About three months ago, I had Copilot create a data table, which to my surprise came out as intended. The problem is, the table couldn't be copy-pasted, edited, printed, or otherwise accessed outside the chat. My only option to "get the table out of Copilot" was a screenshot. I had ChatGPT create it for me instead.

Earlier this week, I asked Copilot to add a couple of events I received via email to my calendar, which it claimed to do. Today, when it was time for the first event, I noticed there was no Outlook event on my phone. I looked for the event for tomorrow, it wasn't there either! It turns out, Copilot just hallucinated the whole thing. A "add this to my calendar" feature is probably one of the few genuinely useful features GenAI could bring to Outlook, but apparently they can't pull off the most obvious use case.

I'm "pushing back" on Copilot because it point-blank lied about adding something to my Outlook calendar and creates outputs that are literally unusable for me (not metaphorically - I needed to get the data table out of Copilot). I'm frustrated just thinking about it.

I said it's rare, not that it's unheard of. Like a true academic, I left myself a nicely sized escape-hatch with that wording.

What you're talking about is usually isolated to certain technical fields that require minimal equipment and can be practiced without hurting anyone, and even then it's the exception not the rule. Good luck finding a non-degreed doctor, dentist, RN, pharmacist, physical therapist, civil engineer, mechanical engineer, rocket scientist, aeronautical engineer, economist, roboticist, CPA, LPC, LBSW, etc. I just named several sizable FIELDS that are pretty much 99.9% occupied by people with those useless degrees. You might be able to find a handful of edge cases where someone managed to get a variant of those titles, but for most part you're not even getting an interview without a degree, and can't even hold the position legally otherwise. And for good reason. Certain fields require sustained, rigorous practice with structured curriculum designed by people who have walked the road, over a protracted period of time. And the higher the cost of learning materials - like patient lives, dangerous engineering equipment, etc. - the higher the need for structured supervision.

And if I may appeal to your ego, I'm going to also bet you've spent a lot of your life feeling that many, if not most, people are downright dumb, because they fail to grasp fairly clear general principles. It's not that they're dumb, it's just that you're unusually smart. Most people aren't going to "pioneer two industries" without sustained supervised practice. If you managed to do it, good on you. But you are the exception, not the rule. I've been in education long enough to see that many highly intelligent people often suffer from the "curse of knowledge," which makes it hard to communicate with people who aren't as knowledgeable as them. What I'm getting at is, you underestimate yourself and overestimate others, which has led you to believe that society can function on the backs of autodidacts. It can't, there aren't enough of them, and therefore structured degrees are needed.

My dad used to work as a repair technician. On more than one occasion, he had enough work to take on an apprentice. Invariably, they'd quit as soon as they got useful to start their own business.

Under the current system, an employer is relieved of the cost of educating the employee (and the risks that go with it) with the trade-off that they pay for the tuition via the employee salary schedule itself. Thus, a college educated worker commands a "higher salary" that covers the upfront education cost. If we switched to employer based training, the employer would need contractual agreements with the employees to avoid what happened to my dad, and the starting salaries would be lower to cover the education costs and risk of an employee not working out. For technical fields, it would also require extensive investment in curriculum development and assessment systems, which would effectively make employers turn into mini colleges anyway (bearing all up front costs, and then passing that down to employees in the form of lower wages).

I'm not saying the employer based system would be entirely bad. But it comes with its own set of problems. In a similar vein, I've seen people tout "barter systems" that completely ignore the downsides (like what, carrying around flour to trade?)

I'm an academic.

I've got a good thing going here, and if you don't keep the lid on it, I might have to get a real job! I make at least half of my living emailing office hours to students, since I have a bad habit of putting them at the top of the syllabus that no one reads. Granted, it was hard work deciding which $100 textbook to make my students buy so I could read my free publisher copy to them verbatim, but that was a one-time adoption activity. And then there's that committee that meets in the printer room when me and the other member happen to be printing at the same time.

In all seriousness though, "very few degrees are useful" is inaccurate. All high technology in society requires extensive, structured study. It is exceedingly rare to see any serious advancement at the fundamental level come from someone without extensive education. The Manhattan project, Internet, air-travel, modern medicine, robotics, industrial systems, etc., are all courtesy of someone with extensive formal study. There are non-degreed "skilled workers" who might take a wrench to an airplane without a degree, but you'll be hard pressed to find someone designing an airplane without a degree.

Now I do agree that a lot of degrees are useless, and it can be hard to tell which ones are a good investment. Ulterior motives have crept into a lot of higher education, just like they've crept into health care.

https://time.com/6273743/think...

There is a huge, fiery asteroid called superintelligence about to crash down on society. We're like the dinosaurs wondering about that new star that appeared in the night sky. Surely it'll just go away. It's not getting bigger. It's not getting closer. Maybe it'll have food? I bet it has food. It'll grant our wishes. Maybe it'll just go away if we think and hope REAL HARD. I bet the asteroid is marketing hype. Calling the asteroid marketing hype will make it leave, everyone! If we all work together to call it marketing hype, it'll just go away. Don't look up, just pretend it isn't there.

I taught high school for years, and my students toward the end outperformed everyone else in my district.

A few things you need to understand. First, admin will not let you just send kids to detention, schools have disciplinary policies in place you have to follow. You definitely could write an office referral at most schools, but if you're the lone teacher writing a million referrals they're going to think you're the problem. Also, if kids don't like you they wont want to learn from you, so your students will perform poorly. This is just an unfortunate fact of life. Also, parents will complain when they find out about that homework punishment, and again, I doubt any admin would back you up. Also, if you made that threat, I bet you most teens would smell blood in the water and the ENTIRE CLASS would in unison chant the meme.

You'd probably have more luck just making a game of it and having fun with the students. If they actually like you (I don't mean as a friend, I mean as an adult they know cares about them) normally when the joke is over they'll actually their work as a reward for you not being a jerk to them.

This is moving toward ChatGPT as an OS. The only problem is, the heavy-hitting apps (Word, Excel, Google Maps) are going to stay away because Google and Microsoft can see where this is going.

Look! It's a vibe-coded version of the Godel Escher Bach "strange loop."

I taught high school math for years. Toward the end, my students consistently outperformed everyone else in my school district. Basically what I'm saying is I've actually taught, extensively, and successfully.

The fact of the matter is, you're dead wrong. A huge room of 25 teens is just not going to behave how you think or want them to. Until you've spent a few years in a classroom, you just don't have a clear mental picture of how managing teens works.

Locking the phones up is exactly what needs to happen. The next thing that needs to happen is for people like you to shut the hell up.

I think there's a bit of a misunderstanding here.

If I work real-hard digging a ditch and filling it back up in my back yard every weekend, I'm not magically getting rich because I "worked real hard." My boss isn't going to say "good job digging that ditch at home and filling it right back up." Hard work in and of itself means nothing. However, many activities that do lead to wealth, or at least definitely help get wealth, require a lot of hard work and persistence. For example, most professional middle class jobs like accountancy, law, software engineering, teaching, nursing, medicine, etc, require a ton of work to master. And when you've mastered a skill like that, it's definitely a lot easier to get wealth, even if it's not guaranteed. Similarly, running a small business is a ton of work (even if it could go under). Climbing the ranks of a corporation is a lot of hard work (even if you could get side-lined, passed over, or stuck).

But it's also possible to bust your butt at a dead-end-job for nothing. Jacob in the Bible gives a pretty serious scolding to his uncle since he was basically put in this exact situation. He busted his butt, but his boss kept changing his salary. In Jacob's case, God looked after him, but otherwise he would have worked hard for nothing.

I think the olde social contract was that if you work hard for your employer, they will reward you hard work with more salary, promotions, etc. That idea, though, does seem to be dead.
 

What kills it for me is it's more expensive. They want to charge a "premium' for not killing the environment or being healthy or whatever. To be blunt, if I could get the double cheese burger for $0.50 off (a reverse up-charge) I would probably go for it. But I'm not paying a "premium" for an inferior tasting product. And I really don't understand how pressed beans actually costs more than a beef patty.

Now plant based meat that costs half price the real thing would make a killing.

I agree base 2 seems most natural in this context. Of course, if we do use different bases it just re-scales the growth by a constant factor via the base change formula, so I think it's probably just ignored for that reason (see first link). As for my notation, if you look under the section Using Common Logarithms in the OpenStax textbook section (see second link), you'll see a discussion on the meaning of "log" in the context of the common logarithm. Usually in math log is the common log, or base 10. But in computer science, sometimes log means log base 2. If you look at the article that talks about the new algorithm, you'll notice they just use log too. Since it's also about computer science, they may have meant base 2, but I'm not sure. Eitherway, I don't think it matters since it just re-scales by a constant factor anyway. Strangely, lb(x) sometimes means log base 2 (binary log) and lg referred to the common log. Ew.

https://cs.stackexchange.com/q...

https://openstax.org/books/pre...

I accidentally switched to y's, that should have read x\logx / x\log(x)^{2/3} = log(x)^{1/3}.

Okay I'll bite. I didn't spend a lot of time proof reading this and re-thinking it, but I think it works. Anyway, it's sketchy to say "for which" values of v and e, since the growth rate is asymptotic. But approximately for any particular implementation, we just set the growth rates equal.

Taking the complexities from:

https://arxiv.org/abs/2504.170...

We get that the two equalize at:

mlog(n)^{2/3}n = m + nlog n

If we switch m for y and n for x, and solve for y, we can write y as a function of x as:

y= \frac{y\log(y)}{y\log(y)^{2/3} - 1}

Which you can copy-paste into Desmos to see the approximate curve along which the two algorithms equalize. To see the behavior as the number of vertices increases without bound, you just take the quotient of the leading terms, which gives y\logy / y\log(y)^{2/3} = log(y)^{1/3}. So asymptotically, the new algorithm wins when the number of edges is less than the cubed root of the log of the number of vertices. But I really didn't carefully read everything so take that with a grain of salt.