We really need to start having a serious conversation about this....

[user 37518]

In my 12 years of graduate education, I've done it countless times in the schooling context, and not a single time in the professional context. In fact, I hadn't studied math extensively until it was time to my my PhD thesis, and I discovered the math I needed wasn't readily available, and thus I had to derive it (e.g. the entire point of a PhD). The number of people working day-to-day in novel mathematical concepts in science and engineering are a minuscule portion of the workforce, yet it's a focus in education.

You are correct about the majority of people not having to work with novel math. But we have to remember that that minuscule portion of the workforce are those who are responsible for the largest advances in science and technollogy. In my opinion, even if most people dont use such level of math, the focus of education should be to kept this way, otherwise, graduates will only be able to become pencil pushers.

On the other hand, I believe not every university should take this approach, community college or job oriented universities should provide more day to day knowledge.

Again, the problem with AI is not AI itself, but how people use it. The other day, I was asking ChatGPT to give me some book recommendations on a particular topic; I found its answers moderately appropriate. The problem, in my opinion, will be when you go to the doctor's office because you feel sick, and the doctor will "ask Siri" or a similar AI. If the doctor is just getting a 2nd opinion, fine, but at some point this most likely will verge into blindly following what the computer says.

You mentioned calculators when they were introduced. I've seen it many, many times. Students push buttons of a calculator, they get results that dont make sense (like a negative value for energy, or a feedback resistor of 0.000002 ohms) and they just blindly write what the calculator says without questioning. This happens with circuit simulators as well. My students sometimes do the simulations wrong, and they do not care to think, they just accept it as it is. With ChatGPT, and the AI that will follow, I believe these sort of errors will be much much greater. Also, calculators, for example, no longer require the individuals to do hand computations, but we could say that the individual still has to do all the thinking, not so with ChatGPT.

 

[JohnRoberts]

You mentioned calculators when they were introduced. I've seen it many, many, times. Students push buttons into a calculator, they get results that dont make sense (like a negative value for energy, or a feedback resistor of 0.000002 ohms) and they just blindly write what the calculator says without questioning. This happens with circuit simulators as well. My students sometimes do the simulations wrong, and they do not care to think, they just accept it as it is. With ChatGPT, and the AI that will follow, I believe these sort of errors will be much much greater. Also, calculators, for example, no longer require the individuals to do hand computations, but we could say that the individual still has to do all the thinking, not so with ChatGPT.

One of the things about working with slide rules is that they provided the most significant digits, but not the decimal point. So we effectively had to do the math in our head to get in the ball park, and then use the slide rule's precision digits.

JR

 

[crazydoc]

I don't know what kind of slide rule you had, but mine only gave 3 significant digits. I had to use the tables in the Handbook of Chemistry and Physics to get usable answers for log and trig functions.

 

[user 37518]

One of the things about working with slide rules is that they provided the most significant digits, but not the decimal point. So we effectively had to do the math in our head to get in the ball park, and then use the slide rule's precision digits.

JR

I am afraid I am too young for slide rules. However, since I was in high school, I've been a huge fan of HP calculators, there is actually a forum on the web, much like this one, where many other geeks like myself discuss HP calculators, buy/sell, etc.. Anyway, I always loved the fact that astronauts used to carry HP calculators in the space shuttle missions, in case the onboard computers failed and they had to find the return trajectory or similar computations by hand.

Here is a picture of Sally Ride with an HP 41C calculator.

 

[JohnRoberts]

3 digits are more than enough resolution to decide whether to use a 6.8K or 7.5K ohm resistor. But the slide rule doesn't tell you whether you need Ohms or kOhms.

I still have the 11" long plastic Pickett (microline 140) that my employer provided me to use as a technician, back in the early 70s. This was before computers or pocket calculators were widely available. By the end of the 70s we had decent HP hand held calculators and early personal computers. (I may still have my first HP35).

JR

PS; Slide rules just have the Log scales printed on them. We mechanically add or subtract log values from each other by moving the slide and reading the result with a cursor. I don't remember (or ever knew) how to use a fraction of over a dozen different scales on my old "slip stick". I vaguely recall that my dad RIP had a small 6" metal pocket slide rule that fit in a shirt pocket.

 

[crazydoc]

A friend got an HP 35 when they first came out ($395), and let me use it to calculate means and standard deviations for a paper I was writing for anthro class. What a time saver!

I have one I found at a yard sale 25 years ago for $1 - doesn't work but it's history.

 

[user 37518]

A friend got an HP 35 when they first came out ($395), and let me use it to calculate means and standard deviations for a paper I was writing for anthro class. What a time saver!

I have one I found at a yard sale 25 years ago for $1 - doesn't work but it's history.

I have an HP35s, which is the re-issue (now discontinued I believe). I use it every day, but the keys are not that great, you some times get missing strokes, which is a drag because the calculator is amazing. That is what RPN will do to you, you can't use any other "regular" or algebraic calculator afterwards. RPN calculators are becoming rarer and rarer now.

Up until 2016-2017 I used to carry an HP15c in my shirt pocket (something irresistible for the ladies /sarcasm), but now I have downloaded the HP48G app for the phone, again, not only because that was my first RPN calculator since I was 15 (16?) years old, but because I can no longer use non-RPN calculators. By now, my brain is programed to type the numbers and then the operation (e.g., 5, enter, 2, +) rather than the usual way (5 + 2 =). So, if I try to use a regular calculator, I will usually end up ranting and taking much more time than usual to get the correct result.

PS the HP35s is being sold for a fortune on Ebay, and to think that I bought one new for less than 80 dollars just 3 years ago

 

[cyrano]

Why would large animals be any different?

You are for a large part foreign DNA. Some of the processes in your body depend upon it.

 

[AnalogPackrat]

My grandfather had an early battery calculator in the 70s (Databrain brand, IIRC). It even had the square root function <edit: found some pics on the web and it did not have sqrt>. I started out with a TI-30 (with the mini 7-seg LED displays) in high school. Summer of 1984 before starting my BS I bought an HP-11C after a friend showed me his during our last semester of HS. At $110 I had to save a week's wages to buy it. I was hooked on RPN after that. The stack-based thinking made a later job writing PostScript code very easy. Stack computers are really neat systems. After I got into programming I lusted after a 16C, but couldn't afford the luxury. Did a summer internship at a civil engineering firm where one guy had a 41C and loads of programs he'd written to do various calculations for footing design, beam loads, etc. He had a case with all the little magnetic cards in it.

Never learned to use a slide-rule but have one that I found in my grandfather's desk when we cleaned out their house a couple of years ago. It's an Acumath 150 with many scales. Has the instructions and a case.

On my phone I use an RPN calculator app that pretty closely mimics my old 11C. I've still got the original, but some of the keys started doubling c.2002 so it's been retired.

 

[crazydoc]

I use a Casio fx-300MS (I've got a couple I got for pennies) because they are solar powered, and besides all the scientific functions, they convert decimals to fractions, which is handy for laying out projects with a ruler, and converting to drill bit sizes.

 

[JohnRoberts]

I have a HP 32s sitting right next to my keyboard when I need to calc something... Over the years I have owned multiple HP hand calcs, including some programmable ones.

RPN is addictive and hard to use anything else after hooked.

JR

 

[Matador]

Again, the problem with AI is not AI itself, but how people use it.

I guess I'm still not seeing the concern. Students (people) will be lazy, ChatGPT or not, I'm not seeing how the existence of ChatGPT will suddenly make students who don't understand a nonsense answer any worse off than they were before.

Once you understand how ChatGPT is trained, you realize that although it's semantic understanding of text is very good, it's ability to draw inferences is actually shockingly terrible. Hence educators will need to rethink how we test students comprehension more than anything else.

Case in point: ChatGPT can provide a very good overview of the events in MacBeth, because a gazillion people have written about it, so the training set is large and varied. Asking a student to provide an overview of the events in MacBeth would be great fodder for being done easily by ChatGPT.

However ask ChatGPT to explain how the events in MacBeth would have unfolded differently if Malcolm had not been declared the heir of Duncan in Act I, and ChatGPT provides nonsensical answers, because it cannot understand how events might have unfolded in a story that wasn't written. It can't provide alternate scenarios, again, unless some (large) number of people have also talked about it prior to training. You can go further, and ask how Lady MacBeth's monologue might have been different if said by Ophelia from Hamlet, and again, ChatGPT provides answers that on the surface might be shallowly interesting, they don't makes any actual sense if you know anything about the characters, and could properly reason about their motivations.

Perhaps one day it will get there, but it's far from being able to make inferences that require nuanced understanding of source material.

 

[JohnRoberts]

chat GPT is the next bright shiny object to distract us from the man behind the curtain.

JR

 

[user 37518]

Why would large animals be any different?

You are for a large part foreign DNA. Some of the processes in your body depend upon it.

I am not entirely sure what you are discussing, but, even though we are in fact made of microscopic molecules, I can't say that just because of that reason I can extrapolate to the macro level. I am made of atoms, but I can't say that my reality follows the wave equation and its is probabilistic.

 

[user 37518]

I guess I'm still not seeing the concern. Students (people) will be lazy, ChatGPT or not, I'm not seeing how the existence of ChatGPT will suddenly make students who don't understand a nonsense answer any worse off than they were before.

Once you understand how ChatGPT is trained, you realize that although it's semantic understanding of text is very good, it's ability to draw inferences is actually shockingly terrible. Hence educators will need to rethink how we test students comprehension more than anything else.

Case in point: ChatGPT can provide a very good overview of the events in MacBeth, because a gazillion people have written about it, so the training set is large and varied. Asking a student to provide an overview of the events in MacBeth would be great fodder for being done easily by ChatGPT.

However ask ChatGPT to explain how the events in MacBeth would have unfolded differently if Malcolm had not been declared the heir of Duncan in Act I, and ChatGPT provides nonsensical answers, because it cannot understand how events might have unfolded in a story that wasn't written. It can't provide alternate scenarios, again, unless some (large) number of people have also talked about it prior to training. You can go further, and ask how Lady MacBeth's monologue might have been different if said by Ophelia from Hamlet, and again, ChatGPT provides answers that on the surface might be shallowly interesting, they don't makes any actual sense if you know anything about the characters, and could properly reason about their motivations.

Perhaps one day it will get there, but it's far from being able to make inferences that require nuanced understanding of source material.

Students cheat all the time, if they can cheat their way through anything, most will. I am not sure you understand what this thing can really do. As of now, I am studying a second masters in mathematics, some of my classmates who can't solve the hardest problems (I am not claiming I can) are asking ChatGPT to prove theorems or solve many of those complex problems that can't be found with just a Google search, and, in many cases, it does do exactly that with great acuracy. A calculator doesn't do that, asking a question on "stack exchange" is also not the same. Yes, both things will ease the effort, but what ChatGPT can do is way beyond that, it is a brain replacement.

 

[user 37518]

RPN is addictive and hard to use anything else after hooked.

JR

John, never thought you were one of the clan!!! one of us, one of us, one of us!

 

[separateness]

Apologies for returning to an earlier subject in this thread but:
I am not sure we should be surprised that any technology reflects the values of the class which produces it. For roughly 100 years now most of the developed world has existed under a sort of technocratic managerialism (the prior liberal or absolutist systems proving far too messy or inept for such massive societies) of one flavor or another. That a created AI gives answers perfectly in line with the valuations of experts who see the problems of society as mere engineering problems seems perfectly predictable. Perhaps some object to this sort of rule but I am not sure that such a complex and massive society could be managed any other way.

As to why so much of this class interests tend toward the Malthusian, I think the answer is pretty clear. The era of cheap energy (barring some unlikely break through like fusion) is, I think, coming to a close. The input costs of oil extraction are slowly increasing and 'sustainable' alternatives offer nowhere near the ROI we have seen in oil production since WW2. We now have huge masses of people in the developed world quite used to what is historically an opulent lifestyle. As that ends, we are bound to see quite a bit of social chaos and a lower population means a) more to go round and b) fewer potential malcontents. I do not say this to endorse this or any perspective really, I just think it is important to have a sober view of where we are in history.

 

[user 37518]

My grandfather had an early battery calculator in the 70s (Databrain brand, IIRC). It even had the square root function. I started out with a TI-30 (with the mini 7-seg LED displays) in high school. Summer of 1984 before starting my BS I bought an HP-11C after a friend showed me his during our last semester of HS. At $110 I had to save a week's wages to buy it. I was hooked on RPN after that. The stack-based thinking made a later job writing PostScript code very easy. Stack computers are really neat systems. After I got into programming I lusted after a 16C, but couldn't afford the luxury. Did a summer internship at a civil engineering firm where one guy had a 41C and loads of programs he'd written to do various calculations for footing design, beam loads, etc. He had a case with all the little magnetic cards in it.

Never learned to use a slide-rule but have one that I found in my grandfather's desk when we cleaned out their house a couple of years ago. It's an Acumath 150 with many scales. Has the instructions and a case.

On my phone I use an RPN calculator app that pretty closely mimics my old 11C. I've still got the original, but some of the keys started doubling c.2002 so it's been retired.

Again, glad that you are an HP/RPN user like myself. I did learn how to write my own programs for HP calculators, it is amazing, it felt like a computer more than a calculator, even when the scientific Casios have more CPU power. Unfortunately, programming into an HP calculator is something that can be very frustrating, not due to the programming itself, but because, inevitably, the memory of the calculator will get wiped or corrupted at some point or another, and all your hard work will be laid to waste. I remember when my HP48G got stuck and it would reboot and then display the following message: "Memory Clear OK", which basically meant "If you had anything stored, you are screwed"

 

[ruffrecords]

Forth Rules, OK!

Cheers

Ian

 

[AnalogPackrat]

Forth Rules, OK!

Cheers

Ian

Yes! I'd almost forgotten about that one. One of my old work colleagues enjoyed playing with Forth and showed me a few things he'd done with it back in the early 90s.