We should but ive been tracking people I know who are addicted to it and I saw a mental decline instantly. It doesn't take a lot of experimenting to show it makes you an idiot, not using your brain will do that.
I'll be the old man yelling at cloud who refuses to use it while the rest of the world straps on bezos vision for 79.95 a month to go to work typing prompts.
We've been trying to do stuff like that for a long time, e.g., the impact of smartphones. My understanding though is that such changes are so widespread it's really hard to have a meaningful control group in the same cohort etc.
See Goodhart's law: “When a measure becomes a target, it ceases to be a good measure”.
If LLMs are judged by some specific metric meant to track their negative impact, they’ll adapt to make the metric look good without improving the underlying impact.
I used to work at a Verizon store, and we were rated by the amount of accessories we could sell with each phone. In order to boost that number, everyone tried to add on a bunch of cheap phone accessories (clearance items and screen protectors) to drive that number up.
Our goal was more than three, but once we figured out how to sell 5 or more for less than $20, they changed it to total cost of accessories per phone.
Problem is we cannot compare them since the tests change over time. My dad had to learn how to use a slide rule in school. By the time he started his first job though, the first electronic calculators were affordable and so he never used for anything real. Anyone just a couple years older than him would have needed it. By the time I got into school nobody knew anything about them, and so we cannot compare.
I'm sure completing the square has been an important algebra question for that last few centuries and so we could potentially compare how students do at this over the years. However the only thing we can ask is how many got it 100% perfect - if I got it wrong but I did most of the steps right how many points do I get - this is expected to vary by teacher. Even then though, it isn't clear if over the years education hasn't change - should we spend more time teaching this important technique at the expense of something else. Also not everybody takes algebra, and I wouldn't be surprised if scores change as we push more/less kids who naturally wouldn't do well in algebra to take it anyway.
I would hope you can think about the above and find a dozen more confounding factors that I haven't thought it.
There is insane correlation between the increasing use of computers (think: iPad kids) and a drop in every measure of academic ability, which had previously only been going up.
I think it's ok to cautiously draw some provisional conclusions.
Last I heard we don't even know how to measure that in the first place. As my intro to psychology class said IQ tests are a reliable measure, but what they measure nobody knows. (which is to say if you take the test more than once you will score about the same every time)
We have better idea now about what it measures. For example about 35% of differences in IQ scores are associated with working memory capacity. And we have very specific tests to measure working memory. There are more very specific parameters that are correlated with IQ that we can measure. We know quite well what IQ actually is but it's a mix of many different things and not a simple parameter indicating who's smart and who's not.
20 replies
We should but ive been tracking people I know who are addicted to it and I saw a mental decline instantly. It doesn't take a lot of experimenting to show it makes you an idiot, not using your brain will do that.
I'll be the old man yelling at cloud who refuses to use it while the rest of the world straps on bezos vision for 79.95 a month to go to work typing prompts.
There are a lot of things we should urgently do for our own good. So no, we're not going to
Maybe in the past when science was still a thing. It's all money and power now.
How would they avoid the brain damage enough to understand it?
We've been trying to do stuff like that for a long time, e.g., the impact of smartphones. My understanding though is that such changes are so widespread it's really hard to have a meaningful control group in the same cohort etc.
Explain how that increases shareholder value?
See Goodhart's law: “When a measure becomes a target, it ceases to be a good measure”.
If LLMs are judged by some specific metric meant to track their negative impact, they’ll adapt to make the metric look good without improving the underlying impact.
Great example of Goodhart:
I used to work at a Verizon store, and we were rated by the amount of accessories we could sell with each phone. In order to boost that number, everyone tried to add on a bunch of cheap phone accessories (clearance items and screen protectors) to drive that number up.
Our goal was more than three, but once we figured out how to sell 5 or more for less than $20, they changed it to total cost of accessories per phone.
It’s too late to record a baseline, so any metrics will need to work retrospectively anyway.
There's still a bunch of us that refuse to use it.
I Think we already are saving stats on grades, education etc
Problem is we cannot compare them since the tests change over time. My dad had to learn how to use a slide rule in school. By the time he started his first job though, the first electronic calculators were affordable and so he never used for anything real. Anyone just a couple years older than him would have needed it. By the time I got into school nobody knew anything about them, and so we cannot compare.
I'm sure completing the square has been an important algebra question for that last few centuries and so we could potentially compare how students do at this over the years. However the only thing we can ask is how many got it 100% perfect - if I got it wrong but I did most of the steps right how many points do I get - this is expected to vary by teacher. Even then though, it isn't clear if over the years education hasn't change - should we spend more time teaching this important technique at the expense of something else. Also not everybody takes algebra, and I wouldn't be surprised if scores change as we push more/less kids who naturally wouldn't do well in algebra to take it anyway.
I would hope you can think about the above and find a dozen more confounding factors that I haven't thought it.
There is insane correlation between the increasing use of computers (think: iPad kids) and a drop in every measure of academic ability, which had previously only been going up.
I think it's ok to cautiously draw some provisional conclusions.
Last I heard we don't even know how to measure that in the first place. As my intro to psychology class said IQ tests are a reliable measure, but what they measure nobody knows. (which is to say if you take the test more than once you will score about the same every time)
We have better idea now about what it measures. For example about 35% of differences in IQ scores are associated with working memory capacity. And we have very specific tests to measure working memory. There are more very specific parameters that are correlated with IQ that we can measure. We know quite well what IQ actually is but it's a mix of many different things and not a simple parameter indicating who's smart and who's not.
The word I've heard for “reliable” is precise.
If a test meaures what it's supposed to measure, it's accurate.
I took the class 30 years ago so my memory of exactly words were used is untellable. The point is memorable and I believe is correct.
If you want to measure seriously, you shouldn't jump to interpretations early.
I don't think that's a very compelling investment thesis.
don't worry, it'll let us know how dumb we've become