CVPR Daily - Tuesday

7 / 30

has to be able to humble down
sometimes. Sometimes the first
implementor, the person that is doing
the actual work, is actually not seeing
that it could be working, and the group
leader sees that it could be working.
And then the group leader should stay
on track. But sometimes the group
leader is just illusioned, is lost in his or
her own ego, is in love with the own
idea, and then you have to say: you
know what, you’re right. And this
dynamic between bottom-up and top-
down leadership, it is a big feature of a
healthy group.
Do you think there are any changes
between your generation of students
and the generation of students that
you see now?
Ok, ehm… Temporally skip. I give you
just a quick comment, we can continue
after. Just off the record comment…
this is such a awesome series of
questions! [we both laugh] It’s just so
fun. Ok.
[Laughing] Why don’t you want to put
this into the interview?
[Still laughing] Ok, put it in. This is so
cool. So before we re-start… what was
the question again?
[I repeat the question]
Ok, this is an awesome question! See,
it’s very interesting. Let me first say
how the environment changed. When I
was a student, everything was much
slower. You had an idea, and you
thought: it’s awesome! Then you had
approximately a year, or a year and a
half, time from an idea to a publication
because the density of people working
in your area was low and, you know, in
the end you wouldn’t be scooped.
These days, if you work on something
hot, especially on neural networks
understanding something - the very
middle of the field - between an idea
and somebody scooping you, you have
maybe six months. So it’s becoming
more stressing. But at the same time,
because it’s so much denser, if you do
something good you get a lot more
citations. And citations are a big
currency, a big mark of success, that
you trade for positions and
professorships. So in a way it’s harder
and easier, at the same time. But it’s
certainly more stressful, and I salute
Vittorio Ferrari 7
“ The noise is noise with respect to the center
point, but it’s signal with respect to scale”