I understand the short-term imperative for such things: it's good to know that there are timelines in which your work has some kind of measurable impact, and even better to know that there's more than one (BPP vs NP, anyone?).
But I get the sense (and maybe I'm just off base here) that this kind of future prediction business is more common in non-theoryCS areas. My archetypical story for what happens if you ask theoreticians about future directions is Jeff Erickson's hilarious tale about his interview at MIT.
Of course the most famous example of future projection is in mathematics ! So maybe my premise is doomed ? But somehow I don't think so. I don't think mathematicians since Hilbert go around proposing future directions for entire areas (although there might be general consensus on key open problems), and I think theoryCS has absorbed much of this ethos (although I don't think that's true in theoretical physics).
I ask because I always feel awkward when asked questions like "where is
For more mathematically oriented areas, asking where a field should head might be like asking, "Which theorems should be proved true?"
ReplyDeleteIt's not an entirely sensible question.
I think when people ask where an area is headed, they're really asking, "What applications do you expect to deliver on? Or, what specific real-world problems do you expect to make progress on?"
If you take the view that theory = !applied, then these questions aren't very sensible either.
What do you think? If asked the latter two questions in a hypothetical job search, how would you respond?
Well given that I'm somewhat applied, I could partially answer both questions. But a theoretician would in general find it difficult to talk about 'real-world apps' unless their work actually had close connection to such problems, or they were willing to go out on a REAL limb ("my work on hardness of approximation DIRECTLY impacts the future of e-commerce").
ReplyDeleteI don't take the view that theory = !applied. I take the view that theory and applied are largely independent with some mutual information (in the probabilistic sense), rather than being mutually exclusive
Actually, one could give sensible answers without succumbing to the impossibility of prediction. For instance, an observation along the lines of "probabilistic methods are becoming mature, computational power is becoming abundant, so I think the next decade will see rapid advances in computational modeling of domain x which has previously been hard" isn't all that vacuous. It involves a description of the present but it also involves the recognition of a future confluence.
ReplyDelete