[Retr0id]

> it's like trying to teach advanced machine learning to someone with an IQ of 100. No matter how well-intentioned the advice or how clear the potential benefits, there's a fundamental mismatch between the cognitive requirements of the task and the available cognitive machinery.

This obviously isn't core to the point OP is making, but I find it hard to believe that someone of "average" intelligence can't learn about advanced machine learning.

[jalalx]

What is obvious is that OP is really into IQ elitism. Looking at the ‘A More Realistic Framework’ suggestions, it asks people to aim for the bare minimum.

[bluSCALE4]

No, it's not suggesting the bare minimum. It's asking the person to openly do what can be proven to provide the best value as long as the missed administrative tasks or fails are down-playable enough to not damage your reputation. This is a very easy framework for me to funnel my thought process through.

Recently, I've been in the position where I'm the only one taking ownership of new work and initiatives, sometimes to my own detriment. For example, I took on a feature that required an entire validation system be put in place. Though I succeeded in successfully completing the system, the feature was not fully complete. The validation system was not visible. Though if leverage, it would lighten everyones load, it largely went unnoticed and what was observed was that I pushed out a buggy feature. Though I resolved the bugs within a sprint, my reputation was damaged to the point where I took on a new feature and promised it would be done in a sprint. I succeeded but wasn't able to complete it due to the api team only finishing 2-3 of 8 apis. This sort of test I made for myself also had low visibility so even though it was an impressive feat by my team's efforts, it also largely went unnoticed.

Though the author is wrong to equate work-throughput with IQ, he's right to find a way to inform trusted management with this work ethic and he's right to make sure when high value tasks are taken on, that it be made clear to product so credit can be given.

I have team members that are doing the bare minimum but cross lots of t's and dot many i's and that gets them equal recognition as me who often misses deadlines but does lots of heavy lifting.

[lukas099]

On the bright side, what you’re doing sounds more fun than crossing lots of t’s

[bluSCALE4]

Funny you say that. I always describe my job as fun. Just have to have the main product manager in your corner.

[Spooky23]

I think it is core — the premise is that ADHD in some presentations makes basic administrative/maintenance activity difficult or impossible.

With machine learning, the concept or layman level explanation to support the application of the tech is definitely doable for the average person. But the average person won’t be able to get the math.

Folks I know with ADHD generally adapt well, with systems or coaching that they get from professionals or folks supporting them. Everyone is different, of course, and folks struggle with different things.

He’s right with the other stuff too imo. People who deliver get grace from non-core failures.

[Aeolun]

Maybe if they really like math? I think there’s probably some correlation to being able to easily reason about these things and enjoyment.

So while you could teach anyone, with enough effort, they would not actually enjoy the experience.

[mewpmewp2]

It is interesting that you did opt to add in the word "about".

[TimTheTinker]

> can't learn about advanced machine learning.

There's a difference between grokking an ELI5 explanation and being able to build, modify, or review an existing ML model

[Retr0id]

And that difference is mostly down to having domain-specific knowledge, which is not something an IQ test measures.

[hilux]

Can someone with a IQ of 100 (whatever that means) understand and implement stochastic gradient descent?

Maybe - but very possibly not.

[nrnrjrjrj]

I think they are right. Advanced ML requires both IQ and career dedication. Doing say the intro YT series building yoyr own LLM and understanding and sucessfully using PyTorch data strucutres in 3 even 4 dimensional (as in an array of array of array of arrays) doing matrix algebra...

I would say this gets possible at 120 and only easy at 140 at a guess.

Edit: i am serious

[calebkaiser]

Anecdotally, as part of my job, I've taught ML concepts to a lot of people. I don't know if I've ever worked with anyone who was simply too "unintelligent" to grasp things. The bottleneck, as for most things in my experience, is time and motivation. In any niche of ML, there's just a lot of things to learn, particularly if you aren't starting with a particularly math-y background.

[hilux]

You were probably working at a fairly advanced tech company. As were the people you were teaching.

If by "ML concepts" you mean "how to implement ML" - do you really think that people who got to a job that required learning this, represent the population IQ distribution?

[jdjdnndn]

I cannot tell if you are serious.

[lr4444lr]

How many people of 100 IQ could even develop a working knowledge of the necessary linear algebra?

Most college students I took it with, probably average a standard deviation higher, did pretty middling in the class.