January 17, 2025 • 26 mins
What if the future of AI didn’t live in the cloud, but right in the palm of your hand?
In this episode of Innovation Heroes, host Ed McNamara uncovers the revolution of on-device AI with Qualcomm's Mike Belcher, Director of North American Commercial Channels.
Discover how this cutting-edge technology is reshaping industries, slashing costs, helping the environment and redefining what’s possible in personalization, efficiency, and security.
Featuring: Mike Belcher, Qualcomm’s Director of North American Commercial Channels
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Ad Spot (00:00):
Ready to elevate your work?
The Snapdragon X Elite processorpowers through demanding tasks with
cutting edge AI, up to 37 percentless power usage, and features like
live translations and enhanced search.
Focus on big ideas whileyour laptop handles the rest.
Learn more at shi.
com slash partners slash snapdragon.
Mike Belcher (00:22):
About 2033, we would actually hit a point where it would
cost more than the GDP of the world torun the internet at the current growth
rates that the internet was growing.
So we have less than a decade tofigure this out and find better ways
to use AI and do so power efficiently.
Um, and honestly, I think, uh, ondevice and endpoint AI is where we're
(00:44):
going to have to migrate the majorityof the work that we do every day.
Ed (00:48):
In the world of technology.
Heroes are everywhere.
They're overcoming disruption, deliveringsustainable outcomes, and fearlessly
forging the future to solve what's next.
Join me, Ed McNamara, as we meet thepeople and businesses driving change
in our constantly disruptive world.
This is Innovation Heroes, apodcast brought to you by SHI.
(01:14):
When you think of the AI you interactwith on a daily basis, what comes to mind?
Chat GPT, Amazon, Google.
What do they all have in common?
They all live in the cloud, but thetechnology world is shifting gears
and a new player is taking stage.
On device AI.
In a time where speed, personalization,security, and efficiency define
(01:34):
success, this game changingtechnology is rewriting the rules
for business and consumers alike.
By processing data locally ondevices, on device AI eliminates
latency issues, enhances performance.
Privacy and enables real time decisionmaking opening doors to innovation
like never before from transformingindustries like healthcare and retail to
(01:55):
powering smarter IOT systems on device.
AI is driving a revolution in howwe interact with technology today.
Mike Belcher, director of North Americancommercial channels at Qualcomm.
We'll help explain why this shift matters,how it's empowering businesses to operate
faster and smarter, and the excitingpossibilities it unlocks for the future.
(02:16):
Mike, welcome to Innovation Heroes.
Oh, Ed, glad to be here.
Thank you for the invite.
Speaker 3 (02:21):
Mike, you've been in IT and computing for a couple decades now, and
I'd love how you got into this space.
Ed (02:28):
Yeah, so, um, you know, I've been in, uh, mostly on the OEM side, uh,
at, uh, Gateway and HP for 25 years,and I've been here now, uh, almost, uh,
three, um, and, and I didn't start offmy career there, uh, but I've been in
tech, uh, since I was in high school, andbefore there was anything called Windows,
(02:48):
and, uh, so, uh, I, I was very lucky.
My mom actually, uh, never went tocollege, but she self taught herself
PCs as they were coming out, so Iwas the first 286s and 386s came out.
Uh, she taught herself computer language,taught me while I was in high school,
and, uh, kind of created a love affairwith the computer with me ever since.
And so I was always,uh, engaged, involved.
(03:11):
Uh, we were doing, I think, someof the first online banking.
My mom worked for Bank of America atthe time, and she was one of the first
guinea pigs of actually doing that,you know, for those who remember,
we would log into the internet, right?
So today we're talking about ondevice AI, um, which seems like
a new idea, but it really isn't.
(03:31):
Um, before the internet, um, allAI was done on device, right?
And can you tell me a little bit aboutthat and what are some early examples?
Mike Belcher (03:40):
Yeah, maybe I'll give some of the examples that,
um, that that I have personally.
Right?
So, uh, in past life, I was responsiblefor looking at emerging technologies
when I was at HP, um, really so that wecould help decipher and where we believe
that technology would would end up.
Right?
And so.
How would we help higher educationinstitutions, for example, one of
(04:01):
them revolved around the amountof research that's being done at
these are one institutions, right?
There is a ton and there's so muchdata that that they're collecting
and there weren't really greatways to analyze that data locally.
And so some of what we starteddoing at HP, we're building these
super high performance workstations.
And so imagine a tower device.
(04:23):
We started working with NVIDIA very early.
In that stage to develop thecapabilities and software to actually
drive the GPUs to process this dataand do that at a localized level.
And so what it allowed to do via fundingthat these institutions would get, they
could buy these 50, 60, 70, 000 PCs andthen use those continually over and over
(04:46):
again for all of their data research.
Right.
And, and mine those, uh, thosenumbers, uh, really come, come
together with better answers, right.
For, for the research they were doing.
But do so in a very safe way, alllocalized and not having to go to the
cloud for the, for, for that data.
So it was one of those kind ofreally unique cases that in me
went, Oh my gosh, there's, there'shuge opportunity for this, right?
Ed (05:08):
Yeah, the, the, the, the role of higher ed in those early days,
it's just cannot be overstated.
And then cloud computingbecame a big thing.
So, um, you know, why did we makethe switch to cloud based AI and,
and what advantages did that providefor businesses and consumers?
Yeah.
So, you know, I, I think the, the,the big turning point is you, you
already alluded to as chat GPT, right?
(05:29):
And when, when that came outwith generative AI and you
could ask any question and get.
Some pretty darn good answers,and even if they weren't good
answers, they sounded good.
So some of that work youjust had to do there.
We couldn't afford to put allthese devices at 50, 60, 70,
000 on everybody's desk, right?
Um, and so the ability for us to takeadvantage of the compute power to take
(05:53):
advantage of the data up in the cloud.
Really became the only way to makethis effective and particularly
where we wanted to ask these reallycomplex, really large questions where
you needed these huge data sets.
We really had to do that in the cloudand the early days of us cutting our
teeth on device, I think, really led theway for them to start to realize, oh,
my gosh, we could do this in the cloud.
(06:14):
In fact, it may be the only waywe could do AI very quickly.
Yeah.
Speaker 3 (06:19):
Right.
So, so now we're migrating back to,um, on device AI and to start, what
are a few examples of on device AI thatexists today and why is the tech world
making the switch from, from cloud,uh, to on, on device for, for AI?
Ed (06:33):
One of the reasons why we have to move from cloud to, uh, to, to endpoint
devices is the cost of what it takes todrive that AI up in the cloud, right?
And so it is very costly for all theAI providers right now, every inquiry
will cost you, you know, between5 cents and 60 cents, depending
on what it is that you're doing.
(06:54):
If you're like me, you aska lot of questions of AI.
So I'm one of those folks driving thecosts up for all of these providers.
But there's so much of this AIthat you don't have to, you don't
need those ginormous data sets.
You don't need these enormouscomputers that require lots of
cooling to keep them operational.
And honestly, the advent ofNPUs or neural processing units
(07:17):
has really led that way, right?
Of how do we do AI and withtechnology that was designed for AI?
And that's exactly what, uh,neural processing units are.
We started development of those 12 plusyears ago and actually deploying them.
If you've got a Samsung phone withone of our chips in it, that's, you
know, five years older or less, youprobably have a pretty powerful NPU
(07:38):
in it and don't even realize it.
Right?
So we've been deploying NPUsfor a number of reasons.
They are so efficient.
And so if you think about what it takesto drive these giant GPUs and power them
and cool them up in the cloud versus whatit takes to run a neural processing unit.
We can run 45 trillion operations persecond on an NPU and do it all day running
(08:00):
less than 300 milliwatts of power, right?
It just sips power to, uh,to, to, to drive itself.
And if we did that same thingwith a GPU, we'd be using
probably 60 or 70 watts of power.
So it's about 150 times moreefficient to do it on the device.
And, uh, if you have a Very largedata center in your neighborhood, your
(08:21):
hometown, because they are sucking thepower out of the grid right now, right?
To run these and to cool them.
And there's a huge concern right nowaround how do we power our homes?
How do we light and heat ourhomes when these data centers
are sucking so much power?
Um, we've actually done some, uh,some, some great research around this.
And by 2040, we would actually hita point where it would cost more
(08:46):
than the GDP of the world to runthe internet at the current growth
rates that the internet was growing.
That was before chat GPT, before theNVIDIA revolution up in the cloud.
Um, we did that same, uh, data research,uh, earlier this spring, like in March.
And we factored that we're going toreach that same level about 2033.
(09:07):
So we have less than a decade tofigure this out and find better ways
to use AI and do so power efficiently.
Um, and honestly, I think, uh, ondevice and endpoint AI is where we're
going to have to migrate the majorityof the work that we do every day.
We started working with Microsoft becauseof our background in this, in the, in
the cellular world of how could we takea much more power efficient model like
(09:30):
ARM, um, and in ARM instruction sets.
And actually, uh, drive windowson arm with a, with a risk
processor, but also embed NPUs.
So now Microsoft could completely redesigntheir OS to take advantage of that NPU.
Now it requires a lot ofchange to make that happen.
(09:50):
A whole lot of improvementsinto the systems.
But boy, the vision for that reallystarted about four years ago, and we're
already starting to see some reallycool things like, uh, if you haven't
seen, uh, in Visual Studio Code,there's, uh, an app called CodeGen.
You basically type in samples of yourcode and then type in a request of
what you want us to do, and it willgo write the code for you, right?
(10:11):
And so you think of what takes days todo, you can do in minutes now, right?
Right.
And so the ability for us to leverage AIto do these things that it does really,
really well is exactly what we can do ondevice and then not worry about the cloud.
This code that might be very discreetand you don't want out in the ether,
it's all self contained on your device.
Speaker 3 (10:32):
And I think our audience, you know, which is a pretty technical
audience, can, can, um, you know,understand the benefits of on device AI.
But, uh, you know, how,how do you describe it?
Ed (10:41):
Yeah, so the, the ability to have on device, and particularly the way
that Microsoft designed Copilot Plus, Isthat it's self contained on the device.
It's designed to operate, not beingconnected to the Internet at all.
Right?
And honestly, uh, it's incrediblysafe and secure that way.
Um, one of the things that they wantedto use AI for, and they needed these
powerful NPUs to be able to drive it, wasin the background AI running all the time.
(11:05):
So when you're runninga Copilot Plus device.
The AI is actually running all the timeand it's learning you, it's, it's seeing
what you do, what, what applicationsyou touch, uh, how you write, how
you communicate all of those things.
It's learning how you operate.
And so I'll give you a really kindof cool example right now for myself.
I travel a ton, way toomuch, 200 plus days a year.
(11:28):
And so I am continually behind on email.
And so when I finally got my device, uh,and now we've, we've set up a co pilot,
uh, M three 65, along with our, ourdevice set, um, I've started running, um.
All of my officeapplications utilizing the A.
I.
And so for outlook.
Now, I will ask inquiries when I comeback from a long trip and say, please
(11:50):
find me all of the last 10 days.
Every email that's come through.
That's asking for a response forme and prioritize them for me.
And so in seconds, I get this list ofwhat I need to go hit right away and I
go hit those and then I do other things.
Right?
So I can actually prioritize my life.
Accomplish more instead of do moreby plowing through all this email
(12:12):
that is probably just looking at it.
You don't do anything with it, right?
I just can't ignore those.
So there's there's these types of thingsthat the personalization of the device.
Um, what will enable that?
Honestly, Windows is the only playerin this space, and we're the only
technology that can drive that.
Currently,
Speaker 3 (12:28):
I won't ask you to set the batting order for the people that
you get to, but you know, in youremail inbox before before others,
because we don't know who's listening.
But when you talk about copilotplus, what do you see is the biggest
differences than the original copilot?
Like how?
How will users use it differently?
Ed (12:44):
Yeah, a couple of unique ways.
First of all, thispersonalization of the device.
I think a year from now, those of usthat deploy this, uh, you know, in
this year, it'll take about a year.
But after you, you start utilizingthis after a year, you're not
going to be able to do without it.
Because the ability to use thingslike recall and find things in
seconds, the ability to havethese, uh, capabilities like, yeah.
(13:07):
Um, you know, uh, organizing yourdata or comparing spreadsheets for
you automatically, right on thedevice, not having to use the cloud
to do that, not worrying about whatdata gets set up into the cloud.
I think those are just going tobecome so embedded and part of what
we do, you won't be able to go awayfrom a device like this, right?
Because you're not going to be nearlyas productive and you certainly
(13:28):
won't be accomplishing more.
So I think long term, that'sdirectly what will happen.
So we're at the earlyinfancy stages of that.
But we're going to see all kinds ofapplications now take advantage of
these NPUs to completely rethink howyou use these applications and I'll
give you a one kind of a cool example.
Yeah.
So Adobe is one of these players, right?
(13:49):
Who, I don't know about you,I've tried to learn Photoshop.
I don't have the time to devote to it andto stick with it to be really good at it.
Right?
And so, you know, for them, the abilityto have almost every user be able
to use Photoshop comes down to AI.
So they've redesigned Photoshop to runnatively on arm and have now, uh, in
(14:09):
process of developing all the AI featuresrunning on our NPU that you can just
say, hey, Photoshop, Take the red eyeout of this picture and you can say it.
You can type it.
You can circle a blemish andsay, remove the blemish, right?
And you're done.
Um, so the ability for us to learnhow to use these tools really
comes down to how do we communicateand using prompts, et cetera.
(14:31):
And so you're going to see allsorts of tools like this come out.
If it's productivity, if it's security.
Everyone will be racing on howdo I leverage the NPU to provide
these really great AI capabilities?
Speaker 3 (14:43):
Yeah, i'm on a marketing team and I have about seven or ten
graphic designers who ensured thatI would never ever learn photoshop
after a few early attempts and theywere like Ed, we've got this for you.
Can we, why don't you put that down,you know, but everything that I hear
them say is that it's just, you know,their, their, their efficiency and
(15:03):
productivity is like going throughthe roof, you know, for exactly the
features that, that you're using.
Um, it, you talked alot about the benefits.
Um, like what aboutchallenges or drawbacks?
You know, what, what, what are the limitsor, or what do we, what do we still
have to be mindful of at this point?
Ed (15:20):
Yeah, so believe it or not, the requirements for
copilot plus are pretty low.
It's like 16 gig 256 on the on thehard drive and ideal is probably like
a half a gig hard drive and 32 gigs.
If you're going to do some heavier dutywork on the device, but if you have large
data sets that you really need to runbigger than about 13 billion parameters,
(15:42):
that's what we can run really effectively.
Um, with our 45 tops NPU, if yourdata sets need to be that large,
that's the kind of stuff we'regoing to have to do in the cloud.
But 30, 13 billion parametersis still pretty darn large.
And so, uh, you know, what?
1 of the things that I thinkwe'll start to see is where
we need that large data set.
We need that super huge compute power.
We're going to do that stuff in thecloud and we'll continue to do so.
(16:05):
But almost everything else.
I think we'll start to migrateto the device because of that
efficiency and cost savings.
And we're going to see a lot moreof, hey, I can download, uh, an SLM,
a small language model instead ofthese large language models, right?
And customize it based on my business.
If I'm, uh, if I'm, uh, uh, an end user,and it may be putting my schematics
(16:26):
and my if, when scenarios when our.
Field tech repairs out, right?
Um, so that the AI can do that for you.
You don't even have to beconnected to the Internet.
The device will just do that foryou while you're out on site.
Um, I think you're gonna see a lot moreof that kind of use case with these SLMs.
Really taking the place of, of what'shappening in the cloud customized
to you and your own use caseinside of your, your organization.
Speaker 3 (16:50):
So I, I, whenever I get to talk to somebody like you who really
sees how these technology is deployed,um, have you had something in the last
like 12 to 18 months where, um, you'vetaken a look, you're like, wait, they're
using this where and how, like, what,what would you say to that question?
Ed (17:06):
God, so, so many.
Um, honestly, I, I think, um, youknow, when I was at HP, we started
using some AI tools to help get tothose zero day attacks, et cetera.
I think what you're going to hear comingout forward and, and how folks, you've
got folks like Sophos and CrowdStrike thatare building for the NPU right now, all
(17:27):
sorts of amazing features and securitythat we just haven't thought of, right?
And part of it is you can use yourAI to help you design this and
help you put scenarios in place ofhow are you going to block this?
How do you develop the softwareto make all of that work?
Um, I think you're going to see somereally cool use cases for how we
use AI in security in particular.
Now you have to have some,some pieces embedded, right?
(17:49):
And running on your, your CPU there.
But I think you're going to start tosee the NPU, which typically operates
about seven times faster than a GPU does.
You're going to start to see a lot ofthat security feature set embedded.
Offload onto the NPU.
It's so efficient, right?
So you don't draw a lot of power.
You're going to draw it on the GPU.
You might just suck the battery outof your device or your fan would
(18:10):
never go off while you are running it.
Right?
And so the ability for us to do thatlocally is a huge boon for security.
And as we all know.
The black hats keep gettingmore and more efficient.
They're going to be using AI,figuring out how to attack us, right?
They
Speaker 3 (18:26):
have it too.
Ed (18:27):
Absolutely.
And so if we think about, right,like, uh, you know, Windows 10
support goes away in October.
I would be really fearful if Iwas, um, an IT decision maker.
Not planning on getting off of Windows10 by October, because guess what?
Those AI black hats will be doing.
They're going to be targeting you.
And so looking for ways to moveforward just on a security basis.
(18:50):
I think it's going to bereally, really important.
And then the ability to use theseAI security features to combat.
The a high black half as they'rethey're going after your business.
Ad Spot (19:01):
Are you ready to take your work to the next level?
The Snapdragon X Elite processoris here to redefine productivity
and power for modern professionals.
With cutting edge AI capabilities, ithandles the most demanding tasks with
up to 37 percent less power usage,so you'll stay productive all day
without worrying about your battery.
Snapdragon X Elite powered laptopsalso bring AI driven features like live
(19:24):
translations and enhance windows search,making multitasking easier than ever.
Let your laptop take care of the busy workwhile you focus on bringing your business
changing ideas to life even faster.
To learn more about the SnapdragonX Elite processor, go to www.
shi.
com slash partners slash snapdragon.
Speaker 3 (19:47):
You know, from a, from a governance perspective, does
having it, um, you know, on device.
Does that make it easier, easierto manage that it's on device just
for obvious reasons, or is thereanything that might be overlooking?
Ed (19:59):
Yeah, it really then boils down to how secure is your device, right?
And so if your users are accessing anyof your sensitive information now on
their devices, your device is as secureand your data is as secure as that is.
Right.
I think I love the way thatMicrosoft designed Copilot Plus to
run encapsulated on your device toeliminate that, that challenge, right?
(20:20):
It was one of those big fears.
It was a big fear I had.
You know, four or five years ago, aswe started looking at chat GPT, Oh, my
God, how do you keep your data safe?
How do you make those things happen?
Right?
And so the ability for usto do this self contained.
And honestly, they did it as muchfor security as they did for the
personalization of the device.
So this AI running all the time,learning you, learning how you operate,
(20:43):
helping you find files, et cetera.
All stay self contained on the device.
And so, as long as your device issecure, all of that is secure, right?
So, this now starts to reallyboil down into how do I ensure
that my end user devices areas safe and secure as possible?
1 of the cool things about windows11 is Microsoft has changed some of
the mandatory things that happen.
(21:04):
And honestly, part of it was inconjunction with us as we were developing.
Our chip sets for them.
We actually developed a securityprocessing unit that is 1 of those
processes along with an NPU andCPU GPU that is loaded with Pluton.
And so now it gives Microsoft abilityto have control over those devices.
Um, I was traveling in the Bostonarea a few months back when a major
(21:27):
malfunction occurred when somebodydid an update and it knocked down all
slew of PCs across the nation, theairlines all went down, et cetera.
I was trapped.
And, uh, you know, so you see it kindof firsthand and part of that was
Microsoft didn't have control over beingable to get those machines booted up.
They had to rely on a third party.
This will now eliminate some of that, uh,sort of challenge and make the devices
(21:48):
much more secure right out of the box.
So that there's some really kind of uniquethings that are happening as Microsoft
has developed this operating systemthat improve almost every aspect, right?
Not only just your user experience,but your security as well.
Speaker 3 (22:01):
So forward looking one of our last questions, you know, um, I always, I
always put way too much time on this year.
I was going to say, what canwe expect this technology
in the next 10 years or so?
I'll say, how about the next two years?
Ed (22:12):
Yeah, I'll give you a couple that I think about and I can
wrap my own brain around, right?
Right.
But, uh, this idea ofthe personalization AI.
I think one of the things you willsee in the next two years is we think
we're going to get to somewhere around60 percent of all devices sold by 2026
will be these co pilot plus AI PC.
(22:32):
So, big majority are going to go, but ifyou think of that divide between the 60
percent that have it in the 40 percentthat don't, I think you're going to have
this big, it'll be a digital divide, butit'll be an AI digital divide, right?
So, if your business is trying tocompete with others who are utilizing
AI, you're going to be at a disadvantage.
And so it's one of those things that,uh, you know, I'm urging everyone
(22:54):
start to do your work now, startto dive into this, understand it,
understand the most important toolsthat you use, talk to those ISVs.
Are they going to bemoving their software?
To an NPU.
Are they going to be leveraging AIbecause you don't want to buy devices
now that you won't be able to do thata year from now, two years from now.
Speaker 3 (23:14):
Right?
Ed (23:14):
Um, so, uh, everything that we're hearing from Microsoft, anything that's
deploying right now will have at leasta three plus year lifespan just for
the AI capabilities, but believe us.
Silicon makers are not going to stop.
It's just going to continue toexplode and give us the ability to
do what I think is my second, um.
Uh, premonition is that we're going tosee a whole lot less importance on CPU
(23:37):
and GPU and we're going to see a wholelot more, uh, emphasis on NPU is the
NPUs can run in the hundreds of trillionoperations a 2nd and do so sipping power.
I think we're going to see a wholelot more of our applications.
Learn that they can run on the NPU and doso, so much more efficiently than they can
on CPU and GPU, which will again, reduceour power draw, get better battery life.
(24:00):
All of those things thatwe hope for out of devices.
I think that's going to take us tillthe end of the decade, but, uh, you will
see a whole lot more importance there.
So if you haven't donesome research around NPUs.
So you know, and understand whythey're going to be so important.
Speaker 3 (24:15):
Yeah.
The end of the decade is only,it's only five years away.
So, uh, not, not too long.
Ed (24:20):
Um, I made that prediction.
I figure I've got five moreyears left in this industry.
Had
Speaker 3 (24:26):
my hair will
Ed (24:27):
be gone
Speaker 3 (24:29):
as you know, from reading the analysts predictions, nobody
actually follows up on those, right?
So like,
we'll sunset this podcast three yearsfrom now, you have nothing to worry about.
Ed (24:40):
Thanks, dad.
I appreciate that.
Speaker 3 (24:41):
Absolutely.
So for those who are not luckyenough to sit next to you on a
plane or at the airport Chili's,you know, 200 days out of the year.
Like, where can listeners go to learnmore about your work and and, uh,
you know, on device AI technology?
Ed (24:56):
Yeah, our website does a lot of great, uh, or provides a
whole lot of great information.
Microsoft has a lot of greatinformation on copilot plus.
So I'd urge you to search Microsoftcopilot plus and find some of the
training and research they have.
If you want to go to our website,it's Qualcomm with two M's dot
com slash AI and we have a hugeamount of data available, right?
(25:19):
To go learn more.
Plus we are adding all sortsof content into portals.
So that you have accessto all of that at SHI.
So you can reach out directly to yourSHI representative and we can ensure
that you get that information as well.
Speaker 3 (25:34):
Excellent.
Well, we're gonna have to leave it there.
So, as we continue to push the boundariesof what's possible, it's clear that on
device AI will play a critical role.
In shaping the next era of computing,whether it's powering personalized
experiences or streamliningindustrial operations, the
potential is limitless on device.
AI isn't just the future.
It's already here, changingthe way we interact with
(25:56):
technology in our everyday lives.
A huge thank you to Mike Belcher forsharing his insights and expertise.
And until next time, keepinnovating and stay ahead of the
curve for innovation heroes and S.
H.
I.
I'm Ed McNamara.
We'll see you again in two weeks,
Ad Spot (26:15):
ready to elevate your work.
The Snapdragon XE processor powers throughdemanding tasks with cutting edge AI up
to 37% less power usage and features likelive translations and enhanced search.
Focus on big ideas whileyour laptop handles the rest.
Learn more at s i.com/partners/snapdragon.
Ready to elevate your work?
The Snapdragon X Elite processorpowers through demanding tasks with
cutting edge AI, up to 37 percentless power usage, and features like
live translations and enhanced search.
Focus on big ideas whileyour laptop handles the rest.
Learn more at shi.
com slash partners slash snapdragon.
Mike Belcher (00:22):
About 2033, we would actually hit a point where it would
cost more than the GDP of the world torun the internet at the current growth
rates that the internet was growing.
So we have less than a decade tofigure this out and find better ways
to use AI and do so power efficiently.
Um, and honestly, I think, uh, ondevice and endpoint AI is where we're
(00:44):
going to have to migrate the majorityof the work that we do every day.
Ed (00:48):
In the world of technology.
Heroes are everywhere.
They're overcoming disruption, deliveringsustainable outcomes, and fearlessly
forging the future to solve what's next.
Join me, Ed McNamara, as we meet thepeople and businesses driving change
in our constantly disruptive world.
This is Innovation Heroes, apodcast brought to you by SHI.
(01:14):
When you think of the AI you interactwith on a daily basis, what comes to mind?
Chat GPT, Amazon, Google.
What do they all have in common?
They all live in the cloud, but thetechnology world is shifting gears
and a new player is taking stage.
On device AI.
In a time where speed, personalization,security, and efficiency define
(01:34):
success, this game changingtechnology is rewriting the rules
for business and consumers alike.
By processing data locally ondevices, on device AI eliminates
latency issues, enhances performance.
Privacy and enables real time decisionmaking opening doors to innovation
like never before from transformingindustries like healthcare and retail to
(01:55):
powering smarter IOT systems on device.
AI is driving a revolution in howwe interact with technology today.
Mike Belcher, director of North Americancommercial channels at Qualcomm.
We'll help explain why this shift matters,how it's empowering businesses to operate
faster and smarter, and the excitingpossibilities it unlocks for the future.
(02:16):
Mike, welcome to Innovation Heroes.
Oh, Ed, glad to be here.
Thank you for the invite.
Speaker 3 (02:21):
Mike, you've been in IT and computing for a couple decades now, and
I'd love how you got into this space.
Ed (02:28):
Yeah, so, um, you know, I've been in, uh, mostly on the OEM side, uh,
at, uh, Gateway and HP for 25 years,and I've been here now, uh, almost, uh,
three, um, and, and I didn't start offmy career there, uh, but I've been in
tech, uh, since I was in high school, andbefore there was anything called Windows,
(02:48):
and, uh, so, uh, I, I was very lucky.
My mom actually, uh, never went tocollege, but she self taught herself
PCs as they were coming out, so Iwas the first 286s and 386s came out.
Uh, she taught herself computer language,taught me while I was in high school,
and, uh, kind of created a love affairwith the computer with me ever since.
And so I was always,uh, engaged, involved.
(03:11):
Uh, we were doing, I think, someof the first online banking.
My mom worked for Bank of America atthe time, and she was one of the first
guinea pigs of actually doing that,you know, for those who remember,
we would log into the internet, right?
So today we're talking about ondevice AI, um, which seems like
a new idea, but it really isn't.
(03:31):
Um, before the internet, um, allAI was done on device, right?
And can you tell me a little bit aboutthat and what are some early examples?
Mike Belcher (03:40):
Yeah, maybe I'll give some of the examples that,
um, that that I have personally.
Right?
So, uh, in past life, I was responsiblefor looking at emerging technologies
when I was at HP, um, really so that wecould help decipher and where we believe
that technology would would end up.
Right?
And so.
How would we help higher educationinstitutions, for example, one of
(04:01):
them revolved around the amountof research that's being done at
these are one institutions, right?
There is a ton and there's so muchdata that that they're collecting
and there weren't really greatways to analyze that data locally.
And so some of what we starteddoing at HP, we're building these
super high performance workstations.
And so imagine a tower device.
(04:23):
We started working with NVIDIA very early.
In that stage to develop thecapabilities and software to actually
drive the GPUs to process this dataand do that at a localized level.
And so what it allowed to do via fundingthat these institutions would get, they
could buy these 50, 60, 70, 000 PCs andthen use those continually over and over
(04:46):
again for all of their data research.
Right.
And, and mine those, uh, thosenumbers, uh, really come, come
together with better answers, right.
For, for the research they were doing.
But do so in a very safe way, alllocalized and not having to go to the
cloud for the, for, for that data.
So it was one of those kind ofreally unique cases that in me
went, Oh my gosh, there's, there'shuge opportunity for this, right?
Ed (05:08):
Yeah, the, the, the, the role of higher ed in those early days,
it's just cannot be overstated.
And then cloud computingbecame a big thing.
So, um, you know, why did we makethe switch to cloud based AI and,
and what advantages did that providefor businesses and consumers?
Yeah.
So, you know, I, I think the, the,the big turning point is you, you
already alluded to as chat GPT, right?
(05:29):
And when, when that came outwith generative AI and you
could ask any question and get.
Some pretty darn good answers,and even if they weren't good
answers, they sounded good.
So some of that work youjust had to do there.
We couldn't afford to put allthese devices at 50, 60, 70,
000 on everybody's desk, right?
Um, and so the ability for us to takeadvantage of the compute power to take
(05:53):
advantage of the data up in the cloud.
Really became the only way to makethis effective and particularly
where we wanted to ask these reallycomplex, really large questions where
you needed these huge data sets.
We really had to do that in the cloudand the early days of us cutting our
teeth on device, I think, really led theway for them to start to realize, oh,
my gosh, we could do this in the cloud.
(06:14):
In fact, it may be the only waywe could do AI very quickly.
Yeah.
Speaker 3 (06:19):
Right.
So, so now we're migrating back to,um, on device AI and to start, what
are a few examples of on device AI thatexists today and why is the tech world
making the switch from, from cloud,uh, to on, on device for, for AI?
Ed (06:33):
One of the reasons why we have to move from cloud to, uh, to, to endpoint
devices is the cost of what it takes todrive that AI up in the cloud, right?
And so it is very costly for all theAI providers right now, every inquiry
will cost you, you know, between5 cents and 60 cents, depending
on what it is that you're doing.
(06:54):
If you're like me, you aska lot of questions of AI.
So I'm one of those folks driving thecosts up for all of these providers.
But there's so much of this AIthat you don't have to, you don't
need those ginormous data sets.
You don't need these enormouscomputers that require lots of
cooling to keep them operational.
And honestly, the advent ofNPUs or neural processing units
(07:17):
has really led that way, right?
Of how do we do AI and withtechnology that was designed for AI?
And that's exactly what, uh,neural processing units are.
We started development of those 12 plusyears ago and actually deploying them.
If you've got a Samsung phone withone of our chips in it, that's, you
know, five years older or less, youprobably have a pretty powerful NPU
(07:38):
in it and don't even realize it.
Right?
So we've been deploying NPUsfor a number of reasons.
They are so efficient.
And so if you think about what it takesto drive these giant GPUs and power them
and cool them up in the cloud versus whatit takes to run a neural processing unit.
We can run 45 trillion operations persecond on an NPU and do it all day running
(08:00):
less than 300 milliwatts of power, right?
It just sips power to, uh,to, to, to drive itself.
And if we did that same thingwith a GPU, we'd be using
probably 60 or 70 watts of power.
So it's about 150 times moreefficient to do it on the device.
And, uh, if you have a Very largedata center in your neighborhood, your
(08:21):
hometown, because they are sucking thepower out of the grid right now, right?
To run these and to cool them.
And there's a huge concern right nowaround how do we power our homes?
How do we light and heat ourhomes when these data centers
are sucking so much power?
Um, we've actually done some, uh,some, some great research around this.
And by 2040, we would actually hita point where it would cost more
(08:46):
than the GDP of the world to runthe internet at the current growth
rates that the internet was growing.
That was before chat GPT, before theNVIDIA revolution up in the cloud.
Um, we did that same, uh, data research,uh, earlier this spring, like in March.
And we factored that we're going toreach that same level about 2033.
(09:07):
So we have less than a decade tofigure this out and find better ways
to use AI and do so power efficiently.
Um, and honestly, I think, uh, ondevice and endpoint AI is where we're
going to have to migrate the majorityof the work that we do every day.
We started working with Microsoft becauseof our background in this, in the, in
the cellular world of how could we takea much more power efficient model like
(09:30):
ARM, um, and in ARM instruction sets.
And actually, uh, drive windowson arm with a, with a risk
processor, but also embed NPUs.
So now Microsoft could completely redesigntheir OS to take advantage of that NPU.
Now it requires a lot ofchange to make that happen.
(09:50):
A whole lot of improvementsinto the systems.
But boy, the vision for that reallystarted about four years ago, and we're
already starting to see some reallycool things like, uh, if you haven't
seen, uh, in Visual Studio Code,there's, uh, an app called CodeGen.
You basically type in samples of yourcode and then type in a request of
what you want us to do, and it willgo write the code for you, right?
(10:11):
And so you think of what takes days todo, you can do in minutes now, right?
Right.
And so the ability for us to leverage AIto do these things that it does really,
really well is exactly what we can do ondevice and then not worry about the cloud.
This code that might be very discreetand you don't want out in the ether,
it's all self contained on your device.
Speaker 3 (10:32):
And I think our audience, you know, which is a pretty technical
audience, can, can, um, you know,understand the benefits of on device AI.
But, uh, you know, how,how do you describe it?
Ed (10:41):
Yeah, so the, the ability to have on device, and particularly the way
that Microsoft designed Copilot Plus, Isthat it's self contained on the device.
It's designed to operate, not beingconnected to the Internet at all.
Right?
And honestly, uh, it's incrediblysafe and secure that way.
Um, one of the things that they wantedto use AI for, and they needed these
powerful NPUs to be able to drive it, wasin the background AI running all the time.
(11:05):
So when you're runninga Copilot Plus device.
The AI is actually running all the timeand it's learning you, it's, it's seeing
what you do, what, what applicationsyou touch, uh, how you write, how
you communicate all of those things.
It's learning how you operate.
And so I'll give you a really kindof cool example right now for myself.
I travel a ton, way toomuch, 200 plus days a year.
(11:28):
And so I am continually behind on email.
And so when I finally got my device, uh,and now we've, we've set up a co pilot,
uh, M three 65, along with our, ourdevice set, um, I've started running, um.
All of my officeapplications utilizing the A.
I.
And so for outlook.
Now, I will ask inquiries when I comeback from a long trip and say, please
(11:50):
find me all of the last 10 days.
Every email that's come through.
That's asking for a response forme and prioritize them for me.
And so in seconds, I get this list ofwhat I need to go hit right away and I
go hit those and then I do other things.
Right?
So I can actually prioritize my life.
Accomplish more instead of do moreby plowing through all this email
(12:12):
that is probably just looking at it.
You don't do anything with it, right?
I just can't ignore those.
So there's there's these types of thingsthat the personalization of the device.
Um, what will enable that?
Honestly, Windows is the only playerin this space, and we're the only
technology that can drive that.
Currently,
Speaker 3 (12:28):
I won't ask you to set the batting order for the people that
you get to, but you know, in youremail inbox before before others,
because we don't know who's listening.
But when you talk about copilotplus, what do you see is the biggest
differences than the original copilot?
Like how?
How will users use it differently?
Ed (12:44):
Yeah, a couple of unique ways.
First of all, thispersonalization of the device.
I think a year from now, those of usthat deploy this, uh, you know, in
this year, it'll take about a year.
But after you, you start utilizingthis after a year, you're not
going to be able to do without it.
Because the ability to use thingslike recall and find things in
seconds, the ability to havethese, uh, capabilities like, yeah.
(13:07):
Um, you know, uh, organizing yourdata or comparing spreadsheets for
you automatically, right on thedevice, not having to use the cloud
to do that, not worrying about whatdata gets set up into the cloud.
I think those are just going tobecome so embedded and part of what
we do, you won't be able to go awayfrom a device like this, right?
Because you're not going to be nearlyas productive and you certainly
(13:28):
won't be accomplishing more.
So I think long term, that'sdirectly what will happen.
So we're at the earlyinfancy stages of that.
But we're going to see all kinds ofapplications now take advantage of
these NPUs to completely rethink howyou use these applications and I'll
give you a one kind of a cool example.
Yeah.
So Adobe is one of these players, right?
(13:49):
Who, I don't know about you,I've tried to learn Photoshop.
I don't have the time to devote to it andto stick with it to be really good at it.
Right?
And so, you know, for them, the abilityto have almost every user be able
to use Photoshop comes down to AI.
So they've redesigned Photoshop to runnatively on arm and have now, uh, in
(14:09):
process of developing all the AI featuresrunning on our NPU that you can just
say, hey, Photoshop, Take the red eyeout of this picture and you can say it.
You can type it.
You can circle a blemish andsay, remove the blemish, right?
And you're done.
Um, so the ability for us to learnhow to use these tools really
comes down to how do we communicateand using prompts, et cetera.
(14:31):
And so you're going to see allsorts of tools like this come out.
If it's productivity, if it's security.
Everyone will be racing on howdo I leverage the NPU to provide
these really great AI capabilities?
Speaker 3 (14:43):
Yeah, i'm on a marketing team and I have about seven or ten
graphic designers who ensured thatI would never ever learn photoshop
after a few early attempts and theywere like Ed, we've got this for you.
Can we, why don't you put that down,you know, but everything that I hear
them say is that it's just, you know,their, their, their efficiency and
(15:03):
productivity is like going throughthe roof, you know, for exactly the
features that, that you're using.
Um, it, you talked alot about the benefits.
Um, like what aboutchallenges or drawbacks?
You know, what, what, what are the limitsor, or what do we, what do we still
have to be mindful of at this point?
Ed (15:20):
Yeah, so believe it or not, the requirements for
copilot plus are pretty low.
It's like 16 gig 256 on the on thehard drive and ideal is probably like
a half a gig hard drive and 32 gigs.
If you're going to do some heavier dutywork on the device, but if you have large
data sets that you really need to runbigger than about 13 billion parameters,
(15:42):
that's what we can run really effectively.
Um, with our 45 tops NPU, if yourdata sets need to be that large,
that's the kind of stuff we'regoing to have to do in the cloud.
But 30, 13 billion parametersis still pretty darn large.
And so, uh, you know, what?
1 of the things that I thinkwe'll start to see is where
we need that large data set.
We need that super huge compute power.
We're going to do that stuff in thecloud and we'll continue to do so.
(16:05):
But almost everything else.
I think we'll start to migrateto the device because of that
efficiency and cost savings.
And we're going to see a lot moreof, hey, I can download, uh, an SLM,
a small language model instead ofthese large language models, right?
And customize it based on my business.
If I'm, uh, if I'm, uh, uh, an end user,and it may be putting my schematics
(16:26):
and my if, when scenarios when our.
Field tech repairs out, right?
Um, so that the AI can do that for you.
You don't even have to beconnected to the Internet.
The device will just do that foryou while you're out on site.
Um, I think you're gonna see a lot moreof that kind of use case with these SLMs.
Really taking the place of, of what'shappening in the cloud customized
to you and your own use caseinside of your, your organization.
Speaker 3 (16:50):
So I, I, whenever I get to talk to somebody like you who really
sees how these technology is deployed,um, have you had something in the last
like 12 to 18 months where, um, you'vetaken a look, you're like, wait, they're
using this where and how, like, what,what would you say to that question?
Ed (17:06):
God, so, so many.
Um, honestly, I, I think, um, youknow, when I was at HP, we started
using some AI tools to help get tothose zero day attacks, et cetera.
I think what you're going to hear comingout forward and, and how folks, you've
got folks like Sophos and CrowdStrike thatare building for the NPU right now, all
(17:27):
sorts of amazing features and securitythat we just haven't thought of, right?
And part of it is you can use yourAI to help you design this and
help you put scenarios in place ofhow are you going to block this?
How do you develop the softwareto make all of that work?
Um, I think you're going to see somereally cool use cases for how we
use AI in security in particular.
Now you have to have some,some pieces embedded, right?
(17:49):
And running on your, your CPU there.
But I think you're going to start tosee the NPU, which typically operates
about seven times faster than a GPU does.
You're going to start to see a lot ofthat security feature set embedded.
Offload onto the NPU.
It's so efficient, right?
So you don't draw a lot of power.
You're going to draw it on the GPU.
You might just suck the battery outof your device or your fan would
(18:10):
never go off while you are running it.
Right?
And so the ability for us to do thatlocally is a huge boon for security.
And as we all know.
The black hats keep gettingmore and more efficient.
They're going to be using AI,figuring out how to attack us, right?
They
Speaker 3 (18:26):
have it too.
Ed (18:27):
Absolutely.
And so if we think about, right,like, uh, you know, Windows 10
support goes away in October.
I would be really fearful if Iwas, um, an IT decision maker.
Not planning on getting off of Windows10 by October, because guess what?
Those AI black hats will be doing.
They're going to be targeting you.
And so looking for ways to moveforward just on a security basis.
(18:50):
I think it's going to bereally, really important.
And then the ability to use theseAI security features to combat.
The a high black half as they'rethey're going after your business.
Ad Spot (19:01):
Are you ready to take your work to the next level?
The Snapdragon X Elite processoris here to redefine productivity
and power for modern professionals.
With cutting edge AI capabilities, ithandles the most demanding tasks with
up to 37 percent less power usage,so you'll stay productive all day
without worrying about your battery.
Snapdragon X Elite powered laptopsalso bring AI driven features like live
(19:24):
translations and enhance windows search,making multitasking easier than ever.
Let your laptop take care of the busy workwhile you focus on bringing your business
changing ideas to life even faster.
To learn more about the SnapdragonX Elite processor, go to www.
shi.
com slash partners slash snapdragon.
Speaker 3 (19:47):
You know, from a, from a governance perspective, does
having it, um, you know, on device.
Does that make it easier, easierto manage that it's on device just
for obvious reasons, or is thereanything that might be overlooking?
Ed (19:59):
Yeah, it really then boils down to how secure is your device, right?
And so if your users are accessing anyof your sensitive information now on
their devices, your device is as secureand your data is as secure as that is.
Right.
I think I love the way thatMicrosoft designed Copilot Plus to
run encapsulated on your device toeliminate that, that challenge, right?
(20:20):
It was one of those big fears.
It was a big fear I had.
You know, four or five years ago, aswe started looking at chat GPT, Oh, my
God, how do you keep your data safe?
How do you make those things happen?
Right?
And so the ability for usto do this self contained.
And honestly, they did it as muchfor security as they did for the
personalization of the device.
So this AI running all the time,learning you, learning how you operate,
(20:43):
helping you find files, et cetera.
All stay self contained on the device.
And so, as long as your device issecure, all of that is secure, right?
So, this now starts to reallyboil down into how do I ensure
that my end user devices areas safe and secure as possible?
1 of the cool things about windows11 is Microsoft has changed some of
the mandatory things that happen.
(21:04):
And honestly, part of it was inconjunction with us as we were developing.
Our chip sets for them.
We actually developed a securityprocessing unit that is 1 of those
processes along with an NPU andCPU GPU that is loaded with Pluton.
And so now it gives Microsoft abilityto have control over those devices.
Um, I was traveling in the Bostonarea a few months back when a major
(21:27):
malfunction occurred when somebodydid an update and it knocked down all
slew of PCs across the nation, theairlines all went down, et cetera.
I was trapped.
And, uh, you know, so you see it kindof firsthand and part of that was
Microsoft didn't have control over beingable to get those machines booted up.
They had to rely on a third party.
This will now eliminate some of that, uh,sort of challenge and make the devices
(21:48):
much more secure right out of the box.
So that there's some really kind of uniquethings that are happening as Microsoft
has developed this operating systemthat improve almost every aspect, right?
Not only just your user experience,but your security as well.
Speaker 3 (22:01):
So forward looking one of our last questions, you know, um, I always, I
always put way too much time on this year.
I was going to say, what canwe expect this technology
in the next 10 years or so?
I'll say, how about the next two years?
Ed (22:12):
Yeah, I'll give you a couple that I think about and I can
wrap my own brain around, right?
Right.
But, uh, this idea ofthe personalization AI.
I think one of the things you willsee in the next two years is we think
we're going to get to somewhere around60 percent of all devices sold by 2026
will be these co pilot plus AI PC.
(22:32):
So, big majority are going to go, but ifyou think of that divide between the 60
percent that have it in the 40 percentthat don't, I think you're going to have
this big, it'll be a digital divide, butit'll be an AI digital divide, right?
So, if your business is trying tocompete with others who are utilizing
AI, you're going to be at a disadvantage.
And so it's one of those things that,uh, you know, I'm urging everyone
(22:54):
start to do your work now, startto dive into this, understand it,
understand the most important toolsthat you use, talk to those ISVs.
Are they going to bemoving their software?
To an NPU.
Are they going to be leveraging AIbecause you don't want to buy devices
now that you won't be able to do thata year from now, two years from now.
Speaker 3 (23:14):
Right?
Ed (23:14):
Um, so, uh, everything that we're hearing from Microsoft, anything that's
deploying right now will have at leasta three plus year lifespan just for
the AI capabilities, but believe us.
Silicon makers are not going to stop.
It's just going to continue toexplode and give us the ability to
do what I think is my second, um.
Uh, premonition is that we're going tosee a whole lot less importance on CPU
(23:37):
and GPU and we're going to see a wholelot more, uh, emphasis on NPU is the
NPUs can run in the hundreds of trillionoperations a 2nd and do so sipping power.
I think we're going to see a wholelot more of our applications.
Learn that they can run on the NPU and doso, so much more efficiently than they can
on CPU and GPU, which will again, reduceour power draw, get better battery life.
(24:00):
All of those things thatwe hope for out of devices.
I think that's going to take us tillthe end of the decade, but, uh, you will
see a whole lot more importance there.
So if you haven't donesome research around NPUs.
So you know, and understand whythey're going to be so important.
Speaker 3 (24:15):
Yeah.
The end of the decade is only,it's only five years away.
So, uh, not, not too long.
Ed (24:20):
Um, I made that prediction.
I figure I've got five moreyears left in this industry.
Had
Speaker 3 (24:26):
my hair will
Ed (24:27):
be gone
Speaker 3 (24:29):
as you know, from reading the analysts predictions, nobody
actually follows up on those, right?
So like,
we'll sunset this podcast three yearsfrom now, you have nothing to worry about.
Ed (24:40):
Thanks, dad.
I appreciate that.
Speaker 3 (24:41):
Absolutely.
So for those who are not luckyenough to sit next to you on a
plane or at the airport Chili's,you know, 200 days out of the year.
Like, where can listeners go to learnmore about your work and and, uh,
you know, on device AI technology?
Ed (24:56):
Yeah, our website does a lot of great, uh, or provides a
whole lot of great information.
Microsoft has a lot of greatinformation on copilot plus.
So I'd urge you to search Microsoftcopilot plus and find some of the
training and research they have.
If you want to go to our website,it's Qualcomm with two M's dot
com slash AI and we have a hugeamount of data available, right?
(25:19):
To go learn more.
Plus we are adding all sortsof content into portals.
So that you have accessto all of that at SHI.
So you can reach out directly to yourSHI representative and we can ensure
that you get that information as well.
Speaker 3 (25:34):
Excellent.
Well, we're gonna have to leave it there.
So, as we continue to push the boundariesof what's possible, it's clear that on
device AI will play a critical role.
In shaping the next era of computing,whether it's powering personalized
experiences or streamliningindustrial operations, the
potential is limitless on device.
AI isn't just the future.
It's already here, changingthe way we interact with
(25:56):
technology in our everyday lives.
A huge thank you to Mike Belcher forsharing his insights and expertise.
And until next time, keepinnovating and stay ahead of the
curve for innovation heroes and S.
H.
I.
I'm Ed McNamara.
We'll see you again in two weeks,
Ad Spot (26:15):
ready to elevate your work.
The Snapdragon XE processor powers throughdemanding tasks with cutting edge AI up
to 37% less power usage and features likelive translations and enhanced search.
Focus on big ideas whileyour laptop handles the rest.
Learn more at s i.com/partners/snapdragon.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.