[skybrian]

Depends what it is? Detecting a cycle is something I learned about a long time ago in a Lisp context, but haven't used. It's simple enough once you know it, but I doubt I could have come up with it on my own without thinking about it for a few days; certainly not on the fly.

For many front-end developers, I don't think graph algorithms matter. I suspect they don't matter for machine learning (though I know little about that field). Would you pass up knowledgeable people in these fields?

This is why you need multiple people to ask a lot of different questions and judge people by what they can do, not what they can't do. And try not to entirely forget about hindsight bias.

[stcredzero]

Detecting a cycle is something I learned about a long time ago in a Lisp context, but haven't used. It's simple enough once you know it, but I doubt I could have come up with it on my own without thinking about it for a few days; certainly not on the fly.

When I took Comp Sci, you received some very general, broadly applicable tools. What I'd expect from my Comp Sci classmates, would be looking at such a problem for a couple of seconds, then they'd say, ok, you can do [X]. We were educated with a toolset that allowed us to do that. If you are thinking of it in terms of cycle detection being a particular specific obscure thing that you'd never have to do, let me say this. 1) If you think of it like that, and you further tell me that it would take you a couple of days, then that immediately tells me you don't have a particular, very broadly applicable toolset. 2) There are contexts where you have to do the kind of systemic thinking where cycles are something you have to account for.

There have been fields where knowledge has been lost. The British Navy figured out how to stop scurvy, then lost the ability to do so. The Japanese figured out how to stop the beri-beri deficiency disease, then failed to propagate the knowledge. I am starting to wonder if academia in the Bay Area and California can effectively compete with startups and big tech companies for Comp Sci expertise commensurate with teaching. Maybe only the very top institutions can do this?

[skybrian]

I don't think graph theory is in any danger of being lost. That seems overly dramatic.

Since people can learn from each other, it's hard to say what people really need to know the day they join the team.

[stcredzero]

I don't think graph theory is in any danger of being lost. That seems overly dramatic.

The Brits and Japanese had no idea of the underlying mechanisms, and our society has the underlying comp sci knowledge in our case, archived in libraries. If an entire generation of computer science grads just fluffs graph theory, it's not like the knowledge will be lost, but it's definitely not a good thing. It's definitely a step in the wrong direction.

However, it was not graph theory which was lost. It's a body of engineering knowledge for how to do applied graph theory; how to recognize you have to deal with graph theory, then quickly clobber the problem with several broadly applicable tools. Why isn't this being passed down from one generation of profs and TAs to the next?

Since people can learn from each other, it's hard to say what people really need to know the day they join the team.

It's one thing to know a high level overview about something, then to be able to go and bone up on a specific area. It's another thing if someone doesn't know enough to recognize the thing without prompting and clearly has no practice solving problems of that general category, whatsoever.

EDIT: Isn't anyone curious about what these tools are?

[tkxxx7]

> EDIT: Isn't anyone curious about what these tools are?

Um, Not when you are making readers feel incompetent for not having them.