The Role of AI in Networking

Introduction

0:00 [Music]

0:11 thank you

0:12 God I hate falling sweet here

0:15 welcome everybody we're so excited to be

0:17 kicking off our second annual and action

0:20 Vegas and for those that were here both

0:22 years thank you for coming back I was

0:24 really stoked when I was asked to kick

0:26 this off but I was also honestly a

0:29 little stumped right how do I get you

0:31 guys as excited as that how do I explain

0:35 what we're gonna have coming in terms of

0:37 how to make AI real right or how to use

0:40 AI how to differentiate with AI how to

0:42 get real outcomes right there's a lot I

0:44 wanted to do but I didn't know how to

0:46 start so I actually called sudir and I

0:49 said so dear how would you do this this

0:50 presentation but I think most you guys

0:52 know Cedar has like no opinions on

0:54 anything so didn't get me anywhere I was

0:56 I was I was lost so I called Bob Friday

0:58 our chief AI officer and you guys know

1:00 this engineering is from Venus marketing

1:02 is from Mars

1:03 didn't really understand what the hell

1:05 he was talking about either so hopefully

1:06 you guys can understand him later so but

1:08 then I realized you mentioned sheriff in

1:10 town there is actually a new sheriff in

1:11 town that's better than all of us so I

1:13 actually booted up my computer and I

1:16 asked chat GPT

1:18 how the hell would you do this thing if

1:20 you know if you were in my shoes it

1:22 actually has some pretty good advice you

1:23 know the first thing is said is start

1:25 with a thought-provoking stat okay you

1:26 know that's like presentation 101 so

1:29 let's let's Okay chat gbt Overlord what

1:31 do you got first here what's a good stat

1:32 so 11 billion dollars in two years AI

1:36 Ops will be 11 billion dollar market

1:37 with a 25 kagger that's pretty cool but

1:40 I'm guessing you guys probably don't

1:42 care about stats and you probably

1:43 already know AI Ops is crushing it

1:45 otherwise you wouldn't be at AI in

1:47 action Vegas to learn more about AI Ops

1:49 so come on chat GPT you got to give

1:51 something a little bit more than that

1:53 all right cool share a personal story I

1:55 like that I can do that

1:57 um the one that immediately comes to

1:59 mind which you're also going to hear

The History of Mist

2:00 from from Tom Wilburn as well so I'll

2:02 touch on the personal experience of this

2:04 is the founding of Miss Right Miss was

2:06 actually founded back in 2014 to

2:09 transition the industry from Network

2:11 experiences

2:12 to user experiences right and that's

2:14 what you're going to hear a lot about

2:15 throughout the course of the day

2:17 um and again Tom will walk you through

2:18 that Journey a little bit more but

2:20 there's one story that actually really

2:22 stuck with me when I joined Mist about a

2:23 year or two after they they had first

2:25 started and we're working with one of

2:26 the the fortune 10 retailers who's

2:28 actually in the room right now and they

2:30 actually were telling a story about

2:31 robots believe it or not that was their

2:32 user for experience robots and their

2:35 robots were going around their

2:36 warehouses and fulfillment centers and

2:38 doing what robots do

2:40 and the problem is if one of those

2:41 robots went down it's millions of

2:43 dollars lost every minute it was down

2:45 and they had a great line robots can't

2:47 call for help the robot's not going to

2:49 pick up the phone and call your help

2:50 desk and say the Wi-Fi sucks right so in

2:53 that instance there's a great example of

2:54 you have to really hone in on user

2:56 experiences and optimize the network for

2:58 that so I love that example right or

3:01 another way to put it is as it guys you

3:03 got to be able to take the blinders off

3:04 and transition from looking at just that

3:07 the fact that the aps are up and running

3:08 or the Network's up and running to how

3:10 are the users on that Network having an

3:12 experience here's a funny example of

3:14 that

3:15 that's strange a ton of Wi-Fi service

3:18 tickets just rolled in but I can't see

3:19 any network issues

3:22 well the lights on so it must be fine

3:24 they're saying the Wi-Fi sucks

3:27 but the light is on

3:30 that was an example of some of the ads

3:31 that we run again to get people to focus

3:33 on the fact that AI is real delivering

3:35 real value and focusing on user

3:37 experiences so thanks Chester PT I think

3:39 we got in a good direction there

3:42 um what else this is the last last

What is Happening with AI Right now?

3:44 recommendation that chat GPT has is pose

3:46 a question or or challenge the audience

3:48 and actually I thought this one was

3:49 really interesting right there's so many

3:51 questions you can go down like is it

3:53 real or how do I trust it Etc but I

3:56 thought the more interesting question

3:57 right now for us to talk about in my

3:59 setup and and again folks we'll double

4:01 click on that is

4:03 why now what's happening right now right

4:05 why is it that all sudden llm GPT AI is

4:09 on everyone's tongues right now why are

4:11 there that kids in Little League are

4:12 coding against open Ai and grandmothers

4:15 in Safeway actually are talking about AI

4:16 so what what happened recently that

4:18 changed things so dramatically so for

4:20 those that don't know Chachi PT is

4:23 generative pre-trained Transformer so

4:26 what that means is you're basically

4:27 creating a large language model of a lot

4:29 of data moving into a system using deep

4:31 learning where I can actually prompt it

4:33 and get real human answers back that's

4:35 the marketing definition of this that's

4:36 the best you're going to get from me but

4:38 the general concept here is very easy to

4:40 use it's very creative has a whole host

4:42 of internet data that it can pull from

4:44 so you can do some really cool things

4:46 like for example hypothetically I may

4:49 have sent my wife a Valentine's Day card

4:51 written by Tupac Shakur

4:54 she appreciated it she works in AI loves

4:56 Tupac I wouldn't necessarily recommend

4:57 it but that was an example where chat

4:58 GPT hooked me up right or here's another

5:00 example right

5:02 I asked chat GPT write a haikub at AI

5:05 because I know you guys are all into

5:06 poetry right so

5:08 should I take my shoes off for this one

5:09 Studio because it's so Zen right

5:11 artificial mind learning evolving

5:13 growing future at your door that's Zen

5:16 is right come on

5:18 TV all right

5:20 good stuff

5:21 um but seriously you know there is a

Why a VNA is Still Required Today

5:23 difference in what llm can provide and

5:25 VNA and actually llm is an augmentation

5:27 to the right software architecture the

5:30 right virtual Network assistant and

5:31 that's one of the things we're going to

5:32 talk about right so while an llm is

5:34 great for certain use cases it actually

5:36 is only um it's hard to validate the

5:39 answers is it accurate is it not

5:40 accurate and it's a snapshot of time

5:42 right it's a specific thing so if you

5:45 want to ask you know what is wrong with

5:46 my network today right or what was wrong

5:48 with Jeff's laptop yesterday or show me

5:51 all unhappy users you're not going to be

5:53 able to do that you need a virtual AI

5:55 driven assistant like junipers Marvis

5:59 I heard I was going to be punked

6:03 there we go Marvis meet Marvis also

6:06 known as Mike Bouchon

6:09 he'll be here all night remember to tip

6:11 your waders

6:13 uh but in all seriousness you know

6:14 Marvis our VNA leveraged a whole host of

6:17 AI driven Technologies you know whether

6:19 it's you know natural language

6:20 processing anomaly detection event

6:22 correlation there's a lot you can do

6:24 with a VNA but I think if you guys are

6:28 down the road at Cisco if you're at

6:29 Aruba atmosphere you've probably heard a

6:32 boatload of buzzwords right all those

6:33 words I just mentioned are probably

6:35 words that you've heard before so if

VNA Requirements

6:37 there's anything that you walk away from

6:38 my presentation is that there's five

6:41 things that are needed in aip's VNA

6:43 solution that really make it real and

6:45 again you're going to hear these

6:46 reinforced by your customers by Bob by

6:48 sudir by Tom as we go down the path but

6:50 the first thing is data bad data in bad

6:55 data out you need the right data to

6:57 build an AI model and again that's why

6:59 back in the days when when Bob and sujay

7:02 hajela created Juniper they rewrote the

7:05 control plane to collect 150 user States

7:07 that's why Juniper acquired Miss because

7:09 now you can bring a Telemetry data from

7:11 routers from switches from firewalls to

7:13 complete the picture so it's all about

7:14 not just the quantity of the data but

7:17 the quality of the data okay secondly

7:19 that has to go somewhere where you can

7:21 process it I am sorry with all due

7:23 respect to the guys that rhyme with

7:25 misco you cannot do AI on premises

7:28 that's not how Amazon does it that's not

7:30 how Google does it that's not how you

7:32 know the world of AI does it you need a

7:34 modern microservices Cloud for agility

7:36 for scale resiliency that's what we

7:39 built from scratch that's what the whole

7:40 Juniper portfolio is leveraging together

7:42 in a unified front

7:45 you need Rich data science I was

7:47 standing up just like this at an event

7:48 not too long ago a field day event and I

7:51 said we were the first to do AI driven

7:52 operations the first to do age of an I.T

7:54 and someone chimed in and said does

7:57 being first really matter you know we

7:58 hear vendors say that all the time and

8:00 my answer is with AI yeah it actually

8:03 does right because the algorithms get

8:05 better there's less false positives you

8:07 add more algorithms to do things like

8:09 service levels or find persistently

8:11 failing clients so actually having a

8:13 Seventh Generation AI algorithm that you

8:15 can now add things like yellow and on

8:17 top of it absolutely matters and that's

8:19 one of the things that we've done

8:20 the fourth pillar is continuous learning

8:23 again Bob Friday compliments to him

8:25 always says this if you want to

8:26 understand if your vendor is using real

8:28 AI ask them where their support

8:30 organization assists with respect to

8:32 their data science Engineers are they

8:34 taking the information back from

8:35 customers like these guys and putting

8:37 you back into the system to learn better

8:38 that's what Juniper is doing that's

8:40 continuous learning pulling in data from

8:42 applications like Zoom that's how you

8:44 build a real VNA

8:45 and then lastly we kind of talked about

8:47 this the full stack portfolio you can't

8:49 do AI in a silo if a team's call goes

8:51 down is it the wireless is it the wired

8:53 is it the WAN is it is it teams is it

8:56 the cloud you need the ability to

8:58 correlate all that information and do it

8:59 on their full stack portfolio so those

9:01 are some of the requirements of a VNA

Driving Sustainability with Innovation and AIOps

9:04 I think what goes into a VNA though may

9:07 be less interesting than what comes out

9:08 of a VNA right what are the actual

9:10 outcomes and again we're going to talk a

9:12 lot about that through this show whether

9:14 it's you know just lower cost savings

9:15 through better Opex whether it's just

9:17 better patient experiences when someone

9:19 shows up at a VA hospital where there's

9:21 better student experience like the folks

9:23 at Dartmouth who just don't want people

9:24 tweeting that the Wi-Fi in the dorm room

9:26 sucks and I can't get at my Xbox running

9:29 right so there's a variety of different

9:31 reasons that you want to leverage a VNA

9:33 and a Ops to deliver Better Business

9:34 outcomes and they're all here and you're

9:36 going to hear from this more Tom's going

9:37 to drill into this as well as the

9:39 customers

9:41 and also an interesting one is about

9:42 sustainability right if the Gap can

9:44 reduce 85 percent of their truck rolls

9:46 to their stores that saves not only I.T

9:48 time but that saves gas and Emissions

9:51 right our friends of service now right

9:53 if they can eliminate which they did 128

9:56 000 metric tons of their carbon

9:58 footprint that's not shabby right and

10:00 that's what comes from again being able

10:01 to do more Automation and more insight

10:03 from your network

10:05 so

What’s Next?/Closing Statements

10:06 enough with chat gbt we have lots of

10:08 human experts here to talk about that as

10:11 we said we got uh sudir mentioned we got

10:13 servicenow we got University of

10:14 Washington HEB ratty children's Lobos

10:17 granted talk about all the different

10:19 ways that they're deploying it and using

10:21 it within their networks we've got Tom

10:22 Wilburn and suder Mata head of sales

10:24 head of products for the a German

10:26 Enterprise they're going to talk about

10:27 the campus Branch we got my good friend

10:29 bhushan Mike Bouchon who your time will

10:32 come up just wait he's going to talk

10:34 about Data Center and we got kicked out

10:36 and talk about security so with all due

10:38 respect

10:39 we're done with you

10:42 really all right we don't need AI driven

10:45 photos I think we're good so with that

10:47 you know Tom why don't we get you on

10:48 sale or should I say Anderson Cooper

10:49 because AI I think makes you think

10:51 thinks you look like Anderson Cooper no

10:52 who thinks Bob looks like Anderson

10:54 Cooper in that picture

10:55 there we go all right

10:57 or top about Tom not bomb sorry so Tom

11:00 come on up stage

11:03 [Music]