My name's Howard Shrobe.
People usually call me Howie.
I'm a principal research scientist here at CSAIL,
and I work in the areas of cybersecurity and artificial
So I work in cybersecurity and often use AI techniques
to work on that area.
This involves ideas like building systems
that can model the way attackers think about attacks
using AI planning technology and then using
that in the context of various kinds of defenses we build.
It also involves fundamental work
on new architectures that are inherently secure.
And from the AI side, I'm currently
working in a new project, which is
focused on what cognitive psychologists call
Theory of Mind.
That is the ability to model what other people are thinking
Cybersecurity is a very rough area
because it's fundamentally involved in conflict
with other planning and intelligent beings.
So we're trying to make a system safe and secure,
but the attacker is also creative.
This is very different than when you're
trying to engineer for safety where you're worried
about natural failure modes.
Nature does not collaborate, whereas attackers do.
And so you have to think about defenses that don't involve
simply using redundancy.
On the AI side, the challenge is that if we're
trying to model how other people think,
the first subproblem is how do people think at all.
And although AI has been very successful in recent years,
on the cognitive side we still have major challenges
And that's one of those.
So if you go back maybe 10 years ago,
we were just at the beginning of seeing
this becoming a major threat.
Today, we have seen actual attacks
that have disabled power grids for example, in Ukraine.
Twice that was done.
The city of Kiev was turned off by cyber means.
We're seeing widespread use of ransomware
against governments and other large organizations.
So what looked like a potential threat
maybe 10 years ago or maybe even a little more
is now a real threat.
It's something we have to deal with day-to-day.
Of course, in this election year we're
talking about it all the time in terms of can
we trust our own elections.
So that landscape has changed completely.
The degree of technology we've developed for defense
has also increased, but the attackers, I think,
have gained capability faster than we're
gaining defensive capability.
One of the things that makes CSAIL relatively unique
is our breadth and depth.
We always talk about the depth of the research done here.
And that's true.
But the other thing is we're just a huge lab
and we cover almost every area of computer science.
So being in this environment is special because no matter
what you might be interested in, and that's
often something you hadn't thought
you were going to be interested in,
there's somebody here who is a world class expert in that.
And usually it's pretty easy to start a collaboration
when there's a good opportunity.
So for me that's one of the really important parts
of being here.