[reachtarunhere]

One of the points of the talk is emphasis on working on "important" problems. It does make sense to not work on incremental things for merely publications so it is definitely good advice. However once you decide to do so what is "important" becomes a difficult question.

I like this article from Daniel Lemire which explores this further

https://lemire.me/blog/2010/03/22/so-you-know-whats-importan...

[Swizec]

The way I’ve always interpreted this is that if even you don’t think it’s important, why are you doing it?

Sure you can’t accurately predict what will and won’t be important long term, but you should think your work is important. Whether you’re right or wrong time will tell.

[reachtarunhere]

Yes my interpretation is that this is more an advice on what not to do - frivolous stuff you don't believe in

[emrah]

> It does make sense to not work on incremental things for merely publications so it is definitely good advice.

True but unfortunately this is how funding works based on what I saw during my time working at various labs.

You need to provide enough evidence that what you are after is going to "work". Most of the brand new ideas get resources by repurposing data from existing funded projects. If you don't have what you need, you finangle the funded project to produce the data the new idea needs.

I'm all for not syncing resources into crazy ideas that will never work but current state of affairs (it's getting worse) is too conservative

[marcosdumay]

It is important to point out that on his speech (and the book) he tells people to work on important problems (always on the plural), and that he never say not to work on non-important ones.

In fact, I remember the book having a very clear assumption that you can't work on important things all the time anyway. But well, there has been some time since I've read it. But it is a very sensible and nuanced advice for highly ambitious people.