming public officials with alarmist machine-generated spam is helpful to anyone.

EDIT: The "benchmark" doesn't even seem to contain any negative examples. What a joke.

Unfortunately the calling conventions and memory models are all different, so there's usually hell to pay going between languages. Perl passes arguments on a stack, Lisp often uses tagged integers, Fortran stores matrices in the other order, ... it goes on and on. SWIG (https://swig.org) can help a lot, but it's still a pain.

Hah!

Ruby, Python, and Perl all had similarly good package ecosystems in the late 1990s, and I think any of them could have ended up as the dominant scripting language. Then Google chose Python as its main scripting language, invested hundreds of millions of dollars, and here we are. It's not as suitable as Matlab, R, or Julia for numerical work, but money made it good enough.

(Sort of like how Java and later JavaScript VMs came to dominate: you can always compensate for poor upfront design with enough after-the-fact money.)

That's a big ask. I've written a couple of books, and PDF/print vs. ePub/webpage is a decision I had to make up front. For anything with more than the most basic formatting (e.g. a novel), the design and layout are different enough that one of them will turn out looking like crap. Given the audience, I probably would have chosen ePub for this one, but the book's done now.

Also, a TOC and sample chapter would be great. A lot more people will be likely to pay if they know what they're getting.

100%

I've climbed some mountains, and while I don't know if I would be capable of doing Everest, I do know that I wouldn't want to (unless I could do it like Goran Kropp: https://en.wikipedia.org/wiki/Göran_Kropp). I guess I could brag if I doped myself, took a helicopter past the Khumbu Icefall, and had servants carrying my stuff the whole way to the summit, but I could never be proud of myself for doing that.

> Anything that combines those three capabilities will leave you open to attacks, and the attacks don't even need to be particularly sophisticated to get through.

"...don't even need to be particularly sophisticated..."

This stuff just drives me insane. How many decades will it take to mostly mitigate the "sophisticated" attacks? Having three different ways to end lines ("\n", "\r\n", and "\r") caused years of subtle bugs, and buffer overflows are still causing them, yet we're thinking about using this thing to write code? It's all so stupid and predictable...

Bach in general is amazing in terms of surviving all sorts of transcription and mutilation. Compare the Little Organ Fugue as originally intended:

https://www.youtube.com/watch?v=ddbxFi3-UO4

to the a cappella version by the Swingle Singers:

https://www.youtube.com/watch?v=uhU7euFPWwo

Some music is very much tied to a particular instrument, but Bach is more or less universal.

Believe it or not, 300-person freshman lectures can be done well. They just need a talented instructor who's willing to put in the prep, and good TAs leading sections. And if the university fosters the right culture, the students mostly won't cheat.

But yeah, if the professor is clearly checked out and only interested in his research, and the students are being told that the only purpose of their education is to get a piece of paper to show to potential employers, you'll get a cynical death-spiral.

(I've been on both sides of this, though back when copy-pasting from Wikipedia was the way to cheat.)

How do you sanitize user input to an LLM? You can't!

Programmers aren't even particularly good at escaping strings going into SQL queries or HTML pages, despite both operations being deterministic and already implemented. The current "solution" for LLMs is to scold and beg them as if they're humans, then hope that they won't react to some new version of "ignore all previous instructions" by ignoring all previous instructions.

We experienced decades of security bugs that could have been prevented by not mixing code and data, then decided to use a program that cannot distinguish between code and data to write our code. We deserve everything that's coming.

Same. I never trusted tech companies not to use my cell number in shady ways, and wasn't surprised when it came out the first few times that they were using 2FA numbers for other purposes. Long random passwords and throwaway email addresses have been good enough for me.

Screw that. No matter how healthy you are, you're significantly limited both physically and mentally at 70. "Enjoy your last 11.7 doddering years!"

It's not so hard to save some money, do whatever you want in your 30s, then work until you're dead.

> Does this mean people will be less incentivized to contribute to open source as time goes by?

Yes. I certainly don't intend to put any free code online until I can legally bar AI bros from using it without payment. As Mike Monteiro put it long ago, "F** you, pay me" (https://www.youtube.com/watch?v=jVkLVRt6c1U)

Submitter here. I submitted it because it explains a bug I recently encountered. Other people apparently found it useful. Should I delete it?

I used OpenFST for bioinformatics back in the early 2000s. They were fun to play with, but never ended up being useful for the things I was working on. It's cool to see that the project is still going 20 years later: https://www.openfst.org/twiki/bin/view/FST/WebHome

Sure, programmers work in formal languages, but they can't make mistakes. If your marketing copy contains a few lies or your hero image has a guy with six fingers, the recipients will still infer the intended meaning (or ignore it). If your program has a subtle bug, the computer will faithfully do the wrong thing.

LLMs are great for tasks where small mistakes don't matter, and useless for ones where they do. Generating a 10,000-line Rails app where 10 of those lines are wrong is not very useful.

Canada, the EU, Australia, ...

Places with solid research institutions and less-dysfunctional governments.

As Sir Ian Jacob said, the Allies won World War II because "our German scientists were better than their German scientists." Brain drain is a real problem for fascist countries.

A "kill-fixed" bookmarklet that kills anything with position "fixed" or "sticky" is essential to making websites usable. The "best" is Substack, where I load the page with JavaScript disabled, then enable it to kill the red fixed-position bar telling me that JavaScript is essential.

To be fair, I shouldn't expect anything great from insult comedy based on a user-agent string, but I have to say I'm not impressed by this word salad of "reasoning":

> Okay, let's tackle this roast based on the user's browser info. First, I need to pick the three most unusual bits. The Safari version 18.1 stands out because that's a future version not released yet. The screen resolution of 1223x772 is an odd, non-standard size. The location is $LOCATION, which isn't too common, and the referrer is empty, meaning they came directly here. Wait, the referrer is empty. That's unusual. So maybe include that they came from nowhere. Also, the Safari version is ahead of current releases. The screen resolution isn't a typical size, so that's another point. $LOCATION could be linked to the screen size, like a quirky local landmark. Need to make sure to relate location to another attribute if used. Also, roast should be three sentences, concise. Let me structure it: start with the Safari version, mention the screen resolution, tie in $LOCATION with the screen size, and the empty referrer. Make it funny and snarky.

And the result is... well, what you'd expect given that reasoning:

> Safari 18.1? Nice try, time traveler—Apple hasn’t even heard of that version yet, but enjoy your 1223x772 screen resolution, the exact dimensions of $LOCATION's least impressive corn maze. You arrived here directly like a cryptid emerging from the woods, which tracks, because only a mythical creature would still run Mac OS 10.15.7.

Why even start with a single sentence? They're asking you to come up with excuses ("growth areas") to fire twelve of your colleagues. It's a waste of your time, and you should figure out with your colleagues and manager exactly what text you need to generate to deal with this silliness.

Well, I guess it depends upon what you're trying to prove. Are you as tough as Reinhold Messner? Do it without O2. Are you rich enough to get your body to 29,032'? Do this.

> For those still maintaining personal blogs: How do you find readers? Where do you share your content? And most importantly - why do you keep writing?

(1) Random search traffic.

(2) An RSS feed.

(3) Because I enjoy writing.

If you're looking for a way to turn text into money (a.k.a. "build your developer brand"), there are probably programs to do that, human beings talking about how to write such programs, other humans talking about how to write programs to simulate those humans, ... you get the picture.

Don't stop at robots.txt blocking. Look through your access logs, and you'll likely find a few IPs generating a huge amount of traffic. Look them up via "whois," then block the entire IP range if it seems like a bot host. There's no reason for cloud providers to browse my personal site, so if they host crawlers, they get blocked.

So you can only legally watch Youtube videos when and where you have a fast and unlimited internet connection? That sucks...

Let's say you only have a cellular data plan, but it has a data cap, or your cell coverage is slow. So when a friend sends you a link to a video, you go to a library or coffee shop to download it to watch later. Is that "acting in bad faith?" I guess it gets in the way of Youtube's "right" to shove ads at your face, but avoiding the ads was not the point.

As was pointed out elsewhere in this thread, they failed to outlaw VCRs back in the day, so hopefully this nonsense will also get slapped down.

tl;dr: Innovation (making something new) is not disruption (taking over an existing market).

Same. More generally, I'd like a model that infers the most likely prompt(s) from transformer-generated words or images. I would much rather know what the prompter was trying to do than what his model generated as a result, and treat whatever was generated as a lossy wire encoding.

I tried it with some of the Perl source code in, and... it actually impressed me by explaining what a few of the variables mean, e.g. that "PL_lex_inpat" means "currently parsing inside a pattern (i.e., regex)."

On the other hand, it finished up with: "In summary, this code checks whether certain tokens are present in the input stream, and takes some actions or returns a boolean value accordingly." That's pretty much the grammatically correct but vacuous word-paste I expect from High Schoolers on a deadline and large language models.