Subscribe
Share
Share
Embed
We explore the risks arising from the use and misuse of digital devices and electronic communication tools. We interview experts in the fields of cybersafety, cybersecurity, privacy, parenting, and technology and share the wisdom of these experts with you!
Welcome to the Cyber Trapps Podcast.
I am Jethro Jones, your host here at the Inch 360 Conference.
I've got James Als, FOSS on the program.
He's a professor of computer science at University of Idaho.
Anything else you wanna say about yourself, James?
I've actually technically distinguished professor
now.
Oh, distinguished.
I'm sorry.
Pardon me.
Distinguished professor.
I started out computer science, wrote my first line of code in 78.
Through my undergraduate mathematics, computer science and physics, but stuck with this computer science.
Came here to Idaho in 91.
Taught our first cybersecurity class in spring of 92.
Wow
Wow.
Helped build our cybersecurity program.
Ran our cybersecurity research lab for 25 years.
So
yeah that's
a little about me
Yeah that's awesome So one of the The first things I'd like to ask you about is things have obviously changed over the last 35 years since you've been teaching there what are some of the things that have been the
biggest changes and how have they impacted what you do and how you teach and things like that Like is AI bigger than the web was in the late nineties early two thousands You know let's talk about some of those things
Yeah so the first change when I got there the internet wasn't even public
Yeah
I mean I had been using it but it was all academic government So the industrialization the ability to make a profit on the internet was the first thing I saw And the web browsers coming of that was a tremendous change we actually saw I did consulting for
nasa and I was there one year and we were monitoring all the network traffic The next year it was a hundred fold that because suddenly browsing and websites everywhere So that was a big change and people trying to figure out how to do that to make money
And then realizing people were also trying take money from them During So yes we saw that big change cell phones was the next big change I saw you know just disruptive technology
I mean you can look there's a famous picture of the inauguration of Pope Benedict in 2005 and Pope Francis seven years later and just The sea of cell phones that occurred social
media came with that And now ai it's we have a tendency I'll say this in technology especially in electronic technology to create something ship it out there and see what works
Material scientists don't do that They don't build oh I got a new carbon nano or Nanofiber let's build a bridge with it They never do
that that We're like Hey we have this thing called dual core Multi-core computation Let's ship and see how it works We have this thing called ai Let's put it out there We have no idea how it works Let's see if we can make some money with it
and people do it What surprises me is people grab onto this shiny object and say how can I make money with it How can my life easier So it's scary it's fascinating I'm playing with it It's amazing what you can do with it and what you can't do
Yeah Well tell me how you are helping your students use or not use and know when to use AI in in what they're planning to do for a profession
Yeah so that part is new we've been doing it in terms of just Google and finding things It's the thing I try to get them to do is how to evaluate the effectiveness of the solution you get like
when I talk about developing ai I give em examples I'm building this new software using ai I'm using the AI also to help me create all these test cases to see if it's doing the right thing
I had one case classic case I gave em where the test case was wrong Test case was saying the wrong answer So the tool is actually correct but it kept giving me an But knowing where to look and not to trust it knowing to draw the questions
and say is this really doing what I want What would where could it go wrong in testing for that Because if they can effectively evaluate the effectiveness tool then they can use it And so yeah that's one thing I'm focused on with them
Really fascinating And then are you using it in your dayday work You said you've been playing with it and trying it out but are you actually using it
So we had a tool we had developed and we needed to do a translation of that tool to a different programming language So I used it for that as part of my experiments of how we can do this And I'm
still doing that because it's mostly much more productive than I could have been Definitely much more productive than my students except there were times where suddenly it deleted A third of the code
Yeah
Just randomly deleted it so I'm using that I also have been building this tutorial I teach intro cyber and I'm working with dual credit So there's K 12 aspects is can we build this tutorial
system that the students can run on their Chromebooks in a browser And I'm using AI to help build all that just to make my life easier And it's very good at that cause it's very basic level
yep
We are just experimenting more with a student gave me a report they did a survey of the literature and I so I asked AI did they miss any And it found two that they missed It said they missed these four but two weren't relevant
Yeah
So little things like that but still figuring it out
/
Yeah for sure And what I find really fascinating is you can look at those four that came back and say two of these aren't relevant but according to the AI
they are and this just to me underscores the need for proficiency and expertise in an area to know whether or not the AI is blowing smoke or not You know
Yes I was talking to somebody in industry who had done a lot of exper experimenting with AI recently and they think it's a they say our AI is a 2.0 GPA intern
Uhhuh
You have to give them very precise instructions and check their work And then it's okay And then his partner was with him said no it's like having 10 interns at once because it's fast
Yep Yeah Yeah I would second that I don't think that Like when the when these AI companies release new models they say this is like PhD level I don't think it's quite there yet and maybe they're giving some really
specified tests that that could make you think that But in everyday work that I'm using it for it's definitely not there for that But I've seen where there are opportunities for it to do things For example I need
to create a website with a database I don't know how to do that myself But I know enough to like that I could probably figure it out but it would just take me a long time to do it So I use an AI tool that makes
websites and I had it do it for me and it worked and people could log in they could create accounts they could do everything that needed to happen But the real problem is if there's a bug And I tell the AI to fix
it then it doesn't and it'll say oh yeah I fixed it And then I'll go in and say no that's still there And they'll say okay I fixed it now And I'm like no it's still there and that's the limitation that it's at
There was the one test case I was talking about It was like generated the sample code and it said okay here's the step The answer's eight
This next step we add another seven This next step if it's correct we add an eight So the total answer should be eight plus seven plus eight
15 which is what it said it was 15 And so my thing was saying 23 and so I told the ai I said Hey you got it wrong eight plus seven plus eight
It's 23 not 15 said oh yeah I'm sorry I made that mistake Let me redo it So redid it Still had 15 Why Because it's a conversation simulator
Yep
And does probabilities and prob the most common thing it sees is seven plus eight is 15 what it throws out the other eight
Yeah
because it is not thinking it is no matter what you say to it like oh you didn't do this or that It's a conversation simulator So it can do things that common that it's seen before like
build that database the website It's not innovating it's just copying something it's seen before maybe in a little different form so I just gotta keep reminding people it's not thinking
Yeah
you know I asked it this question it said it wasn't biased No
Yeah
It's a conversation simulator If I told you Hey you're biased you'd say no I'm not That's what it's whether you are
especially if it's trying so hard to please you And if it if its goal is to make you happy and feel like you are you are the in the right and then you say you're being You're being
biased And then it says oh you know what I probably am You're right Let me tell you again how I'm not biased And then it will still do the same thing because it's not actually responding
to that in the way that you and I would respond to it It's responding to it in a mechanical probabilistic way that is unique and different than what how you and I would respond to it
And but it comes off so human sounding
Yeah so
thing
I gotta keep reminding my students it's not thinking
Yeah It's
it is a probabilistic process It's very good at certain things Use it as a tool
Yeah
but understand what it's doing not doing So it's fun And when it comes to cyber it's like make sure it actually didn't write code filled with vulnerabilities cause I've seen do that too
Yeah So what's one prediction for the future of where you think this is going
oh where AI is going?
Yeah Or how we use it or I'm leaving this open for you James
Oh I can find a lot of people I have a family member who has used it to increase his productivity on certain things because it was a complex but Specifiable task
Yeah
Yeah I think the effective people using this will learn how to do that give that good prompt give that specification say this is what I need So somebody who's a mid-level engineer or higher can be able to specify what they need and get the effective
out of this
Companies are gonna be assuming it's a silver bullet
Yeah
and we're gonna have problems initially
We have those now I think in terms of cyber or computer science just code generation I did a study years ago and 60 of the people in Jobs department of Labor statistics say need a bachelor in computer science Didn't have a bachelor in
computer science They come from other disciplines a lot of em don't have any cyber background Don't know about buffer flows don't know about that They're writing code that ends up being on our internet of things devices whatever And vulnerable
Yeah
AI's gonna be doing the same thing And companies are like Hey just We're gonna use this we're gonna make money let's move on cause that's the bottom
line for companies and you know I keep telling my students the other thing I tell em is just because you can do something doesn't mean you should
Yes absolutely
but they do Hey I can make you know internet enabled toaster why not You know let's do it
Yeah
Yeah the kitchen appliances that are internet enabled are one of my favorite things to poke fun at because So like there's so little utility for that and and yet people still will are making those so
James this was a great conversation Thanks for taking the time thanks for coming to Inch 360 this year and anything you wanna plug or anything for people to get in touch with you or learn from you Vandals
Go Vandals we have a bachelor's of cyber actually a bachelor's master's PhD in cybersecurity we were one of the five first five universities to be designated by this NSA as a
center of excellence in cybersecurity education We're smaller we're not as well known but we're doing really good work and our so anybody hiring We got some students graduating
Yeah.
Let's go.
Very good.
Thank you, James.
Appreciate you being here.
Thank you.