December 15, 2025 • 51 mins
Retired homicide detective Graeme Simpfendorfer reveals how artificial intelligence could finally crack Australia’s most notorious cold cases. As AI revolutionises murder investigations and predicts crimes before they occur, criminals are learning to weaponise the same technology - so who will master it first?
Listen to Graeme Simpfendorfer’s earlier I Catch Killers episodes - part one here and part two here.
Want to hear more from I Catch Killers? Visit news.com.au.
Watch episodes of I Catch Killers on our YouTube channel here.
Like the show? Get more at icatchkillers.com.au
Advertising enquiries: newspodcastssold@news.com.au
Questions for Gary: icatchkillers@news.com.au
Get in touch with the show by joining our Facebook group, and visiting us on Instagram or Tiktok.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
The public has had a long held fascination with detectives.
Detective see aside of life the average person is never
exposed to. I spent thirty four years as a cop.
For twenty five of those years I was catching killers.
That's what I did for a living. I was a
homicide detective. I'm no longer just interviewing bad guys. Instead,
I'm taking the public into the world in which I operated.
(00:23):
The guests I talk to each week have amazing stories
from all sides of the law. The interviews are raw
and honest, just like the people I talk to. Some
of the content and language might be confronting. That's because
no one who comes into contact with crime is left unchanged.
Join me now as I take you into this world.
(00:46):
In part two of my chat, we've retired Detective Graham Simfendorfer.
We dive further into the world of AI and how
it's going to change the world of crime. It's quite
mind blowing where we're headed. My sense is, if pleasing
in investigations, the life of a criminal is going to
get a lot tougher. But maybe I'm looking at things simplistically,
(01:07):
because no doubt the crooks are also going to embrace
AI as well. That may counterbalance the advantages law enforcement have.
We also had a close look at how AI may
help solve cold cases and how it could be used
in the court system. It's a really interesting chat that
I learned a lot from Graham. Welcome back to part
(01:27):
two of Eye Catch Killers. I'm scared AI is taking
over the world. Yeah, I want to jump straight into
cold cases because homicide was my passion in my career,
and I know you worked homicide and people are just
fascinated with cold cases and the thing with homicide until
they're solved. It doesn't matter if they're fifty years old.
(01:49):
People want results when someone's life has been taken. How
can we use AI in the cold case space?
Speaker 2 (01:57):
Yeah, great question, and some significant pieces can play a
really important part in this. So there's the aspect of
reviewing all the old information, laying a lens across what
we know, open source around that information as well. So
it's large data sets of information that, like you said,
to get your head around a cold case, you literally
sit with a box for a month.
Speaker 3 (02:18):
Or so trying to understand it all.
Speaker 2 (02:19):
If you can put all that data into AI, it'll
pull out for you what the key issues are where
the gaps are perhaps some avenues of inquiry. Even so, again,
like we said before, you don't want to lose the
skills as an investigator, but it can process that large
set of information really quickly and accurately with what's been
given in those statements and may even be able to say, hey,
look we need to go back further because we've seen
(02:41):
with cold case is the way statements were taken ten
years ago, twenty years ago, very almost almost leading, if
you will, because you're trying to cover your points of
proof and not that whole story that you get. So
one of those aspects is going back if you can
to speak to some of those witnesses and the AI can,
So let's go back and explore this part a bit more,
or the person that witnessed that vehicle drive pilet's go
(03:04):
back and see that. There's that aspect of understanding the
investigation and I guess the investigation plan. But then there's
also review of either CCTV or face images, and that's
where the facial recognition technology could come in and play
a part.
Speaker 1 (03:18):
Right, So the cold cases, okay, and when you said
you know, the AI might say, have you thought of
this or line of inquiry, not taking that personal as
an investigator, because what you're dealing with AI, they're drawing
on the experience of all the in theory at its best,
(03:39):
all the world's greatest investigators looking at this case, all
the information that they've learned from past things. Because this
am I describing it correctly. Basically, they're drawing on all
the information.
Speaker 3 (03:51):
That's right.
Speaker 2 (03:52):
You're taking every into account and thinking outside the box
if it needs to, and that's the skills of great
investigator processing that, and you can teach it to think
about the laws of evidence and making sure that whatever
strategy or tactic that it's thinking of is complying with
the law and complying with the rules of evidence. So
it'll do all that for you. But yeah, you're right,
(04:13):
it's taking the best of everything and applying it to
your case. So for cold cases, it's worth a review.
It's worth looking at a ability to either you know,
fix fix some footage or make the CCTV less grainy,
or looking at a face image that was previously no
one knew who this was or couldn't be solved. And
(04:34):
that's where this journey for me kicked off. Actually was
looking at. A friend of mine showed me a documentary
series that I think was on Netflix about the there
was nineteen seventy nine Lunar Park fire.
Speaker 3 (04:47):
Oh yeah, I remember that.
Speaker 1 (04:48):
There seven people died in the on the Gates train.
Speaker 3 (04:52):
Is that one?
Speaker 2 (04:54):
And I watched it. Got a degree in fire investigation,
So the scientific brain of me was very intrigued by
by that aspect. But then as the documentary went on,
it was Caro Meldrem Hannah that did that documentary, an
amazing documentary by the way, And towards the end, there's
this face image of who they believed lit the fire,
and there was allegations of being linked to by keys
(05:16):
all the like, but there was this face image, and
I just could not understand how in this day and
age we don't have technology that could turn that face
image from just the sketch image that we're used to
seeing into a actual person and then use the data
sets and AI two search for that person through whether
criminal networks or you know, there's some ethical considerations around
(05:37):
searching against open source, but how can we not have that?
And my rest that sent me on the journey of
researching who surely someone could be doing this and my
investigations globally is no one.
Speaker 3 (05:49):
No one is.
Speaker 2 (05:50):
So we started that journey with program AI around turning
a cold case composite sketch image of the old hand
drawn ones, using AI to turn that into a three
D verse of a person, and then you've got a
much better identity of who that person could be searching
against either open source or criminal databases for an identity,
and you need to potentially age that person. Obviously, if
(06:12):
you're talking about a cold case now we've all seen
it on your Instagram reels, is what are you going
to look like when you're seventy The AI will aid you,
and it can aid you, but again it's only as
good as what's put into it. So you're relying on
the witnesses version of what they remembered, and we all
know we would see you know, this is a seventy
percent likeness or this is an eighty percent likeness. There
(06:32):
I can turn that seventy percent likeness into something and
use it as a tool to search the sources, but
it's still the investigator's job to actually is that the
person or not?
Speaker 1 (06:42):
Okay, And so there's a responsibility of the type of
information you put in And you've said with creating that
or the image and whether you put it into open
source when you're talking open source, you put in out
there on social media everywhere.
Speaker 3 (06:57):
Yeah, that's right.
Speaker 2 (06:58):
Yeah, so open sources, open source intelligences is the web. Yeah, Instagram, Facebook.
Speaker 1 (07:05):
So Google put put out there anyone.
Speaker 2 (07:07):
Yeah, And there symmetical considerations because you know, when you
upload your photo to Facebook, you're not really giving permission
for that to be used or scraped is the term.
You're not you're not giving anyone authorized to scrape all that.
So symmetical considerations that we looked at. But that's the
stuff we need.
Speaker 3 (07:22):
To work through.
Speaker 2 (07:23):
If we're looking for a murderer or a rapist, are
the laws going to be sufficient to help support law
enforcement to identify that person?
Speaker 1 (07:31):
Well, it should be like we've had to as technology
evolves and the powers need you need the authority and
it might be for listening devices from the Supreme Court,
that type of thing to get permissioned on the ghost
train one, did that evolve into anything that what you're
looking at.
Speaker 2 (07:50):
Still a work in progress. So we're still using that
as a test case too. It's a great test case actually,
to turn that image into a person, but also age
that person and what they might look like years ago.
Because seventy nine there was no Facebook. You need to
aid that person to see if that person is actually
out there, either on a license photo or a passport photo.
So law enforcement has access to all that already. So
(08:11):
that's where we're at, and we're just trying to train
that so to train the AI to do it. I've
had some friends sketch myself and then running across open source,
and it's showing some really positive results.
Speaker 1 (08:21):
We've got positive in terms of the way you're aging.
Speaker 2 (08:24):
That's right, not like a fine wine though positive In yeah,
it's around ninety six to nine. Anything over ninety six
ninety seven percent is showing as accurate okays speak.
Speaker 1 (08:36):
If we break it down in percentages, that's a good
percentage to work with, good.
Speaker 3 (08:40):
Tool, isn't it.
Speaker 2 (08:41):
So to either rule that person in or rout, you
might get a hit on this likeness is like this person,
but then your investigation says, actually that can't be that person.
That person left the country two weeks before the offense.
Speaker 3 (08:52):
Okay.
Speaker 1 (08:53):
Another thing that comes in so with AI, like we
used to get when I first started place. Are the
big penries. You know, someone during the sketch, this is
what Graham looks like on that, And there was some
good and there was Subjectivity came into it when people
are doing the sketches. But AI, when it's doing, they're
dragging all that information, isn't it. Where there's been failings,
(09:17):
where there's been successes. That's the type of access to
the material we're having.
Speaker 2 (09:23):
And that's how you want it to learn. So you
don't want it with a bias towards a certain culture
or a certain a certain inference. I guess you want
it to be accurate to what it is you're looking for.
So your sketch that's only seventy percent accurate of the
person is only going to get seventy percent accuracy on
what you're searching for using the AI. So you've still
it to use investigative brain to think about who it is,
(09:45):
et cetera. But it'll do it so quickly, and I'll
search your license or your your prison photos that are
taken time and time again, I'll do it in seconds.
Speaker 1 (09:56):
This is hypothetical, and it's hypothetical, and it's talking about
actual investigations, but it's just got my mind ticking over
the barrable. I call it cyril killing of three Aboriginal children,
Colin Walker, Evelyn green Up and Clinton Speedy back in
nineteen ninety one, over a five month period, they disappeared
(10:17):
and their bodies were found. That case has been worked
on for over thirty years. I became involved in the investigation,
I think five or six years after the children disappeared,
for a reinvestigation, and we revisited like you do on
a cold case, revisited, and the big argument initially was
were these matters linked? And we went through tendency and
(10:40):
coincidence evidence and strongly show that the likelihood they have
been linked. I'm just thinking of what you're talking about.
With all that and my memory, there's probably about the
lever arch folders sometimes I presented to the courts, because
it's been through courts time and time again. It might
be fifteen twenty lever arch folders information. And there was
(11:02):
an analyst that worked on the case, Bianchor, and myself
and another detective Jerry Bowden and Jason Evers, and we
were walking. We were the walking talking knowledge of that
investigation only because we had to do an old school
and read the documents and everything, so we retain and
we summarize the information, We analyze the information, and Bianca
did some extraordinary work as the analyst. But with Ai
(11:26):
Fai looked at that there's a very real possibility they
could pick up something despite our best efforts that we
might have missed. Would that be fair the.
Speaker 3 (11:33):
Same, Yeah, definitely fair.
Speaker 2 (11:35):
It may pick up an avenue of inquiry that or
a link that you didn't quite put together, or it
might question a link.
Speaker 1 (11:41):
Yeah.
Speaker 2 (11:42):
It could work in the opposite too, can't it. So
it's a search for the truth. We all know that
around investigations. It could actually do both, could find a
new avenue of acquiry or so actually you went down
the wrong path on that one. Let's go back to
that intersection where you went down there and provide you
with some insight. But it'll process all of those lever
arch folders rapidly and so thoroughly that it could map
(12:02):
it out.
Speaker 1 (12:02):
And there were legal issues and legislation was changed for
that case, and there was high court decisions and all
those things could be broken down into it. And like
we had some of the four and agains, but some
of the best legal minds in the country working on it.
But AI might even take it to the next level.
Speaker 3 (12:21):
It can if it looked at it.
Speaker 2 (12:23):
That's right, and it can, and I think there's a
there's a place for it. Again, coming back to it's
got to be the appetite I guess of the agencies
around the country to want to lean into that and
explore it, because without their assistance, without their access to
these files or these data sets, you can't train the model.
But again, the first point's got to be safe, secure
(12:44):
and for the right purpose. And that's why you know,
we're looking at partnering with the right investors for this country.
Not people just want to make profit. It is about
solving crime and just just even if it solves one
cold case, how good is that for the.
Speaker 1 (12:58):
Fairt Well, you know that you've been there, you've seen
that the pain that if anyone's case hasn't been solved,
or someone's missing and there's no answers. I also think
and I talk William Tyrell case, and that's the case.
It's very close to me and I'm heavily involved in it,
and I think of all the information, we were just
overwhelmed with information because it was a high profile and
(13:21):
when I was running it for the four years that
there was so much information that could have gone through
and I could see something like I'd like AI to
look through it and has anything been missed? And it's
been running now for six years, and I think I'm
as confused as the rest of the public what's happening
on the investigation. But it'd be interesting for.
Speaker 2 (13:41):
AI and it could in another positive there in that case,
In that scenario, and without knowing it to be anywhere
near the little that you have lived, it could take
some of the conscious or unconscious bias, yes, out of
the situation and deal with the facts and case law
and evidence processes, procedures and just apply.
Speaker 1 (13:58):
That's interesting that you say, like the biases, because that's
been part of the conflict in that investigation people, so
it didn't get on board with what I was looking
at and vice versa. That would be good. I would
love an objective opinion on it. I'm calling for a
public inquiry on the teral investigation or the judicial inquiry,
(14:19):
some form of independent review of what's been done on
that case. Yeah, maybe we call for an AI.
Speaker 2 (14:25):
Inquiry, maybe the cysts and another tool to assist and
that investigation again removing any of those biases and going
right back to the facts and the information at the
time that was given, and it can unpack it all
as incredible.
Speaker 1 (14:39):
Is AI capable of removing the biases? If how here's
the brief of evidence, give an objective opinion. I'm thinking,
what would I be typing in the GPT if I
had access to all the evidence that was gathered from
the wim terial investigation. Could you please review this investigation
(14:59):
and give an unbiased opinion or an objective opinion on the.
Speaker 2 (15:04):
Me in accordance with the rules of evidence, facts of law,
and you could even look at it from that criminal
point of view. But you then put that lens of
as an investigator, where are the avenues of inquiry that
we may have missed or where are some choke points
here that we can go back and revisit, et cetera.
So yeah, and it's about training it. But you've got
to have access to some of these older, probably less intense,
(15:25):
cold cases to train your model to think as an investigator.
Speaker 1 (15:29):
I would like to do it. And I'm mainly referring
to those two investigations because I'm very familiar with them.
But I'd like to put all the information from both
those investigations, put it into AI and identify. The question
would be identify the person most likely to have committed
these offenses. That'd be interesting.
Speaker 2 (15:49):
What would exercise? Yeah, yeah, and one hundred percent it
would be And it can do it. It just got
to be able to you know, as I said at
the start, it does cost a lot of money to
build them things. But the money is there in how
has it cost to solve a murder?
Speaker 1 (16:05):
Well, I look at there's always it's always a contentious
environment the cold case units in any state, and I
see it in all states because you've got family members
that are pushing for we want everything done to have
this case solved, and you've got police saying, well we've
looked at this, and yeah, there's pushback. There's always you're
not going to satisfy everyone. Like if one case is unsolved,
(16:26):
there's going to be a family that's unhappy. That's the
nature of it. But yeah, it just gets me thinking.
Quite often the cold case units are restructured, renamed, and
we're going to do this and going to do that.
But wouldn't it be interesting if they said, Okay, we're
going to get put all these cold cases through AI
and see if anything's been missed, because invariably they bring
(16:47):
a fresh team of investigators and you know the old saying,
the fresh set of eyes. Sometimes it works, but we're
all detectives, we're all probably trained in the same line.
The thinking it would be good having that pendance of
AI looking through the briefs.
Speaker 2 (17:02):
Couldn't a more I think that's having the appetite and
the courage to go down that path because it obviously
potentially will create more work for the police.
Speaker 3 (17:10):
But you start to trade.
Speaker 2 (17:12):
Off the benefits of using AI and the administrative aspect
we discussed earlier. You're freeing up some more time to
potentially revisit here, but it's a case by case. Traditionally
the cold cases are brought up because there's a new
piece of information. Well, this is a new tool that
may be able to assist, and I'm all for it,
and there are you know you mentioned chat CHAIRBT, but
I'll give them a plug the main code.
Speaker 3 (17:31):
Guys.
Speaker 2 (17:32):
They've built Matilda, the Australian version that's going to compete
with chat CHAIRBT because it's going to be built on
our bias. Australian culture, Australian knowledge and using the data
sets to build that that are from here, okay, and
we have access to and everyone has ability to input
to that, not using a UK or an American or
an agent.
Speaker 1 (17:53):
So it's based on our environment or relable to it.
So it's what's that.
Speaker 2 (17:58):
Called Matilda Matilda the guys main Code Dave and his
team there have built Matilda. They launched it recently. So
it's exciting Fustralia and we could be global leaders in
this space of AI. We don't need to rely on
what's being done elsewhere.
Speaker 1 (18:11):
Well, it's something that I think we need to embrace.
And as I said, I'm not the first to adopt technology,
but I'm starting to look and think you're going to
get left behind if we don't look at this. But
breaking down what we just talked about, my mind's spinning
in terms of what could be done with cold cases,
and quite often it's just that small piece, that missing
(18:33):
piece of information that with all the good intent of
all the investigators working on it tirelessly, that is missed.
Speaker 2 (18:39):
Yeah, exactly that. What's that old game was mind Sweeper? Yeah,
we just click on that one little square that opens
up open all up. So I think that's part of
As long as we're adhering to and working with the
partners in the real moral, ethical sense, adhering to the
principles that they want, it can be built. But we're
just going to get the right investors with the right
(19:01):
purpose and integrity that we think we share to come
with us on that journey to build it for the
investigators and so will solve the heat.
Speaker 1 (19:09):
And we keep coming back to that the integrity and
the ethics of it that's crucial to it, isn't it
because it can could go off track very quickly if
it's not used properly.
Speaker 2 (19:18):
Yeah, And as you said, you know, you get one
wrong because for whatever reason it may be, well, the
whole thing comes crashing down.
Speaker 1 (19:25):
AI in the courts that we had a magistrate on
a few weeks ago on by catch killers and we're
talking about AI in the court system, and he made
the comment that say sentencing of drink driving charges pc
I prescribe concentration of alcohol for sentencing, you could access
what were the sentences on everyone that's been sentenced for
(19:48):
that defense and then make your determination on what sentence
you should give, whether it's a suspension, of license and
that type of thing. I think it's going to have
to change the face of courts.
Speaker 2 (19:58):
Yeah, I agree, I think it's I have to. You know,
if they're not thinking about it, they should be, but
I'm sure they are, and you know, very smart minds
in there. And what a way to be able to
draw all that information from such a massive amount of
history or data for the judges to have at their
fingertips or barristers to have to suggest to judges around
sentencing parameters. Makes everything more efficient, more effective and actually comparable.
(20:24):
And that's what the whole point is, if we could
have that with an URL judiciary or the.
Speaker 1 (20:29):
I suppose the argument is that you've got to hang
on to the human side of it, the human part
of it. But in my limited experience playing with AI,
it comes back with some very emotional awareness that quite
surprises me. Like if it's things I'm thinking, I didn't
expect a computer to spit that back at me. So, yeah,
(20:52):
you feel all that into court. I don't know what
the role of a solicitor. I know, if I was charged,
and I hope not to be charged again, I'm just
put that record, But if I charged again, like I
worked on my defense and worked with my great legal
team for recording a conversation on the telephone that almost
seems like I did embrace technology leader on it. I
(21:14):
would put it in AI and go, where is the
strength for the case against me? Wants a brief of
evidence being served? What is the strength of the case
against me? And this is my case, and then present
the cagun argument.
Speaker 3 (21:25):
For sure yep.
Speaker 1 (21:26):
So is there a role for solicits anymore?
Speaker 3 (21:29):
Well, do we need less solicitors? We won't go if
we put it.
Speaker 2 (21:34):
Look like we were saying before on my lens through
policing was a certain bias, and that's so changed now
that I've come out on the other side and really
respect and admire the work that I've done by defense
teams because they've got to go through so much information
on these serious cases, and you know, is there an
avenue there to explore or not? Is this the right
(21:55):
person or not? That's our judicial system. But there's definitely
a place for AI. And imagine the hours that are
spent on that. But you know, the research solicitors, the well.
Speaker 1 (22:06):
The poweralegals, legals, so find this case law and all that.
Speaker 2 (22:11):
Yeah, sorry, guys, it might be, it might be changing
for you, but you know it'll shift to another focus.
We saw this with computers and everyone thought, oh, it's
going to be the end of it all. Even's going
to be out of jobs. But we've seen it now
advanced manufacturing. It's just shifted and I dare also use
that word that I hate from COVID pivot.
Speaker 1 (22:31):
You're clearly from down Victoria.
Speaker 2 (22:33):
That's right.
Speaker 1 (22:33):
We heard it every stingy Daniel Andrews every day.
Speaker 2 (22:38):
But yeah, we've seen the advancements and new technology does
bring new challenges and brings new things, but it takes
that to the next step and next generation. Who knows
what we're talking about in twenty years, like we look
now at you know, you look at some images and
videos now and you go, that's a I think in three,
four or five years, you're not about.
Speaker 1 (22:56):
To tell we went well, the young child down in
South Australia that disappeared gus that they were getting a
lot of AI and a lot of media. Different organizations
reached out to me to comment on it. Just based
on my experiences with the Willim Turreal matter and what
I acknowledged with the Willim Tural matter, it was very
difficult because people there were a lot of sightings of
(23:19):
William Tyrell and yeah, you had to sift through it,
you had to follow it up. And we ended up
working out the triarche system on what weight we'd carry
a siding of wim. But then another layer came in
with the gust thing because there was AI generated images
of and yeah, you could work it out relatively easily
that it was AI generated, But as you said, it's
(23:42):
going to get better and that's going to be very
hard for investigators to sift through what's real and what's not.
Speaker 2 (23:48):
Yeah, it's pretty scary, not just the images, but then voice,
the use of people's voice. And you know, we did
do a bit of this on the show Hunted where
we cloned the person and we're able to write the
script and rang their mother as if we were the
fugitive sneak. Yeah, but you know there's a criminal element
(24:09):
that comes in on this too, doesn't it. So you know,
we see a lot of frauds that that happen. Organized
crime internationally will use this technology to try and commit fraud,
which is why I have with my kids. The best
way to try and get around is you have your
code word. Right, So if my kids are overseas, Hey Dad,
I'm stuck on need some money, which is your classic case, son,
what's the code word for the day. Yeah, I won't
(24:31):
know that, so it might sound like my son might
be purporting to be here.
Speaker 1 (24:34):
Because that's your protection. That's a very simple, practical.
Speaker 3 (24:38):
Easy way. But then they've got to remember the code.
But no, that's that's our simple one.
Speaker 2 (24:43):
And I know if I meet the talking to them
on the phone, or you know, they've sent me a
voice message, you know, then fact check what's the what's
the code?
Speaker 1 (24:50):
So with AI and voice copying the voice, there's all
these hours and you're you're the same in the stuff
that you've done in the media, and you've got me
talking here on the podcast. They in effect could copy
that voice and at this point it could be leaving
the message. But I dare say it would get to
the point where they could have a conversation.
Speaker 3 (25:11):
In my voice exactly.
Speaker 1 (25:13):
Yeah.
Speaker 3 (25:14):
Scary stuff, isn't it.
Speaker 1 (25:15):
Especially if they fhm, my mother.
Speaker 2 (25:18):
That's right, And that's that's the vulnerability. We always see
organized crime will keep shifting to where the gap is.
We've seen that in the tobacco industry. Yes, we saw
it through the armed robbery days. You know, when was
the last time we saw a bank robbery. Well that's
because we target harden, didn't We know now that the
real opportunity is online fraud and scams and targeting our
elderly that aren't up with the tech. And there's a
(25:39):
lot of work being done by the banks to educate,
you know, fact check, you know, if that email doesn't
look quite right, and don't do it. So there's a
lot of space. So that's why it's so critical to
have this developed. It's going to be developed. The crooks
are going to use it to law enforcement needs to
be right up there with them, not falling behind.
Speaker 1 (25:58):
Well, we have to because I thing about the crookxy
creative in the way they can exploit the situation and
moving as you said with the tobacco wars or whatever,
if they see an area they can exploit. But it
makes me concerning that even before you'd go to court,
if you had the phato of someone there or something
(26:18):
and there's a hard evidence, we don't know. Sound like
Donald Trump, fake news, fake.
Speaker 3 (26:24):
Image, that's right?
Speaker 1 (26:26):
Where is it? Where's it going to?
Speaker 3 (26:27):
Where does it end.
Speaker 2 (26:28):
Yeah, Yeah, that's a scarity there are you know, there
are ways of trying to find if that is AI
generated or not. And we're getting into a complete area
where which is outside my wheelhouse of how that's done.
But that's part of the reason of building it, building
it with the ethics that it needs to be.
Speaker 1 (26:45):
Getting back to two investigations, let's just put their heads
to how else it could help the fact sheet and
all these things seem relatively little, but in a high turnover,
the high volume crimes, the type of things that the
these police are out doing on a daily basis, and
it might be arrest arresting someone for a break and enter,
(27:06):
and that should in theory free police up if we
embrace what AI can do for us.
Speaker 2 (27:14):
Yeah, I think that's a no brainer. Yeah, it should
free up countless hours. How do you We'll just give
one example of the one homicide detective, then a few
times at by the volume crime as well, which you know,
potentially the crime mapping, the predictive data of where the
hot spots are. We've seen a lot of great work
done in that space in policing over the years. But
there's another argument for another aspect of AI to look
(27:36):
at the crime mapping. You know, where where's all your
cars getting broken into? What do we used to say that,
You know, you'd see a crime spike of those volume crimes.
You go, well, there's a dealer around here somewhere, because
they're not going to hold onto the stolen gear for long.
They want to offload that really quick.
Speaker 1 (27:50):
They'd be someone could be as simple as someone targeting
a shopping center car park. Car has been broken into,
and you know we'd put a car in there, settle
up with clear left unlocked and wait till the person
come to arrest them. But identifying patterns in crime would
be good. Linking crimes is another thing, and we've got
a lot better, I think in law enforcement because there's
(28:12):
more of an exchange of information between jurisdictions at state
police and that. But serial offenders the type of thing
that quite often went unnoticed because there was no link
to the crime that this person had committed in another location.
Speaker 3 (28:30):
Yeah, exactly.
Speaker 2 (28:31):
And I think you just speak to the core of
what I'm hopefully trying to push is that all law
enforcements need to get on board this and once the
border they don't care about the borders. We've seen that
and we've sadly learned the hard lessons of those serial
offenders that will jump jurisdictions because we weren't communicating. We're
so much better now, but we can still improve and
if we're going to do this, I think everyone needs
(28:52):
to get on board because you know, the I guess
the armed robbery or the sex offense that's committed over
in way, maybe the exactly the same or the way
of offending. Then all of a sudden it started up
here in Queensland and there's a real space there as
well to understand that and learn from that and market
as a flag and go, well that needs to be
looked at. Is that possible that that could be the
(29:13):
same offender?
Speaker 1 (29:13):
Okay, interview So I keep coming back to the interview room.
You're playing with my head thinking I could sit here
with interview you for a couple of hours, and then
my phone tells me you've forgot to ask this or
forgot to ask that.
Speaker 3 (29:27):
You could.
Speaker 1 (29:28):
A lot of the preparation for the interview room was
preparing for the interview, and that quite often meant you
had to read through the whole broef and get all
the details so you had it in your head, I
would imagine AI could help with that.
Speaker 2 (29:42):
Yeah, the same question as what we asked before, rather
than the question to the AI is put this together
for the district court or Supreme court? Is with all
the available evidence to me at this stage? Draft mean,
and if you plan based on these offenses and in
line with the points of proof to prove the charge,
you can't.
Speaker 1 (29:59):
Now that I could see the benefits of it. But
I also think, and you said, you need that human oversite.
You need an experienced police officer there listening to how
you answered the questions, because I've seen mistakes people have made,
and I've probably made it myself that you work so
hard to Okay, I got to ask these questions. You've
got twenty questions written out. Good, I've asked that one,
(30:21):
I've asked that one. But you're not listening to the stances.
Speaker 2 (30:23):
Yeah.
Speaker 1 (30:24):
So yeah, it's a balance between the human oversite and
what AI can bring.
Speaker 2 (30:29):
Yeah. I've been caught by that, I've got to say embarrassingly,
because you're so focused on your interview plan and where
you're going. But that's why I would have your corroborator
next to you to hopefully pick up on that. And
when I first started homicide, you know, I'm on my
introduction to the homicide world. I think it was day
two I was there. We'd actually had a fellow in
for interview, you know, and your teams watching through the
(30:49):
other window, listening, and they're all contributing because everyone has
different aspects of the brief, aren't they. So that was
how it was done and here's the next step to that.
So if you you know, you still have people listening,
you still have your boss listening or checking, but you've
got your AI assistant. They're making sure that you've covered
everything off and hopefully validate where you're at.
Speaker 1 (31:10):
Well in what we're talking about. You would think that,
because I'm not sure if we spoke on camera or
off camera, we're certainly talking about that the struggles that
most police forces in Australia are getting recruiting people at
the moment. So the police that we've got, we've got
to use them smart, use them most efficiently. And if
(31:31):
law enforcement embrace this, one would think that it would
free up police from a lot of the work behind
the desk.
Speaker 2 (31:38):
Yeah, I definitely feel that, and I feel that it's
about trying to identify what's going to be your big
spang for buck, and right now I think that's the
administration piece. My heart goes to helping solve crime. Obviously
that's what I want. But the biggest piece I would
think here, I guess for the senior police is getting
rid of the administrative tasks that are day to day.
(31:59):
If you LANs your brief preparation, your fact checking, your rosters,
all those administrative tasks that could be taken away and
all of a sudden, you know, your analysts hours a cutdown,
investigators hour as a cutdown. We're trying to work through
some data now on actually putting a monetary value on
how much that is to be able to sell. This
is why we need to invest in this, and this
(32:20):
is why we need you to come on this journey
with us.
Speaker 1 (32:22):
That would be interesting. And I think with an analyst,
and I've worked with some great analysts over the years,
and I think how they would readily adapt to using AI,
like as in not making them obsolete by using it.
It just frightens me the amount of information they'd be
able to spit out so it doesn't have to be
the dumb lead investigator. The analysts and the analyst is
(32:48):
embracing it.
Speaker 2 (32:49):
Yeah, another tool for them to use, and they're forward
facing to the investigations that matter and using their skills
that matter, not just a monthly reporting. It's an important
part of the piece for the big strategic vision. But
again that doesn't need to be done by a human
can be fact checked and looked at. But if you
train it appropriately and it's all there for everyone to
(33:10):
see live. We've seen all the advancements in body worn
cameras and well that it's all there. It just needs
to come together.
Speaker 1 (33:17):
Well, people might think this is an exaggeration, but I
think you might be able to corroborate me here. I
retired as a detective chief inspector. I was an inspector
for the last fifteen years, or a detective inspector in
the police. I reckon thirty to forty percent of my
time spent when I'm leading investigations, leading all different murder investigations,
(33:41):
thirty to forty percent of my time with been administration
and menial administration. I would I'd cringe saying it, but
it was the reality that once a month I'd be
spending two or three hours adding up card diries making
sure the car diaries calculated. And if I missed doing
the car dory, that was probably worse than not solving
(34:04):
the crime. I should submit the car dories on time.
They had to check all the people under me, so
things like that that should just be by the wayside.
Speaker 3 (34:13):
Yeah, again, no brainer.
Speaker 2 (34:14):
Yeah, make it happen for everyone up and you know
there was a joke about it, but there is that
stress of the ADMIN that comes with it as the
team leader as well. The monthly times come around and
you know you need to make more time for your admin.
Need to sit down do this not Well, we're just
going from job to job to job and if we
need that now you need to make these red lights green.
Speaker 1 (34:35):
I don't think I'm unique. I don't think you speak
to any operational police officers and the frustration I used to,
in fact joke sometimes say about one o'clock, okay, admin done,
let's start investigating this murder. But it was quite often
we're all office bound because we're catching up on record
keeping administrators.
Speaker 2 (34:54):
Yeah, it's definitely got its place. Yeah, look, it's a
no brainer for me. But we're just we're really close
to I think getting on that right path it's got
to have some courage to do it. And there's been
a lot of changes in senior management and chief commissions
around the country, so you know there's time to come
in here and lean on.
Speaker 1 (35:13):
I don't blame anyone that might have fallen behind or
not realized because it is changing so rapidly. We wouldn't
be having this conversation two years ago. It would be
completely different. We'd be talking like it's a sci fi movie,
but now it's reality. So we talk about saving time.
But the thing that excites me most about is how
(35:33):
it can enhance investigations.
Speaker 3 (35:35):
Yeah, definitely.
Speaker 1 (35:36):
It just seems like a no brain of the human factor,
the sharp minded analysts that you'd have on an investigation
and the detectives pouring over briefs that can potentially be done.
Speaker 2 (35:49):
So much quicker, and imagine what we could do preventing
crime and the.
Speaker 3 (35:56):
Unseeing cost of prevention.
Speaker 2 (35:58):
You know that it was always that movement towards the
end of my career was let's try and stop the
crime before it even happens.
Speaker 1 (36:03):
Well, that's an interesting point too, and I haven't covered
and that's something that I wanted to wanted to cover,
is that we've been focusing on solving crimes, but how
can AI be used to prevent crimes? Give us some
examples there.
Speaker 2 (36:19):
Yeah, well, I think the clearest example is then predicting
where the hot spots are, And we see a lot
of work with that at the moment, but that's generally
data predicting where these hot spots are. The AI aspect
might come into that potentially, And where I see it
more so is you know you'd see two, three, four,
five street robberies or that's one example where if we
(36:42):
can identify this offender after the first one or the
second one, we know they'll generally keep going because either
their motivation is either greed or drug induced, camera or whatever.
If we can identify that person on day one or two,
how many offenses have we stopped by being committed by
getting that person here not down the road. That's one
aspect to it. There are some other discussions I'm having
(37:04):
with with other developers around the human behavior models. So
you know we've all seen you know, the kids or
anyone just before you know, you've watched the footage and go, okay,
I can see they're starting to get a thousand yards there.
They're doing this and it's about to be on. So
you can see the human behaviors through footage that something's
about to happen.
Speaker 1 (37:24):
Right.
Speaker 2 (37:25):
You know, when we've investigated either you know, your machety
attacks or the like that we see a shopping centers
these days, there's a lead up to that. It just
does not all of a sudden explode as a general leader.
So we can train AI to look at those human
behaviors and what are the signals here, what's happening here.
Speaker 3 (37:40):
So we're getting into a.
Speaker 2 (37:41):
Whole other area now of non policing. But you know,
I know private security is very interested in those aspects.
Speaker 1 (37:47):
And that's interesting, and we'll use it just by way
of an example, the attacks in the shopping centers because
it's been in the media a bit. But yes, it
would be a gathering.
Speaker 3 (37:57):
Yeah, I'm just featuring like a thing.
Speaker 1 (38:00):
Spirits investigator might be able to pick that up, but
AI might be able to do it even better.
Speaker 2 (38:05):
And we're trying to monitor I don't want to name
particular shopping centers, but if you're trying to monitor hundreds
and hundreds of cameras, yeah, or big public events, the
AI can do that, and it can alert the person
that's there in charge of looking at those cameras. We're
seeing something here on camera forty six and forty seven
of a gathering of some it could.
Speaker 1 (38:22):
Be let's break it down. That could be put if
there's one hundred cameras around the shopping center and AI
had put in place that if you see people dressed
in this style of more than four gather at a location,
alert that saw draw correcting. We want to know and yeah.
Speaker 2 (38:42):
That operator and then look at that and go oh,
that's nothing, or well, we've got a problem here and
you're onto it twenty thirty seconds a minute beforehand. Hopefully
you can get there and disperse them before anything even happens,
so you prevented the crime. That's always So there's an
aspect of human behavior. So you biomechanics that you need
to train with data sets, so you need access to
all these different ones that have occurred in the past,
(39:04):
and the data is there. We've got a partner with
the right people to build that data and train the
AI to Okay, that's all we're looking for, or weapons identification.
I know there's a great project that's been done through
one of the people I mentioned before. Around identifying firearms,
similar to what I said about vehicles. Yeah, so you
know what gun you're looking for or exactly which one
it is, and.
Speaker 1 (39:25):
What you're talking here, Just so I understand and hopefully
other people understand, Like, identify a gun. If I get
out of the glock, I can tell it's a glock,
but it might have a scratch on the barrel or
scratch on the pistol grip. It might be something like that.
But AI would pick up on that and go, this weapon,
(39:45):
there's the same weapon that was used in this shooting.
Speaker 2 (39:48):
Yeah, if it's all talking holistically to each other and
you've got good footage, so yes, you've got good footage.
But even post offense, if you've got your suspects arrested
with is you know, with twenty two or a glock
or whatever the case may be, that's that particular weapon.
All of a sudden, you're plugging into all these unsolved
ballistic reports of this actually was a clock that was
(40:10):
the weapon that killed this person in this murder, because
we've got the ejected casing, et cetera. All of a sudden,
you're linking all these cases again cross border stuff that
is volumes and volumes of forensic material that it can
do in seconds if you're focusing on the right problem.
What is the problem we're trying to solve here? So
weapons identification is one similar to vehicles, they're definitive. There
(40:34):
is an exhaustive list. There's only so many makes and
models of firearms on the planet, so you can teach
it too to come up with those options of this
is what we definitely think it is, or this is
potentially what the type of weapon is. And again then
you look at your execution of search rights. We know
we're looking for a particular maker, model firearm as opposed
(40:54):
to the guesswork. So there's a lot there. I'm sort
of overwhelming the conversation with a different aspects, but I
think it really comes back to that initial piece of
what is the issue that we're trying to use AIEN
to solve? Is it an administration? Is it prevention of crime?
Is it cold cases?
Speaker 1 (41:13):
And there's all these different things that could enhance.
Speaker 3 (41:16):
Or assist working with it.
Speaker 1 (41:19):
Hey guys, it's Gary jubilin here. Want they get more
out of I Catch Killers? Then you should head over
to our new video feed on Spotify, where you can
watch every episode of I Catch Killers. Just search for
I Catch Killers video in your Spotify app and start watching.
Today you said about overwhelmed with information. It made me
(41:41):
think while you were talking about how we were overwhelmed
with the availability of information, that when I first started
as a homicide detective, there was a couple of people
had mobile phones. You could find out or listen to
their mobile phones, or they had landlines and you could
track them that way, and that was you know, that
that was the line of inquiry. A bank might have
(42:04):
a CCTV camera out the front, but only a bank
or a licensed premises. And I did actually see towards
the latter part of my career we were getting overwhelmed
with information, like there was so someone's murdered. You'd get
called out. And the instructions I was given in the
early days as a young homicide detective go out and
(42:26):
collect all the CCTV footage from businesses and you'd come back, Oh,
that service station had one, and you'd be all proud
of yourself, and yeah, there was one one at the bank.
Now you say that private homes cars have got you know.
And then as my career progressed, I'm seeing at the top,
and I'm looking at all the information we've got and
I'm going I'd be pulling my hair out if I
(42:47):
had had hair, just overwhelmed with the information. And AI
swings the pendulum back in favor of being able to
handle all that information, doesn't it, because we did get
to the point where we were able well.
Speaker 3 (43:00):
Well, information intelligence is key, isn't it.
Speaker 1 (43:02):
Yeah.
Speaker 2 (43:02):
So the sooner you can get that in investigator's hands
and understand it, yeah, the better you position for your strategy.
So all that information that you're getting from CCTV, from
that murder, getting it all together, if you miss that
bit of footage that shows it turned left onto the freeway,
then you know that's a complete game change, isn't it.
So you won't miss that. And whether you're plugging in
one hundred cameras worth the footage into this AI, it'll
(43:26):
track that you tell it. I want you to track
that car. That's the one we want out of all
this information. We've got timeline and tell me where it's
been and what it's done and bang.
Speaker 1 (43:34):
And what people didn't When I say people, the public
if you're not involved directly in it. When we've got
listening devices in houses, you put a listening device in here,
and you've got someone monitoring it. They can even monitor
it live and if something happens or invariably you've got
to come back and review it because it's not been
(43:54):
monitored live. So you've got twenty four hours of listening. Yes,
like people going, have you gone gone through that? Well,
it's basically in real time. And then when you hear
the conversation, then you've got to document the conversation.
Speaker 2 (44:08):
It was so time, and then they're more blow our
minds with then your hat a layer of you know,
if you need a translator, Yeah, so yeah, I get
one hundred percent your point. And that's that's those complex
pieces of information and large amount of data that it
can deal with. And that definition of AI you said
right off the top, that's what it's doing. It's looking
(44:29):
at all of those conversations and if you're after key
words or if you're after a key you want them
talking about the victim of a planned murder. You know,
you've got a contract, kill job, whatever you want to
know whenever that pops up.
Speaker 1 (44:43):
And we would get to the point where I've used yeah,
limited times when I have people look at the terminology
used by by an offender from a psychological or suspect
from a psychological point of view. You could even feed
that into yeah, couldn't you. It appears that and he's
talking about that Issue's demeanor or his language is changing.
(45:04):
Little things like that, and it's only incremental victories, but
they're the little things that make the difference.
Speaker 2 (45:11):
And gives the advantage back to the police. Yeah, there's
one other piece that I will talk to and it
can be quite confronting, so I don't want to trigger
anyone on here, but a really I think crucial piece
of where AI can help as well as in the
child exportation material space, you know you would a basic example,
police do a search one on a laptop and you've
gond a laptop and you've got hundreds or thousands of
(45:33):
really horrific images of child exportation material. Some of that
you may be concerned is contact defending. So you're trying
to identify this kid that's clearly in the footage or
in the images with you're accused. You need to find
out who this kid is to stop the contact defending.
So then you're madly trying to figure out how how
is this possible? Where is that you're looking at the
metadata of the images, you're looking at all this sort
(45:56):
of stuff. The AI should be able to work out that,
and it may be to use facial recognition technology of
the child to try and identify who is this child.
So that's the piece. I feel that that's where the
trigger for scraping social media is important, because that kid's
photo might be in the background at someone's birthday or
at a any event, and all of a sudden, you've
(46:17):
got a match on this kid that's being offended against,
and you can immediately stop that offending through the use
of the AI to identify that kid. There is some
work being done in that space by the AFP and
monash Uni, and that's fantastic, but you know, it needs
more support, it needs more funding, and it needs more
attention to help protect these kids because it's crucial because
there's nothing worse than investigator. I can't find this kid
(46:39):
and we know they're being offended against, and you can't
release it to the media. You know that you've got
to protect the identity of these kids.
Speaker 1 (46:44):
That seems like a no brainer that if I can
help in that space, and I understand what you're saying
the kid that's been sexually abused and you've got to
identify that child. I could help.
Speaker 2 (46:57):
That's why I want to write. Partner with the right
investors that have the same morals and purpose that we
have at Perigrin to get this done because we can
protect kids.
Speaker 1 (47:06):
We touched just briefly. Fiscally, this seems like a no
brainer for law enforcement. Yeah, freeing up, freeing up more success,
reducing crime. There seems to be benefits all over.
Speaker 3 (47:20):
Yeah, I think so.
Speaker 2 (47:21):
But it's also challenging times, isn't it. And I can
only talk in the Victorian context because that's what I
saw when they were seeking a new Chief commissioner. You
to have to save ordered by the government to save
five hundred million per year in your budget, so you know,
good luck, chief, but better you than me. But that's
the first step off point. So you're then coming along
(47:43):
or a company like Peramn's coming along, going we want
you to partner with us because we want to invest
in this. You really can't expect the government to be
stepping or the police to be stepping up in their
budget to assist. Hence what we need corporate support. But
it is a no brainer but eventually they'll get there
to see the benefits. We've saved so many hours in
investigators time and administrative tasks. But right now it'd be
(48:06):
very challenging and for them to commit.
Speaker 1 (48:07):
I hope they see it that way because, as we said,
sometimes police, when technology changes, we drag our Yeah, I'm
sure there's not many detectives that have to g to
a private company to get something that the police. You
would assume police had capabilities of but didn't.
Speaker 2 (48:25):
Well, like I said, I was at a conference eighteen
months two years ago and it was around the child
expectation and very senior police and I won't name the organization.
We're there going when is corporate? When is private sector
going to step up and help us? Went, oh, let's
have a conversation FORFE So well yeah, so yeah, it's close.
We just need to keep it going, keep the conversation going,
(48:46):
and keep developing it. But look, coming back to it,
I just thought of when were speaking before around the
use of AI and cameras and that now it's actually present.
Now I've got your cameras around my place and they're
they're giving a plug to the UFI system, but they've
got AI in there, So you can now load someone's
face onto your home CCTV system and it'll alert me
when that person is there, whether I want them there
(49:08):
or not. So the AI is already plugged into some
of these things that are there, so you know, as
you said, it's coming, it's here.
Speaker 1 (49:14):
Well that that could you just made me think we'll
go into an area, very real area that like domestic
violence and then that type of thing. But you could
forewarned if someone's someone's turning up.
Speaker 2 (49:27):
So that's that's that natural evolution of what used to
just be, you know, an old school c CDV camera.
They've got chunky ones on there that are now these
small little ones that are either powered by solo or battery,
and they've got AI into them that can alert you
to know whether you do want them or don't want
them there, can give you an alert to your phone,
and you can be anywhere in the world. I think
I was lucky enough to be overseas not too long ago,
(49:47):
and I'm getting alerts from my camera back home. Oh
that's thank you. That's the person turn up to feed
my turtle. Thanks mate, Yeah, fair cool turtles.
Speaker 3 (49:57):
Turtles, turtle's got the weight.
Speaker 1 (49:59):
Well, look I really enjoyed the chat, and thanks for
coming on. You're the right person, because I know that
someone that brings me crashing into this brave, brave new world.
But the way it's evolved, you know, I might have
to have you back in six months time, and it
might even be you. It might be a hologram or.
Speaker 3 (50:15):
Something I might remote in. Virtually, Yeah, yeah.
Speaker 1 (50:19):
I think if you've given you've given me an understanding.
I hope the listener is an understanding of what's out there.
Speaker 2 (50:25):
I hope so, and I hope that point of just
looking a bit further about what I can do. It's
not just a one shop or one scenario fits all.
It's what's the problem, what is it? And to have
some courage to lean into it. But good luck trying
to sleep tonight. Carry I've blown your mind.
Speaker 1 (50:42):
I'm down the rabbit hole. Now I'm gone. But we
finish off with because I do think that's important, because
we know how power can corrupt in police and that
the ethics and the integrity that needs with this type
of type of.
Speaker 2 (50:58):
Thing part of part of our lives absolutely critical, which
is why again I've said countless times need to be
built here for here.
Speaker 1 (51:06):
Just hold on for a second. I'm just going to
put in chat GPT. How I say goodbye to Graham.
I love it, New Graham for coming on I catch Killers.
It's been a very entertaining discussion. Thanks Hasburt Chees
The public has had a long held fascination with detectives.
Detective see aside of life the average person is never
exposed to. I spent thirty four years as a cop.
For twenty five of those years I was catching killers.
That's what I did for a living. I was a
homicide detective. I'm no longer just interviewing bad guys. Instead,
I'm taking the public into the world in which I operated.
(00:23):
The guests I talk to each week have amazing stories
from all sides of the law. The interviews are raw
and honest, just like the people I talk to. Some
of the content and language might be confronting. That's because
no one who comes into contact with crime is left unchanged.
Join me now as I take you into this world.
(00:46):
In part two of my chat, we've retired Detective Graham Simfendorfer.
We dive further into the world of AI and how
it's going to change the world of crime. It's quite
mind blowing where we're headed. My sense is, if pleasing
in investigations, the life of a criminal is going to
get a lot tougher. But maybe I'm looking at things simplistically,
(01:07):
because no doubt the crooks are also going to embrace
AI as well. That may counterbalance the advantages law enforcement have.
We also had a close look at how AI may
help solve cold cases and how it could be used
in the court system. It's a really interesting chat that
I learned a lot from Graham. Welcome back to part
(01:27):
two of Eye Catch Killers. I'm scared AI is taking
over the world. Yeah, I want to jump straight into
cold cases because homicide was my passion in my career,
and I know you worked homicide and people are just
fascinated with cold cases and the thing with homicide until
they're solved. It doesn't matter if they're fifty years old.
(01:49):
People want results when someone's life has been taken. How
can we use AI in the cold case space?
Speaker 2 (01:57):
Yeah, great question, and some significant pieces can play a
really important part in this. So there's the aspect of
reviewing all the old information, laying a lens across what
we know, open source around that information as well. So
it's large data sets of information that, like you said,
to get your head around a cold case, you literally
sit with a box for a month.
Speaker 3 (02:18):
Or so trying to understand it all.
Speaker 2 (02:19):
If you can put all that data into AI, it'll
pull out for you what the key issues are where
the gaps are perhaps some avenues of inquiry. Even so, again,
like we said before, you don't want to lose the
skills as an investigator, but it can process that large
set of information really quickly and accurately with what's been
given in those statements and may even be able to say, hey,
look we need to go back further because we've seen
(02:41):
with cold case is the way statements were taken ten
years ago, twenty years ago, very almost almost leading, if
you will, because you're trying to cover your points of
proof and not that whole story that you get. So
one of those aspects is going back if you can
to speak to some of those witnesses and the AI can,
So let's go back and explore this part a bit more,
or the person that witnessed that vehicle drive pilet's go
(03:04):
back and see that. There's that aspect of understanding the
investigation and I guess the investigation plan. But then there's
also review of either CCTV or face images, and that's
where the facial recognition technology could come in and play
a part.
Speaker 1 (03:18):
Right, So the cold cases, okay, and when you said
you know, the AI might say, have you thought of
this or line of inquiry, not taking that personal as
an investigator, because what you're dealing with AI, they're drawing
on the experience of all the in theory at its best,
(03:39):
all the world's greatest investigators looking at this case, all
the information that they've learned from past things. Because this
am I describing it correctly. Basically, they're drawing on all
the information.
Speaker 3 (03:51):
That's right.
Speaker 2 (03:52):
You're taking every into account and thinking outside the box
if it needs to, and that's the skills of great
investigator processing that, and you can teach it to think
about the laws of evidence and making sure that whatever
strategy or tactic that it's thinking of is complying with
the law and complying with the rules of evidence. So
it'll do all that for you. But yeah, you're right,
(04:13):
it's taking the best of everything and applying it to
your case. So for cold cases, it's worth a review.
It's worth looking at a ability to either you know,
fix fix some footage or make the CCTV less grainy,
or looking at a face image that was previously no
one knew who this was or couldn't be solved. And
(04:34):
that's where this journey for me kicked off. Actually was
looking at. A friend of mine showed me a documentary
series that I think was on Netflix about the there
was nineteen seventy nine Lunar Park fire.
Speaker 3 (04:47):
Oh yeah, I remember that.
Speaker 1 (04:48):
There seven people died in the on the Gates train.
Speaker 3 (04:52):
Is that one?
Speaker 2 (04:54):
And I watched it. Got a degree in fire investigation,
So the scientific brain of me was very intrigued by
by that aspect. But then as the documentary went on,
it was Caro Meldrem Hannah that did that documentary, an
amazing documentary by the way, And towards the end, there's
this face image of who they believed lit the fire,
and there was allegations of being linked to by keys
(05:16):
all the like, but there was this face image, and
I just could not understand how in this day and
age we don't have technology that could turn that face
image from just the sketch image that we're used to
seeing into a actual person and then use the data
sets and AI two search for that person through whether
criminal networks or you know, there's some ethical considerations around
(05:37):
searching against open source, but how can we not have that?
And my rest that sent me on the journey of
researching who surely someone could be doing this and my
investigations globally is no one.
Speaker 3 (05:49):
No one is.
Speaker 2 (05:50):
So we started that journey with program AI around turning
a cold case composite sketch image of the old hand
drawn ones, using AI to turn that into a three
D verse of a person, and then you've got a
much better identity of who that person could be searching
against either open source or criminal databases for an identity,
and you need to potentially age that person. Obviously, if
(06:12):
you're talking about a cold case now we've all seen
it on your Instagram reels, is what are you going
to look like when you're seventy The AI will aid you,
and it can aid you, but again it's only as
good as what's put into it. So you're relying on
the witnesses version of what they remembered, and we all
know we would see you know, this is a seventy
percent likeness or this is an eighty percent likeness. There
(06:32):
I can turn that seventy percent likeness into something and
use it as a tool to search the sources, but
it's still the investigator's job to actually is that the
person or not?
Speaker 1 (06:42):
Okay, And so there's a responsibility of the type of
information you put in And you've said with creating that
or the image and whether you put it into open
source when you're talking open source, you put in out
there on social media everywhere.
Speaker 3 (06:57):
Yeah, that's right.
Speaker 2 (06:58):
Yeah, so open sources, open source intelligences is the web. Yeah, Instagram, Facebook.
Speaker 1 (07:05):
So Google put put out there anyone.
Speaker 2 (07:07):
Yeah, And there symmetical considerations because you know, when you
upload your photo to Facebook, you're not really giving permission
for that to be used or scraped is the term.
You're not you're not giving anyone authorized to scrape all that.
So symmetical considerations that we looked at. But that's the
stuff we need.
Speaker 3 (07:22):
To work through.
Speaker 2 (07:23):
If we're looking for a murderer or a rapist, are
the laws going to be sufficient to help support law
enforcement to identify that person?
Speaker 1 (07:31):
Well, it should be like we've had to as technology
evolves and the powers need you need the authority and
it might be for listening devices from the Supreme Court,
that type of thing to get permissioned on the ghost
train one, did that evolve into anything that what you're
looking at.
Speaker 2 (07:50):
Still a work in progress. So we're still using that
as a test case too. It's a great test case actually,
to turn that image into a person, but also age
that person and what they might look like years ago.
Because seventy nine there was no Facebook. You need to
aid that person to see if that person is actually
out there, either on a license photo or a passport photo.
So law enforcement has access to all that already. So
(08:11):
that's where we're at, and we're just trying to train
that so to train the AI to do it. I've
had some friends sketch myself and then running across open source,
and it's showing some really positive results.
Speaker 1 (08:21):
We've got positive in terms of the way you're aging.
Speaker 2 (08:24):
That's right, not like a fine wine though positive In yeah,
it's around ninety six to nine. Anything over ninety six
ninety seven percent is showing as accurate okays speak.
Speaker 1 (08:36):
If we break it down in percentages, that's a good
percentage to work with, good.
Speaker 3 (08:40):
Tool, isn't it.
Speaker 2 (08:41):
So to either rule that person in or rout, you
might get a hit on this likeness is like this person,
but then your investigation says, actually that can't be that person.
That person left the country two weeks before the offense.
Speaker 3 (08:52):
Okay.
Speaker 1 (08:53):
Another thing that comes in so with AI, like we
used to get when I first started place. Are the
big penries. You know, someone during the sketch, this is
what Graham looks like on that, And there was some
good and there was Subjectivity came into it when people
are doing the sketches. But AI, when it's doing, they're
dragging all that information, isn't it. Where there's been failings,
(09:17):
where there's been successes. That's the type of access to
the material we're having.
Speaker 2 (09:23):
And that's how you want it to learn. So you
don't want it with a bias towards a certain culture
or a certain a certain inference. I guess you want
it to be accurate to what it is you're looking for.
So your sketch that's only seventy percent accurate of the
person is only going to get seventy percent accuracy on
what you're searching for using the AI. So you've still
it to use investigative brain to think about who it is,
(09:45):
et cetera. But it'll do it so quickly, and I'll
search your license or your your prison photos that are
taken time and time again, I'll do it in seconds.
Speaker 1 (09:56):
This is hypothetical, and it's hypothetical, and it's talking about
actual investigations, but it's just got my mind ticking over
the barrable. I call it cyril killing of three Aboriginal children,
Colin Walker, Evelyn green Up and Clinton Speedy back in
nineteen ninety one, over a five month period, they disappeared
(10:17):
and their bodies were found. That case has been worked
on for over thirty years. I became involved in the investigation,
I think five or six years after the children disappeared,
for a reinvestigation, and we revisited like you do on
a cold case, revisited, and the big argument initially was
were these matters linked? And we went through tendency and
(10:40):
coincidence evidence and strongly show that the likelihood they have
been linked. I'm just thinking of what you're talking about.
With all that and my memory, there's probably about the
lever arch folders sometimes I presented to the courts, because
it's been through courts time and time again. It might
be fifteen twenty lever arch folders information. And there was
(11:02):
an analyst that worked on the case, Bianchor, and myself
and another detective Jerry Bowden and Jason Evers, and we
were walking. We were the walking talking knowledge of that
investigation only because we had to do an old school
and read the documents and everything, so we retain and
we summarize the information, We analyze the information, and Bianca
did some extraordinary work as the analyst. But with Ai
(11:26):
Fai looked at that there's a very real possibility they
could pick up something despite our best efforts that we
might have missed. Would that be fair the.
Speaker 3 (11:33):
Same, Yeah, definitely fair.
Speaker 2 (11:35):
It may pick up an avenue of inquiry that or
a link that you didn't quite put together, or it
might question a link.
Speaker 1 (11:41):
Yeah.
Speaker 2 (11:42):
It could work in the opposite too, can't it. So
it's a search for the truth. We all know that
around investigations. It could actually do both, could find a
new avenue of acquiry or so actually you went down
the wrong path on that one. Let's go back to
that intersection where you went down there and provide you
with some insight. But it'll process all of those lever
arch folders rapidly and so thoroughly that it could map
(12:02):
it out.
Speaker 1 (12:02):
And there were legal issues and legislation was changed for
that case, and there was high court decisions and all
those things could be broken down into it. And like
we had some of the four and agains, but some
of the best legal minds in the country working on it.
But AI might even take it to the next level.
Speaker 3 (12:21):
It can if it looked at it.
Speaker 2 (12:23):
That's right, and it can, and I think there's a
there's a place for it. Again, coming back to it's
got to be the appetite I guess of the agencies
around the country to want to lean into that and
explore it, because without their assistance, without their access to
these files or these data sets, you can't train the model.
But again, the first point's got to be safe, secure
(12:44):
and for the right purpose. And that's why you know,
we're looking at partnering with the right investors for this country.
Not people just want to make profit. It is about
solving crime and just just even if it solves one
cold case, how good is that for the.
Speaker 1 (12:58):
Fairt Well, you know that you've been there, you've seen
that the pain that if anyone's case hasn't been solved,
or someone's missing and there's no answers. I also think
and I talk William Tyrell case, and that's the case.
It's very close to me and I'm heavily involved in it,
and I think of all the information, we were just
overwhelmed with information because it was a high profile and
(13:21):
when I was running it for the four years that
there was so much information that could have gone through
and I could see something like I'd like AI to
look through it and has anything been missed? And it's
been running now for six years, and I think I'm
as confused as the rest of the public what's happening
on the investigation. But it'd be interesting for.
Speaker 2 (13:41):
AI and it could in another positive there in that case,
In that scenario, and without knowing it to be anywhere
near the little that you have lived, it could take
some of the conscious or unconscious bias, yes, out of
the situation and deal with the facts and case law
and evidence processes, procedures and just apply.
Speaker 1 (13:58):
That's interesting that you say, like the biases, because that's
been part of the conflict in that investigation people, so
it didn't get on board with what I was looking
at and vice versa. That would be good. I would
love an objective opinion on it. I'm calling for a
public inquiry on the teral investigation or the judicial inquiry,
(14:19):
some form of independent review of what's been done on
that case. Yeah, maybe we call for an AI.
Speaker 2 (14:25):
Inquiry, maybe the cysts and another tool to assist and
that investigation again removing any of those biases and going
right back to the facts and the information at the
time that was given, and it can unpack it all
as incredible.
Speaker 1 (14:39):
Is AI capable of removing the biases? If how here's
the brief of evidence, give an objective opinion. I'm thinking,
what would I be typing in the GPT if I
had access to all the evidence that was gathered from
the wim terial investigation. Could you please review this investigation
(14:59):
and give an unbiased opinion or an objective opinion on the.
Speaker 2 (15:04):
Me in accordance with the rules of evidence, facts of law,
and you could even look at it from that criminal
point of view. But you then put that lens of
as an investigator, where are the avenues of inquiry that
we may have missed or where are some choke points
here that we can go back and revisit, et cetera.
So yeah, and it's about training it. But you've got
to have access to some of these older, probably less intense,
(15:25):
cold cases to train your model to think as an investigator.
Speaker 1 (15:29):
I would like to do it. And I'm mainly referring
to those two investigations because I'm very familiar with them.
But I'd like to put all the information from both
those investigations, put it into AI and identify. The question
would be identify the person most likely to have committed
these offenses. That'd be interesting.
Speaker 2 (15:49):
What would exercise? Yeah, yeah, and one hundred percent it
would be And it can do it. It just got
to be able to you know, as I said at
the start, it does cost a lot of money to
build them things. But the money is there in how
has it cost to solve a murder?
Speaker 1 (16:05):
Well, I look at there's always it's always a contentious
environment the cold case units in any state, and I
see it in all states because you've got family members
that are pushing for we want everything done to have
this case solved, and you've got police saying, well we've
looked at this, and yeah, there's pushback. There's always you're
not going to satisfy everyone. Like if one case is unsolved,
(16:26):
there's going to be a family that's unhappy. That's the
nature of it. But yeah, it just gets me thinking.
Quite often the cold case units are restructured, renamed, and
we're going to do this and going to do that.
But wouldn't it be interesting if they said, Okay, we're
going to get put all these cold cases through AI
and see if anything's been missed, because invariably they bring
(16:47):
a fresh team of investigators and you know the old saying,
the fresh set of eyes. Sometimes it works, but we're
all detectives, we're all probably trained in the same line.
The thinking it would be good having that pendance of
AI looking through the briefs.
Speaker 2 (17:02):
Couldn't a more I think that's having the appetite and
the courage to go down that path because it obviously
potentially will create more work for the police.
Speaker 3 (17:10):
But you start to trade.
Speaker 2 (17:12):
Off the benefits of using AI and the administrative aspect
we discussed earlier. You're freeing up some more time to
potentially revisit here, but it's a case by case. Traditionally
the cold cases are brought up because there's a new
piece of information. Well, this is a new tool that
may be able to assist, and I'm all for it,
and there are you know you mentioned chat CHAIRBT, but
I'll give them a plug the main code.
Speaker 3 (17:31):
Guys.
Speaker 2 (17:32):
They've built Matilda, the Australian version that's going to compete
with chat CHAIRBT because it's going to be built on
our bias. Australian culture, Australian knowledge and using the data
sets to build that that are from here, okay, and
we have access to and everyone has ability to input
to that, not using a UK or an American or
an agent.
Speaker 1 (17:53):
So it's based on our environment or relable to it.
So it's what's that.
Speaker 2 (17:58):
Called Matilda Matilda the guys main Code Dave and his
team there have built Matilda. They launched it recently. So
it's exciting Fustralia and we could be global leaders in
this space of AI. We don't need to rely on
what's being done elsewhere.
Speaker 1 (18:11):
Well, it's something that I think we need to embrace.
And as I said, I'm not the first to adopt technology,
but I'm starting to look and think you're going to
get left behind if we don't look at this. But
breaking down what we just talked about, my mind's spinning
in terms of what could be done with cold cases,
and quite often it's just that small piece, that missing
(18:33):
piece of information that with all the good intent of
all the investigators working on it tirelessly, that is missed.
Speaker 2 (18:39):
Yeah, exactly that. What's that old game was mind Sweeper? Yeah,
we just click on that one little square that opens
up open all up. So I think that's part of
As long as we're adhering to and working with the
partners in the real moral, ethical sense, adhering to the
principles that they want, it can be built. But we're
just going to get the right investors with the right
(19:01):
purpose and integrity that we think we share to come
with us on that journey to build it for the
investigators and so will solve the heat.
Speaker 1 (19:09):
And we keep coming back to that the integrity and
the ethics of it that's crucial to it, isn't it
because it can could go off track very quickly if
it's not used properly.
Speaker 2 (19:18):
Yeah, And as you said, you know, you get one
wrong because for whatever reason it may be, well, the
whole thing comes crashing down.
Speaker 1 (19:25):
AI in the courts that we had a magistrate on
a few weeks ago on by catch killers and we're
talking about AI in the court system, and he made
the comment that say sentencing of drink driving charges pc
I prescribe concentration of alcohol for sentencing, you could access
what were the sentences on everyone that's been sentenced for
(19:48):
that defense and then make your determination on what sentence
you should give, whether it's a suspension, of license and
that type of thing. I think it's going to have
to change the face of courts.
Speaker 2 (19:58):
Yeah, I agree, I think it's I have to. You know,
if they're not thinking about it, they should be, but
I'm sure they are, and you know, very smart minds
in there. And what a way to be able to
draw all that information from such a massive amount of
history or data for the judges to have at their
fingertips or barristers to have to suggest to judges around
sentencing parameters. Makes everything more efficient, more effective and actually comparable.
(20:24):
And that's what the whole point is, if we could
have that with an URL judiciary or the.
Speaker 1 (20:29):
I suppose the argument is that you've got to hang
on to the human side of it, the human part
of it. But in my limited experience playing with AI,
it comes back with some very emotional awareness that quite
surprises me. Like if it's things I'm thinking, I didn't
expect a computer to spit that back at me. So, yeah,
(20:52):
you feel all that into court. I don't know what
the role of a solicitor. I know, if I was charged,
and I hope not to be charged again, I'm just
put that record, But if I charged again, like I
worked on my defense and worked with my great legal
team for recording a conversation on the telephone that almost
seems like I did embrace technology leader on it. I
(21:14):
would put it in AI and go, where is the
strength for the case against me? Wants a brief of
evidence being served? What is the strength of the case
against me? And this is my case, and then present
the cagun argument.
Speaker 3 (21:25):
For sure yep.
Speaker 1 (21:26):
So is there a role for solicits anymore?
Speaker 3 (21:29):
Well, do we need less solicitors? We won't go if
we put it.
Speaker 2 (21:34):
Look like we were saying before on my lens through
policing was a certain bias, and that's so changed now
that I've come out on the other side and really
respect and admire the work that I've done by defense
teams because they've got to go through so much information
on these serious cases, and you know, is there an
avenue there to explore or not? Is this the right
(21:55):
person or not? That's our judicial system. But there's definitely
a place for AI. And imagine the hours that are
spent on that. But you know, the research solicitors, the well.
Speaker 1 (22:06):
The poweralegals, legals, so find this case law and all that.
Speaker 2 (22:11):
Yeah, sorry, guys, it might be, it might be changing
for you, but you know it'll shift to another focus.
We saw this with computers and everyone thought, oh, it's
going to be the end of it all. Even's going
to be out of jobs. But we've seen it now
advanced manufacturing. It's just shifted and I dare also use
that word that I hate from COVID pivot.
Speaker 1 (22:31):
You're clearly from down Victoria.
Speaker 2 (22:33):
That's right.
Speaker 1 (22:33):
We heard it every stingy Daniel Andrews every day.
Speaker 2 (22:38):
But yeah, we've seen the advancements and new technology does
bring new challenges and brings new things, but it takes
that to the next step and next generation. Who knows
what we're talking about in twenty years, like we look
now at you know, you look at some images and
videos now and you go, that's a I think in three,
four or five years, you're not about.
Speaker 1 (22:56):
To tell we went well, the young child down in
South Australia that disappeared gus that they were getting a
lot of AI and a lot of media. Different organizations
reached out to me to comment on it. Just based
on my experiences with the Willim Turreal matter and what
I acknowledged with the Willim Tural matter, it was very
difficult because people there were a lot of sightings of
(23:19):
William Tyrell and yeah, you had to sift through it,
you had to follow it up. And we ended up
working out the triarche system on what weight we'd carry
a siding of wim. But then another layer came in
with the gust thing because there was AI generated images
of and yeah, you could work it out relatively easily
that it was AI generated, But as you said, it's
(23:42):
going to get better and that's going to be very
hard for investigators to sift through what's real and what's not.
Speaker 2 (23:48):
Yeah, it's pretty scary, not just the images, but then voice,
the use of people's voice. And you know, we did
do a bit of this on the show Hunted where
we cloned the person and we're able to write the
script and rang their mother as if we were the
fugitive sneak. Yeah, but you know there's a criminal element
(24:09):
that comes in on this too, doesn't it. So you know,
we see a lot of frauds that that happen. Organized
crime internationally will use this technology to try and commit fraud,
which is why I have with my kids. The best
way to try and get around is you have your
code word. Right, So if my kids are overseas, Hey Dad,
I'm stuck on need some money, which is your classic case, son,
what's the code word for the day. Yeah, I won't
(24:31):
know that, so it might sound like my son might
be purporting to be here.
Speaker 1 (24:34):
Because that's your protection. That's a very simple, practical.
Speaker 3 (24:38):
Easy way. But then they've got to remember the code.
But no, that's that's our simple one.
Speaker 2 (24:43):
And I know if I meet the talking to them
on the phone, or you know, they've sent me a
voice message, you know, then fact check what's the what's
the code?
Speaker 1 (24:50):
So with AI and voice copying the voice, there's all
these hours and you're you're the same in the stuff
that you've done in the media, and you've got me
talking here on the podcast. They in effect could copy
that voice and at this point it could be leaving
the message. But I dare say it would get to
the point where they could have a conversation.
Speaker 3 (25:11):
In my voice exactly.
Speaker 1 (25:13):
Yeah.
Speaker 3 (25:14):
Scary stuff, isn't it.
Speaker 1 (25:15):
Especially if they fhm, my mother.
Speaker 2 (25:18):
That's right, And that's that's the vulnerability. We always see
organized crime will keep shifting to where the gap is.
We've seen that in the tobacco industry. Yes, we saw
it through the armed robbery days. You know, when was
the last time we saw a bank robbery. Well that's
because we target harden, didn't We know now that the
real opportunity is online fraud and scams and targeting our
elderly that aren't up with the tech. And there's a
(25:39):
lot of work being done by the banks to educate,
you know, fact check, you know, if that email doesn't
look quite right, and don't do it. So there's a
lot of space. So that's why it's so critical to
have this developed. It's going to be developed. The crooks
are going to use it to law enforcement needs to
be right up there with them, not falling behind.
Speaker 1 (25:58):
Well, we have to because I thing about the crookxy
creative in the way they can exploit the situation and
moving as you said with the tobacco wars or whatever,
if they see an area they can exploit. But it
makes me concerning that even before you'd go to court,
if you had the phato of someone there or something
(26:18):
and there's a hard evidence, we don't know. Sound like
Donald Trump, fake news, fake.
Speaker 3 (26:24):
Image, that's right?
Speaker 1 (26:26):
Where is it? Where's it going to?
Speaker 3 (26:27):
Where does it end.
Speaker 2 (26:28):
Yeah, Yeah, that's a scarity there are you know, there
are ways of trying to find if that is AI
generated or not. And we're getting into a complete area
where which is outside my wheelhouse of how that's done.
But that's part of the reason of building it, building
it with the ethics that it needs to be.
Speaker 1 (26:45):
Getting back to two investigations, let's just put their heads
to how else it could help the fact sheet and
all these things seem relatively little, but in a high turnover,
the high volume crimes, the type of things that the
these police are out doing on a daily basis, and
it might be arrest arresting someone for a break and enter,
(27:06):
and that should in theory free police up if we
embrace what AI can do for us.
Speaker 2 (27:14):
Yeah, I think that's a no brainer. Yeah, it should
free up countless hours. How do you We'll just give
one example of the one homicide detective, then a few
times at by the volume crime as well, which you know,
potentially the crime mapping, the predictive data of where the
hot spots are. We've seen a lot of great work
done in that space in policing over the years. But
there's another argument for another aspect of AI to look
(27:36):
at the crime mapping. You know, where where's all your
cars getting broken into? What do we used to say that,
You know, you'd see a crime spike of those volume crimes.
You go, well, there's a dealer around here somewhere, because
they're not going to hold onto the stolen gear for long.
They want to offload that really quick.
Speaker 1 (27:50):
They'd be someone could be as simple as someone targeting
a shopping center car park. Car has been broken into,
and you know we'd put a car in there, settle
up with clear left unlocked and wait till the person
come to arrest them. But identifying patterns in crime would
be good. Linking crimes is another thing, and we've got
a lot better, I think in law enforcement because there's
(28:12):
more of an exchange of information between jurisdictions at state
police and that. But serial offenders the type of thing
that quite often went unnoticed because there was no link
to the crime that this person had committed in another location.
Speaker 3 (28:30):
Yeah, exactly.
Speaker 2 (28:31):
And I think you just speak to the core of
what I'm hopefully trying to push is that all law
enforcements need to get on board this and once the
border they don't care about the borders. We've seen that
and we've sadly learned the hard lessons of those serial
offenders that will jump jurisdictions because we weren't communicating. We're
so much better now, but we can still improve and
if we're going to do this, I think everyone needs
(28:52):
to get on board because you know, the I guess
the armed robbery or the sex offense that's committed over
in way, maybe the exactly the same or the way
of offending. Then all of a sudden it started up
here in Queensland and there's a real space there as
well to understand that and learn from that and market
as a flag and go, well that needs to be
looked at. Is that possible that that could be the
(29:13):
same offender?
Speaker 1 (29:13):
Okay, interview So I keep coming back to the interview room.
You're playing with my head thinking I could sit here
with interview you for a couple of hours, and then
my phone tells me you've forgot to ask this or
forgot to ask that.
Speaker 3 (29:27):
You could.
Speaker 1 (29:28):
A lot of the preparation for the interview room was
preparing for the interview, and that quite often meant you
had to read through the whole broef and get all
the details so you had it in your head, I
would imagine AI could help with that.
Speaker 2 (29:42):
Yeah, the same question as what we asked before, rather
than the question to the AI is put this together
for the district court or Supreme court? Is with all
the available evidence to me at this stage? Draft mean,
and if you plan based on these offenses and in
line with the points of proof to prove the charge,
you can't.
Speaker 1 (29:59):
Now that I could see the benefits of it. But
I also think, and you said, you need that human oversite.
You need an experienced police officer there listening to how
you answered the questions, because I've seen mistakes people have made,
and I've probably made it myself that you work so
hard to Okay, I got to ask these questions. You've
got twenty questions written out. Good, I've asked that one,
(30:21):
I've asked that one. But you're not listening to the stances.
Speaker 2 (30:23):
Yeah.
Speaker 1 (30:24):
So yeah, it's a balance between the human oversite and
what AI can bring.
Speaker 2 (30:29):
Yeah. I've been caught by that, I've got to say embarrassingly,
because you're so focused on your interview plan and where
you're going. But that's why I would have your corroborator
next to you to hopefully pick up on that. And
when I first started homicide, you know, I'm on my
introduction to the homicide world. I think it was day
two I was there. We'd actually had a fellow in
for interview, you know, and your teams watching through the
(30:49):
other window, listening, and they're all contributing because everyone has
different aspects of the brief, aren't they. So that was
how it was done and here's the next step to that.
So if you you know, you still have people listening,
you still have your boss listening or checking, but you've
got your AI assistant. They're making sure that you've covered
everything off and hopefully validate where you're at.
Speaker 1 (31:10):
Well in what we're talking about. You would think that,
because I'm not sure if we spoke on camera or
off camera, we're certainly talking about that the struggles that
most police forces in Australia are getting recruiting people at
the moment. So the police that we've got, we've got
to use them smart, use them most efficiently. And if
(31:31):
law enforcement embrace this, one would think that it would
free up police from a lot of the work behind
the desk.
Speaker 2 (31:38):
Yeah, I definitely feel that, and I feel that it's
about trying to identify what's going to be your big
spang for buck, and right now I think that's the
administration piece. My heart goes to helping solve crime. Obviously
that's what I want. But the biggest piece I would
think here, I guess for the senior police is getting
rid of the administrative tasks that are day to day.
(31:59):
If you LANs your brief preparation, your fact checking, your rosters,
all those administrative tasks that could be taken away and
all of a sudden, you know, your analysts hours a cutdown,
investigators hour as a cutdown. We're trying to work through
some data now on actually putting a monetary value on
how much that is to be able to sell. This
is why we need to invest in this, and this
(32:20):
is why we need you to come on this journey
with us.
Speaker 1 (32:22):
That would be interesting. And I think with an analyst,
and I've worked with some great analysts over the years,
and I think how they would readily adapt to using AI,
like as in not making them obsolete by using it.
It just frightens me the amount of information they'd be
able to spit out so it doesn't have to be
the dumb lead investigator. The analysts and the analyst is
(32:48):
embracing it.
Speaker 2 (32:49):
Yeah, another tool for them to use, and they're forward
facing to the investigations that matter and using their skills
that matter, not just a monthly reporting. It's an important
part of the piece for the big strategic vision. But
again that doesn't need to be done by a human
can be fact checked and looked at. But if you
train it appropriately and it's all there for everyone to
(33:10):
see live. We've seen all the advancements in body worn
cameras and well that it's all there. It just needs
to come together.
Speaker 1 (33:17):
Well, people might think this is an exaggeration, but I
think you might be able to corroborate me here. I
retired as a detective chief inspector. I was an inspector
for the last fifteen years, or a detective inspector in
the police. I reckon thirty to forty percent of my
time spent when I'm leading investigations, leading all different murder investigations,
(33:41):
thirty to forty percent of my time with been administration
and menial administration. I would I'd cringe saying it, but
it was the reality that once a month I'd be
spending two or three hours adding up card diries making
sure the car diaries calculated. And if I missed doing
the car dory, that was probably worse than not solving
(34:04):
the crime. I should submit the car dories on time.
They had to check all the people under me, so
things like that that should just be by the wayside.
Speaker 3 (34:13):
Yeah, again, no brainer.
Speaker 2 (34:14):
Yeah, make it happen for everyone up and you know
there was a joke about it, but there is that
stress of the ADMIN that comes with it as the
team leader as well. The monthly times come around and
you know you need to make more time for your admin.
Need to sit down do this not Well, we're just
going from job to job to job and if we
need that now you need to make these red lights green.
Speaker 1 (34:35):
I don't think I'm unique. I don't think you speak
to any operational police officers and the frustration I used to,
in fact joke sometimes say about one o'clock, okay, admin done,
let's start investigating this murder. But it was quite often
we're all office bound because we're catching up on record
keeping administrators.
Speaker 2 (34:54):
Yeah, it's definitely got its place. Yeah, look, it's a
no brainer for me. But we're just we're really close
to I think getting on that right path it's got
to have some courage to do it. And there's been
a lot of changes in senior management and chief commissions
around the country, so you know there's time to come
in here and lean on.
Speaker 1 (35:13):
I don't blame anyone that might have fallen behind or
not realized because it is changing so rapidly. We wouldn't
be having this conversation two years ago. It would be
completely different. We'd be talking like it's a sci fi movie,
but now it's reality. So we talk about saving time.
But the thing that excites me most about is how
(35:33):
it can enhance investigations.
Speaker 3 (35:35):
Yeah, definitely.
Speaker 1 (35:36):
It just seems like a no brain of the human factor,
the sharp minded analysts that you'd have on an investigation
and the detectives pouring over briefs that can potentially be done.
Speaker 2 (35:49):
So much quicker, and imagine what we could do preventing
crime and the.
Speaker 3 (35:56):
Unseeing cost of prevention.
Speaker 2 (35:58):
You know that it was always that movement towards the
end of my career was let's try and stop the
crime before it even happens.
Speaker 1 (36:03):
Well, that's an interesting point too, and I haven't covered
and that's something that I wanted to wanted to cover,
is that we've been focusing on solving crimes, but how
can AI be used to prevent crimes? Give us some
examples there.
Speaker 2 (36:19):
Yeah, well, I think the clearest example is then predicting
where the hot spots are, And we see a lot
of work with that at the moment, but that's generally
data predicting where these hot spots are. The AI aspect
might come into that potentially, And where I see it
more so is you know you'd see two, three, four,
five street robberies or that's one example where if we
(36:42):
can identify this offender after the first one or the
second one, we know they'll generally keep going because either
their motivation is either greed or drug induced, camera or whatever.
If we can identify that person on day one or two,
how many offenses have we stopped by being committed by
getting that person here not down the road. That's one
aspect to it. There are some other discussions I'm having
(37:04):
with with other developers around the human behavior models. So
you know we've all seen you know, the kids or
anyone just before you know, you've watched the footage and go, okay,
I can see they're starting to get a thousand yards there.
They're doing this and it's about to be on. So
you can see the human behaviors through footage that something's
about to happen.
Speaker 1 (37:24):
Right.
Speaker 2 (37:25):
You know, when we've investigated either you know, your machety
attacks or the like that we see a shopping centers
these days, there's a lead up to that. It just
does not all of a sudden explode as a general leader.
So we can train AI to look at those human
behaviors and what are the signals here, what's happening here.
Speaker 3 (37:40):
So we're getting into a.
Speaker 2 (37:41):
Whole other area now of non policing. But you know,
I know private security is very interested in those aspects.
Speaker 1 (37:47):
And that's interesting, and we'll use it just by way
of an example, the attacks in the shopping centers because
it's been in the media a bit. But yes, it
would be a gathering.
Speaker 3 (37:57):
Yeah, I'm just featuring like a thing.
Speaker 1 (38:00):
Spirits investigator might be able to pick that up, but
AI might be able to do it even better.
Speaker 2 (38:05):
And we're trying to monitor I don't want to name
particular shopping centers, but if you're trying to monitor hundreds
and hundreds of cameras, yeah, or big public events, the
AI can do that, and it can alert the person
that's there in charge of looking at those cameras. We're
seeing something here on camera forty six and forty seven
of a gathering of some it could.
Speaker 1 (38:22):
Be let's break it down. That could be put if
there's one hundred cameras around the shopping center and AI
had put in place that if you see people dressed
in this style of more than four gather at a location,
alert that saw draw correcting. We want to know and yeah.
Speaker 2 (38:42):
That operator and then look at that and go oh,
that's nothing, or well, we've got a problem here and
you're onto it twenty thirty seconds a minute beforehand. Hopefully
you can get there and disperse them before anything even happens,
so you prevented the crime. That's always So there's an
aspect of human behavior. So you biomechanics that you need
to train with data sets, so you need access to
all these different ones that have occurred in the past,
(39:04):
and the data is there. We've got a partner with
the right people to build that data and train the
AI to Okay, that's all we're looking for, or weapons identification.
I know there's a great project that's been done through
one of the people I mentioned before. Around identifying firearms,
similar to what I said about vehicles. Yeah, so you
know what gun you're looking for or exactly which one
it is, and.
Speaker 1 (39:25):
What you're talking here, Just so I understand and hopefully
other people understand, Like, identify a gun. If I get
out of the glock, I can tell it's a glock,
but it might have a scratch on the barrel or
scratch on the pistol grip. It might be something like that.
But AI would pick up on that and go, this weapon,
(39:45):
there's the same weapon that was used in this shooting.
Speaker 2 (39:48):
Yeah, if it's all talking holistically to each other and
you've got good footage, so yes, you've got good footage.
But even post offense, if you've got your suspects arrested
with is you know, with twenty two or a glock
or whatever the case may be, that's that particular weapon.
All of a sudden, you're plugging into all these unsolved
ballistic reports of this actually was a clock that was
(40:10):
the weapon that killed this person in this murder, because
we've got the ejected casing, et cetera. All of a sudden,
you're linking all these cases again cross border stuff that
is volumes and volumes of forensic material that it can
do in seconds if you're focusing on the right problem.
What is the problem we're trying to solve here? So
weapons identification is one similar to vehicles, they're definitive. There
(40:34):
is an exhaustive list. There's only so many makes and
models of firearms on the planet, so you can teach
it too to come up with those options of this
is what we definitely think it is, or this is
potentially what the type of weapon is. And again then
you look at your execution of search rights. We know
we're looking for a particular maker, model firearm as opposed
(40:54):
to the guesswork. So there's a lot there. I'm sort
of overwhelming the conversation with a different aspects, but I
think it really comes back to that initial piece of
what is the issue that we're trying to use AIEN
to solve? Is it an administration? Is it prevention of crime?
Is it cold cases?
Speaker 1 (41:13):
And there's all these different things that could enhance.
Speaker 3 (41:16):
Or assist working with it.
Speaker 1 (41:19):
Hey guys, it's Gary jubilin here. Want they get more
out of I Catch Killers? Then you should head over
to our new video feed on Spotify, where you can
watch every episode of I Catch Killers. Just search for
I Catch Killers video in your Spotify app and start watching.
Today you said about overwhelmed with information. It made me
(41:41):
think while you were talking about how we were overwhelmed
with the availability of information, that when I first started
as a homicide detective, there was a couple of people
had mobile phones. You could find out or listen to
their mobile phones, or they had landlines and you could
track them that way, and that was you know, that
that was the line of inquiry. A bank might have
(42:04):
a CCTV camera out the front, but only a bank
or a licensed premises. And I did actually see towards
the latter part of my career we were getting overwhelmed
with information, like there was so someone's murdered. You'd get
called out. And the instructions I was given in the
early days as a young homicide detective go out and
(42:26):
collect all the CCTV footage from businesses and you'd come back, Oh,
that service station had one, and you'd be all proud
of yourself, and yeah, there was one one at the bank.
Now you say that private homes cars have got you know.
And then as my career progressed, I'm seeing at the top,
and I'm looking at all the information we've got and
I'm going I'd be pulling my hair out if I
(42:47):
had had hair, just overwhelmed with the information. And AI
swings the pendulum back in favor of being able to
handle all that information, doesn't it, because we did get
to the point where we were able well.
Speaker 3 (43:00):
Well, information intelligence is key, isn't it.
Speaker 1 (43:02):
Yeah.
Speaker 2 (43:02):
So the sooner you can get that in investigator's hands
and understand it, yeah, the better you position for your strategy.
So all that information that you're getting from CCTV, from
that murder, getting it all together, if you miss that
bit of footage that shows it turned left onto the freeway,
then you know that's a complete game change, isn't it.
So you won't miss that. And whether you're plugging in
one hundred cameras worth the footage into this AI, it'll
(43:26):
track that you tell it. I want you to track
that car. That's the one we want out of all
this information. We've got timeline and tell me where it's
been and what it's done and bang.
Speaker 1 (43:34):
And what people didn't When I say people, the public
if you're not involved directly in it. When we've got
listening devices in houses, you put a listening device in here,
and you've got someone monitoring it. They can even monitor
it live and if something happens or invariably you've got
to come back and review it because it's not been
(43:54):
monitored live. So you've got twenty four hours of listening. Yes,
like people going, have you gone gone through that? Well,
it's basically in real time. And then when you hear
the conversation, then you've got to document the conversation.
Speaker 2 (44:08):
It was so time, and then they're more blow our
minds with then your hat a layer of you know,
if you need a translator, Yeah, so yeah, I get
one hundred percent your point. And that's that's those complex
pieces of information and large amount of data that it
can deal with. And that definition of AI you said
right off the top, that's what it's doing. It's looking
(44:29):
at all of those conversations and if you're after key
words or if you're after a key you want them
talking about the victim of a planned murder. You know,
you've got a contract, kill job, whatever you want to
know whenever that pops up.
Speaker 1 (44:43):
And we would get to the point where I've used yeah,
limited times when I have people look at the terminology
used by by an offender from a psychological or suspect
from a psychological point of view. You could even feed
that into yeah, couldn't you. It appears that and he's
talking about that Issue's demeanor or his language is changing.
(45:04):
Little things like that, and it's only incremental victories, but
they're the little things that make the difference.
Speaker 2 (45:11):
And gives the advantage back to the police. Yeah, there's
one other piece that I will talk to and it
can be quite confronting, so I don't want to trigger
anyone on here, but a really I think crucial piece
of where AI can help as well as in the
child exportation material space, you know you would a basic example,
police do a search one on a laptop and you've
gond a laptop and you've got hundreds or thousands of
(45:33):
really horrific images of child exportation material. Some of that
you may be concerned is contact defending. So you're trying
to identify this kid that's clearly in the footage or
in the images with you're accused. You need to find
out who this kid is to stop the contact defending.
So then you're madly trying to figure out how how
is this possible? Where is that you're looking at the
metadata of the images, you're looking at all this sort
(45:56):
of stuff. The AI should be able to work out that,
and it may be to use facial recognition technology of
the child to try and identify who is this child.
So that's the piece. I feel that that's where the
trigger for scraping social media is important, because that kid's
photo might be in the background at someone's birthday or
at a any event, and all of a sudden, you've
(46:17):
got a match on this kid that's being offended against,
and you can immediately stop that offending through the use
of the AI to identify that kid. There is some
work being done in that space by the AFP and
monash Uni, and that's fantastic, but you know, it needs
more support, it needs more funding, and it needs more
attention to help protect these kids because it's crucial because
there's nothing worse than investigator. I can't find this kid
(46:39):
and we know they're being offended against, and you can't
release it to the media. You know that you've got
to protect the identity of these kids.
Speaker 1 (46:44):
That seems like a no brainer that if I can
help in that space, and I understand what you're saying
the kid that's been sexually abused and you've got to
identify that child. I could help.
Speaker 2 (46:57):
That's why I want to write. Partner with the right
investors that have the same morals and purpose that we
have at Perigrin to get this done because we can
protect kids.
Speaker 1 (47:06):
We touched just briefly. Fiscally, this seems like a no
brainer for law enforcement. Yeah, freeing up, freeing up more success,
reducing crime. There seems to be benefits all over.
Speaker 3 (47:20):
Yeah, I think so.
Speaker 2 (47:21):
But it's also challenging times, isn't it. And I can
only talk in the Victorian context because that's what I
saw when they were seeking a new Chief commissioner. You
to have to save ordered by the government to save
five hundred million per year in your budget, so you know,
good luck, chief, but better you than me. But that's
the first step off point. So you're then coming along
(47:43):
or a company like Peramn's coming along, going we want
you to partner with us because we want to invest
in this. You really can't expect the government to be
stepping or the police to be stepping up in their
budget to assist. Hence what we need corporate support. But
it is a no brainer but eventually they'll get there
to see the benefits. We've saved so many hours in
investigators time and administrative tasks. But right now it'd be
(48:06):
very challenging and for them to commit.
Speaker 1 (48:07):
I hope they see it that way because, as we said,
sometimes police, when technology changes, we drag our Yeah, I'm
sure there's not many detectives that have to g to
a private company to get something that the police. You
would assume police had capabilities of but didn't.
Speaker 2 (48:25):
Well, like I said, I was at a conference eighteen
months two years ago and it was around the child
expectation and very senior police and I won't name the organization.
We're there going when is corporate? When is private sector
going to step up and help us? Went, oh, let's
have a conversation FORFE So well yeah, so yeah, it's close.
We just need to keep it going, keep the conversation going,
(48:46):
and keep developing it. But look, coming back to it,
I just thought of when were speaking before around the
use of AI and cameras and that now it's actually present.
Now I've got your cameras around my place and they're
they're giving a plug to the UFI system, but they've
got AI in there, So you can now load someone's
face onto your home CCTV system and it'll alert me
when that person is there, whether I want them there
(49:08):
or not. So the AI is already plugged into some
of these things that are there, so you know, as
you said, it's coming, it's here.
Speaker 1 (49:14):
Well that that could you just made me think we'll
go into an area, very real area that like domestic
violence and then that type of thing. But you could
forewarned if someone's someone's turning up.
Speaker 2 (49:27):
So that's that's that natural evolution of what used to
just be, you know, an old school c CDV camera.
They've got chunky ones on there that are now these
small little ones that are either powered by solo or battery,
and they've got AI into them that can alert you
to know whether you do want them or don't want
them there, can give you an alert to your phone,
and you can be anywhere in the world. I think
I was lucky enough to be overseas not too long ago,
(49:47):
and I'm getting alerts from my camera back home. Oh
that's thank you. That's the person turn up to feed
my turtle. Thanks mate, Yeah, fair cool turtles.
Speaker 3 (49:57):
Turtles, turtle's got the weight.
Speaker 1 (49:59):
Well, look I really enjoyed the chat, and thanks for
coming on. You're the right person, because I know that
someone that brings me crashing into this brave, brave new world.
But the way it's evolved, you know, I might have
to have you back in six months time, and it
might even be you. It might be a hologram or.
Speaker 3 (50:15):
Something I might remote in. Virtually, Yeah, yeah.
Speaker 1 (50:19):
I think if you've given you've given me an understanding.
I hope the listener is an understanding of what's out there.
Speaker 2 (50:25):
I hope so, and I hope that point of just
looking a bit further about what I can do. It's
not just a one shop or one scenario fits all.
It's what's the problem, what is it? And to have
some courage to lean into it. But good luck trying
to sleep tonight. Carry I've blown your mind.
Speaker 1 (50:42):
I'm down the rabbit hole. Now I'm gone. But we
finish off with because I do think that's important, because
we know how power can corrupt in police and that
the ethics and the integrity that needs with this type
of type of.
Speaker 2 (50:58):
Thing part of part of our lives absolutely critical, which
is why again I've said countless times need to be
built here for here.
Speaker 1 (51:06):
Just hold on for a second. I'm just going to
put in chat GPT. How I say goodbye to Graham.
I love it, New Graham for coming on I catch Killers.
It's been a very entertaining discussion. Thanks Hasburt Chees
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Betrayal: Weekly
Betrayal Weekly is back for a brand new season. Every Thursday, Betrayal Weekly shares first-hand accounts of broken trust, shocking deceptions, and the trail of destruction they leave behind. Hosted by Andrea Gunning, this weekly ongoing series digs into real-life stories of betrayal and the aftermath. From stories of double lives to dark discoveries, these are cautionary tales and accounts of resilience against all odds. From the producers of the critically acclaimed Betrayal series, Betrayal Weekly drops new episodes every Thursday. Please join our Substack for additional exclusive content, curated book recommendations and community discussions. Sign up FREE by clicking this link Beyond Betrayal Substack. Join our community dedicated to truth, resilience and healing. Your voice matters! Be a part of our Betrayal journey on Substack. And make sure to check out Seasons 1-4 of Betrayal, along with Betrayal Weekly Season 1.